I was in Bengaluru at 1am on a Tuesday, three tabs deep into reinsurer outreach prep, when the IUMI EYE Issue 52 PDF landed. IUMI is the International Union of Marine Insurance. Their quarterly newsletter is written by and for the people who insure the world’s ships, cargo, and offshore energy infrastructure. Underwriters, surveyors, maritime lawyers, loss prevention specialists. These are the people who think about what happens when a container catches fire mid-Pacific or a tanker grounds off Turkey carrying soya beans.

I don’t insure ships. I build parametric insurance settlement rails. But the traditional insurance industry’s own publications are the best primary source for why parametric architecture needs to exist. And this issue delivered the data.

$92M to $377M in four years

The centrepiece is a special feature on “fake carriers.” IUMI and TAPA EMEA issued a joint warning: criminals are creating shell companies, cloning legitimate firms with stolen credentials, forging insurance certificates, and using generative AI to scale all of it.

The data behind the warning:

TAPA’s intelligence system recorded nearly 160,000 cargo-related crimes across 129 countries between 2022 and 2024. The total losses amount to billions of euros. In the US and Canada alone, estimated cargo theft losses hit $725 million in 2025. The average per-incident value rose over 30% year-on-year, even as total incidents dropped 6%.

Good news: Fewer hits.

Bad news: Higher yield per hit.

The adversaries are getting more selective and more capable at the same time. And this is something to worry about.

The industry’s response today? Continuous vetting of carriers and drivers. Deeper verification of contacts and documentation. Enable multi-factor authentication on freight exchange platforms. Implement more secure ID cards and scanners to read them. Cost, cost, cost and more cost.

Every single countermeasure proposed above is a human-verification step bolted onto an existing process. More checking of people. More scrutiny of people. More people examining more documents. Against unknown adversaries who are using genAI and/or LLMs to probably generate those documents.

The unsaid structural problem

The entire claims-based insurance model depends on documents being trustworthy. A “claim” requires a claimant to submit evidence. An adjuster evaluates that evidence. A decision is made. And then the money moves.

Every link in that chain is an attack surface. From the claimant’s identity, to the documentation, to the adjuster’s judgment, the communication channel; every layer is open to attack. Fake carriers exploit all four simultaneously.

This is what I call a need for the elimination thesis: parametric insurance doesn’t compress the claims adjudication process. It removes it.

When a payout is triggered by an external oracle (flight tracking data, weather stations, IoT sensors, port telemetry), there is no claim to file. No document to forge. No adjuster to deceive. No identity to spoof. The payout trigger is a fact about the world, not a narrative constructed by the claimant.

And every year that AI makes document forgery cheaper, the elimination thesis gets stronger. The fraud arms race is asymmetric in the wrong direction for traditional models, and symmetric in the right direction for parametric ones. Oracles don’t get easier to spoof as LLMs improve. Claims documents do.

2.79 out of 5

On Page 25 of the newsletter, there is an IUMI industry survey on digital transformation. The results are quietly devastating.

Only 31% of respondents use AI beyond internal experiments. Self-assessed digital maturity averages 2.79 on a five-point scale. The top barriers to AI adoption aren’t workforce resistance or training gaps. They’re legacy systems (22%) and data quality (26%). Core system modernisation doesn’t even top the priority list, despite being the prerequisite for everything else the industry projects and says it wants.

In the survey authors’ own words: technological ambition appears to be advancing faster than the underlying system architecture that would sustainably support it.

That sentence is the whole story. You can’t digitise a process whose fundamental architecture assumes and is fully reliant on paper-era trust models. You can only replace it.

64% of companies have dedicated internal digital teams, and 89% of those rely purely on internal capabilities. They’re trying to self-perform a paradigm shift. This almost never works. The railroad companies didn’t build the airlines. The postal services didn’t build email. The film studios didn’t build streaming.

The orphan function

There’s one more interesting data point in the report. When asked about digital priorities, over a third of insurers identified underwriting as their top focus. Claims transformation ranked well below.

If you think about it, this is exactly backwards from first principles. Underwriting determines premium income. Claims determines the payout. The technical result, the thing that actually decides profitability, is shaped more by claims than by underwriting.

But claims transformation is messy. It touches operations, legal, customer relationships, regulatory compliance across jurisdictions. Underwriting optimisation is comparatively clean: better data in, better pricing out. So the industry polishes the intake funnel and defers the structural problem to tomorrow.

The gap this creates is specific and exploitable. If you build settlement infrastructure that makes the claims function disappear for a defined product category, you’re solving the problem the industry has decided not to solve internally.

Signals from adjacent verticals

The rest of the newsletter is dense with expansion signals for anyone thinking in parametric terms.

Autonomous vessels. HGK Shipping got the first permit for a remotely controlled hazardous goods tanker in Flanders. German insurers say they’ll cover autonomous ships but demand full operational data access.

Their open questions (eg., who is legally navigating from a Shore Control Centre? how to allocate liability between shipowner and tech provider? how to establish causation in system failures?) are exactly the questions that oracle-based models answer by construction. The telemetry data IS the adjudication. There’s no “establishing causation” when the trigger is a sensor reading.

Supply chain disruption. The US-Israel Iran conflict has pushed container shipping surcharges to $3,000 per forty-foot unit. Maersk and CMA CGM suspended Red Sea routes. India and Pakistan are the most exposed to LNG disruption from the Gulf, and chaos is simmering with hyperlocal worries about shortages. Every supply chain shock creates demand for fast, automatic protection that doesn’t require a three-month claims process.

Port accumulation risk. A Latin American analyst applied Little’s Law (Inventory = Flow × Dwell Time) to port exposure. Colombian ports today hold $580-660M in cargo at baseline, rising to 2.25x under a 10-day disruption. That’s a clean parametric frame: if dwell time exceeds threshold X, payout triggers.

No loss adjuster walks through the port with a clipboard.

The conference theme tells us everything we need to know

IUMI’s annual conference in Rotterdam this September is titled “Anchoring Trust in a Contested World.”

The word “trust” is doing all the load-bearing work here. The system continues to depend on trust in the claimant’s identity, trust in documentation, trust in the adjuster’s judgment. The system’s own global trade body is publishing evidence that each of these trust layers is eroding fast, or becoming less dependable at current costs.

Parametric insurance doesn’t require trust in the claimant. It requires trust in the oracle. That’s a different engineering problem, and a solvable one.

The traditional insurance industry knows the current architecture is under stress. They document it meticulously, every quarter, in professionally typeset PDFs. They just can’t publish PR around the conclusion that follows from their own evidence.

So I will: the claims process, as an institution, has a shelf life. And the countdown is being accelerated by the very AI tools that the industry scores itself 2.79/5 at adopting.

I’m building OrbitCover, parametric insurance settlement infrastructure. We automate payouts using external data triggers and settle via UPI in under four minutes, zero human claims handling. If you’re thinking about what comes after claims, let’s talk.

Here’s a scenario 300 million Indians experience every year.

Your flight gets delayed four hours. You’re staring at the floor at IGI Terminal 3, laptop dying, meeting cancelled, just mildly irritated co-passengers and hyper energetic kids past their bed-time buzzing all around. You have travel insurance. You paid for it. You pull up the policy document and start reading the claims process:

1. File a claim within 48 hours

2. Attach boarding pass, delay certificate from airline, original policy document

3. Wait anywhere from 2 to 15 business days for acknowledgment

4. Insurer may request additional documentation

5. Settlement within 45 days of final document submission

You do the math. Four hours of delay. 2-3 forms to fill. Six to eight weeks of paperwork. For maybe ₹2,000.

Most people don’t file. The ones who do, roughly 40% get rejected on technicalities. The insurer calls this “claims management.” I call it a clogged pipe at best, or scummy at worst.

Pipes, not promises

Let’s borrow a concept from a field that solved a version of this problem sixty years ago. In 1948, Claude Shannon published “A Mathematical Theory of Communication.” He was trying to figure out how much information you could reliably push through a telephone wire. His answer became the channel capacity theorem:

C = B × log₂(1 + S/N)

C is capacity: how many bits per second you can transmit reliably.

B is bandwidth: the range of frequencies the wire can carry.

S/N is the signal-to-noise ratio: how much of what arrives is actual message versus static.

The theorem says something profound: every communication channel has a hard ceiling. No amount of clever engineering lets you exceed it. You can approach the ceiling with better encoding. You cannot break through it.

Shannon was talking about copper wire. But the structure applies to anything that transfers information between two parties.

Including insurance.

Reframing insurance as a channel

Insurance is, at its core, a risk transfer channel. One party (the insured) has a risk they want to offload. Another party (the insurer) is willing to absorb that risk for a price. The “message” being transmitted is:

Every channel has the three Shannon variables.

Bandwidth (B): How many distinct states can the channel carry per unit time?

In traditional indemnity insurance, the answer is shockingly low.

A single claim cycle produces one binary distinction (pay or reject) over weeks or months. That’s roughly 0.001 distinctions per hour. A telephone wire from 1948 could do billions of times better.

Signal (S): How much of the transmitted information is actual, meaningful content?

The “signal” in a claim is simple: what happened, and how much did it cost? This could be communicated in a few hundred bits.

For eg., a flight delay claim is ~50 bits of information: flight number, scheduled time, actual time, delay duration, policy number.

Noise (N): How much interference, ambiguity, and distortion corrupts the message?

This is where the system collapses. The noise sources in traditional insurance claims are extraordinary:

• Document fraud (the insurer doesn’t trust the insured’s evidence)

• Interpretive ambiguity in policy wording (”act of God” means what, exactly?)

• Subjective loss assessment (two or more departments across TPA and insurer acting as adjusters, two different numbers)

• Adversarial incentives (insured inflates, insurer deflects)

• Bureaucratic friction (forms, approvals, escalations, each adding delay and error)

The noise doesn’t just degrade the signal. It overwhelms it. In the flight travel insurance analogy, a simple 50-bit message gets wrapped in thousands(20x) of bits of verification overhead, each of which introduces its own noise.

The capacity equation for risk transfer

Shannon’s formula generalizes beyond telephone wires. For any exchange system:

Effective Capacity = D(t) × log₂(1 + T/N)

where:

• D(t) is distinctions per unit time (how many settlement decisions per hour).

• T is trust between parties.

• N is noise (fraud, ambiguity, friction).

Note that I’ve replaced Shannon’s “signal power” with “trust.” This isn’t a randomly defined, loose metaphor. It is a structural equivalence.

So in a physical channel, signal power determines how much the receiver updates on what arrives. A loud signal cuts through static. A weak signal gets lost. In a social/economic channel, trust does the same job.

A message from a trusted source carries more usable information than the same message from an untrusted source. If I don’t trust you, every message you send is ambiguous to me. I can’t distinguish signal from noise. The channel capacity collapses toward zero.

Low trust equals high noise.

This isn’t a metaphor. This is the mechanism.

Measuring the collapse

Let’s put rough numbers on traditional indemnity insurance as a risk transfer channel.

• D(t) ≈ 0.001 distinctions/hour. ie., one pay/reject decision stretched over weeks. That’s the throughput through the “claims: channel

• T = Low. Both parties assume adversarial intent. The insured expects rejection. The insurer expects inflation.

• N = High. Document fraud, subjective loss assessment, wording ambiguity, bureaucratic friction. Noise everywhere.

The channel is almost fully clogged, to almost ensure it doesn’t function ideally.

This explains something that has puzzled insurance economists for decades: why does insurance penetration in India sit at ~4% of GDP when the need for risk transfer is enormous? The standard explanations are “lack of awareness” and “affordability.”

Both are wrong, or at least incomplete. Hundreds of millions of Indians know what insurance is. Premiums for basic travel or health cover are within reach of the middle class.

The real technical answer is: the risk transfer channel has such low effective capacity that the cost of using it (in time, friction, and uncertainty) exceeds the expected value of the payout for most small-to-medium losses.

People aren’t unaware. They’re making a very rational calculation based on what they see around: the pipe is too clogged to be worth their time and effort pushing claims through.

This is why informal risk-sharing persists. Family lending, community pooling, or just simply absorbing losses. These channels have lower bandwidth than formal insurance in theory, but in reality; they have much higher trust (family T >> insurer T), which means their effective capacity can actually be higher for small losses.

What changes when you replace the channel

Parametric insurance doesn’t improve the traditional channel. It replaces it entirely. The mechanism is simple:

instead of : “file a claim, prove your loss, wait for assessment”;
the contract says: “if event X occurs as measured by data source Y, payment Z is automatically triggered.”

No need to “raise” a claim. No adjudication. No human in the loop.

The channel variables shift dramatically:

• D(t) jumps from ~0.001/hour to ~3,600/hour. One data/oracle check per second is achievable. That’s a six-order-of-magnitude increase in throughput.

• T decouples from the insurer entirely. It’s now determined by data/oracle trust (more on this below).

• N drops from high and diffuse (fraud, ambiguity, subjectivity) to low and measurable (basis risk + data quality). Two specific, quantifiable noise sources instead of dozens of unmeasurable ones.

Therefore, C_eff goes from near zero to orders of magnitude higher.

And three major things happen when you make this switch.

First, the noise sources change qualitatively: Traditional noise is diffuse and unmeasurable: how do you quantify “interpretive ambiguity in policy wording”?

Parametric noise reduces to exactly two measurable quantities: One is basis risk (the gap between the parametric trigger and the actual loss) and the second is oracle data quality (how accurate is the data feed).

Basis Risk is now a noise engineering problem. Both can now be precisely quantified, tracked, and engineered against. You go from fighting fog to calibrating instruments.

Second, the trust variable decouples from the insurer: In traditional insurance, T(trust) is “do I trust THIS company to pay?”

That’s hard to build, easy to destroy, and non-transferable. In parametric, T is “do I trust the data source?”

Flight delay data from OAG or FlightAware is trusted because airlines themselves consume it. The trust is borrowed from an existing high-trust channel rather than built from scratch. This is not a small point. It means parametric insurance can achieve high T on day one in verticals where trusted oracles already exist.

Third, speed becomes a trust-building mechanism: When a parametric payout lands in your bank account four minutes after your flight delay crosses the threshold, that experience itself raises T for the next interaction. Speed is not just a convenience feature. It’s a trust-formation event.

Each fast, accurate payout increases the T term in the capacity equation, which increases C_eff, which means more people use the channel, which produces more trust-formation events. The flywheel is information-theoretic, not just commercial.

The constraints that remain

If this analysis is right, two hard constraints determine where parametric insurance can and can’t work:

Constraint 1: Data/Oracle availability: The entire system’s trust rides on the data source. For flight delays, high-trust oracles exist (airline schedule data, airport systems). For hotel cancellations or e-commerce delivery failures, they don’t, yet.

“Did my package arrive damaged?” doesn’t have a FlightAware equivalent today. This means vertical expansion is gated by oracle infrastructure, not by market demand or regulatory approval.

Constraint 2: Basis risk as irreducible noise: No parametric trigger perfectly matches every individual’s actual loss. Your flight was delayed 3 hours 50 minutes; the trigger is 4 hours. You got nothing.

This isn’t a bug to be fixed. It’s the thermodynamic floor of the channel: the minimum noise below which you cannot go while maintaining the speed and objectivity that make parametric work. Reducing basis risk to zero means measuring actual individual losses, which means... you’re back to traditional claims adjudication.

The engineering challenge today is to minimize basis risk (N) without reintroducing the noise sources you eliminated. This is an optimization problem with a formal structure:

min E[|L_actual - L_parametric|²]
subject to: no subjective assessment, no human adjudication

Every trigger threshold decision is an instance of this optimization. Set the threshold too low and you overpay (noise from false positives). Set it too high and you underpay (noise from false negatives, which also destroy trust).

The optimum exists, and it’s findable today with existing data.

What this lens predicts

If risk transfer is a channel capacity problem, several things follow:

Distribution channels are trust channels: Where you sell parametric insurance isn’t just a go-to-market decision. Each distribution channel has its own T/N (trust/risk) profile. Embedding in an airline booking flow borrows the airline’s trust (high T).

Selling direct-to-consumer on your own website means building trust from scratch against Indian consumers’ well-calibrated scam detectors (low T, high N).

This predicts a strict sequencing: high-trust embedded channels first, own-brand D2C only after trust has been established externally.

Insurance penetration is a channel capacity metric, not a demand metric: India’s 4% penetration doesn’t mean 96% of the population doesn’t want risk transfer. It means the available channels have such low capacity that most risk transfer demand goes unserved.

Widen the channel (increase C_eff), and penetration rises mechanically.

This reframes the entire insurtech thesis from “how do we sell more policies?” to “how do we increase the effective capacity of the risk transfer channel?”

Whoever certifies the oracle controls the market: If trust in parametric insurance flows from trust in the oracle, then the entity that standardizes oracle certification holds a structural position equivalent to a ratings agency. This role doesn’t exist yet in parametric insurance. It will need to, someone will build it out soon (or someone is already building it).

Reinsurance is a meta-channel. Reinsurers need to trust your triggers, your oracles, your loss ratios before they’ll allocate capital. Their trust formation process has its own bandwidth and noise characteristics.

Established reinsurance brands (Lloyd’s syndicates, Swiss Re) function as trust amplifiers: their endorsement raises T for our entire system at once. Getting a Lloyd’s coverholder stamp isn’t just a commercial milestone. It’s a discontinuous jump in our meta-channel’s capacity.

The general principle

Travel insurance is just one instance. The structure is everywhere.

Any sector or system that transfers value, risk, or information between parties is a channel. Every channel has bandwidth, noise, and capacity constraints. Trust is the social equivalent of signal power: it determines how much information actually gets through.

When you find a channel with near-zero effective capacity despite high demand, you’ve found an opportunity for parametric insurance. The question is always the same: can I build a replacement channel with higher D(t), higher T, and lower N?

Payments had this, twice. In the US, accepting online payments meant merchant accounts, payment gateways, PCI compliance, and weeks of integration. Stripe replaced that entire channel with seven lines of code. The noise (compliance friction, integration complexity, settlement opacity) collapsed. Transaction capacity exploded.

In India, the channel for transferring money was clogged differently (bank visits, cash handling, multi-day settlement). UPI replaced it. Volume went from millions to billions per month. Same structure, different pipe. In both cases, the demand was always there. The channel wasn't.

Risk transfer is having its UPI moment. The demand for loss protection exists. The current channel can’t serve it. Parametric triggers, trusted oracles, and instant settlement rails are the replacement channel.

The math says it should work. The question is where the oracles are good enough to make it real.

I build parametric insurance settlement infrastructure at OrbitCover. If you’re thinking about risk transfer, oracle design, or channel capacity in financial services, I’d like to hear from you.

About every teacher I had growing up in Chennai told me I’d amount to nothing.

The rare ones who didn’t, were few and far between, probably 3 in my whole 12 years of schooling. The messaging of “you’ll amount to nothing” was almost a chorus. Consistent, almost coordinated, like they’d compared notes. And for a long time, I let that chorus be the soundtrack. Then I stopped.

This is not a villain origin story. I coped in the only way I could - singed up at a squeaky old Renga lending library where 10Rs in the 90s would get you a dozen books to carry home and survive the week, of course I couldn’t decide to pick in most cases. So the latest TinTin or Enid Blyton or Archies were out of consideration; high demand, too expensive. The dusty biographies and profile books were cheap. The business and behavioural economics books that rarely saw anyone grab them, were cheap. And so I consumed what I could afford.

I read. I worked on myself from what I could pick and consume. Reading not only helped me escape, it helped me discover, it helped me grow, and more importantly, it introduced me to similar folks who stood out, who were written off, and went onto change the world. I wasn’t worried about being told “You’ll be upto nothing!”. Einstein’s teachers told him the same. Edison and Ben Franklin at early stages of life were told the same. Larry Ellison’s father told him the same damn thing. Steve Jobs was given up for adoption. Andy Grove fled Hungary as a refugee with nothing to his name. I am not borrowing their pain. I had my own. But they made me feel I wasn’t misplaced. That being different wasn’t so bad if you could think, study and channel. I mention them and their early experiences only because there is a specific kind of fuel that comes from being written off early: a fuel that doesn’t expire, doesn’t dilute, and burns cleanest when everything else has run out.

I’ve spent ten years building companies in India. A doctor network. A health data API. Insurance middleware. A parametric flight delay product. On paper, it all looks scattered. Underneath, it’s the same blueprint I’d drawn across different materials: find the missing infrastructure layer, build the rails, make the system work.

This is the first time I’m writing it all down. Not because I’ve figured it out, but because I’m building in public now; not just products, but an evolving version of me. Manifestos aren’t supposed to be finished documents. They’re flags planted in uncertain ground. You plant them so that when you drift, you can look back and see where you meant to stand.

Here are 44 principles I have used to discover and course correct myself to where I am today. They come from Ellison and Bezos, from Jobs and Grove, from Bose and the Collison brothers. I have taken more from Kahneman and Taleb and Cialdini. I have distilled them through my experiences, my learnings, my failure. From the school years in Chennai. From six years of building insurance infrastructure in a country where parametric settlement didn’t exist meaningfully until we built it.

Some of these I believe deeply. Some I’m still testing. That’s the point of this exercise.

On burning boats and the geometry of risk

1. The safest-looking path is the most dangerous one.

Ellison said the bigger the apparent risk, the fewer people will try to go there.

He’s right, but the deeper insight is structural: crowded paths create competition that erodes any meaningful returns to near or absolute zero.

I chose parametric insurance in India; highly regulated, complex, deeply boring to most folks. That complexity is the moat. It keeps the field near empty.

2. Once the course is set, burn the boats. Then burn the dock.

When we pivoted from MedPiper’s doctor network to insurance infrastructure, I didn’t keep the old product alive “just in case.” Ellison sailed far off course and burned his boats. I learned the same lesson the expensive way: half-commitment is full failure. When you straddle two strategies, you execute neither. The back door is where your energy leaks out; and by the time you notice, you’re near death.

3. Regret is a compass, not a feeling.

Bezos projected himself to age 80 and asked what he’d regret more: trying and failing, or never trying. He called it the Regret Minimization Framework.

I use a simpler version: the pain of failure is acute and temporary. The pain of “what if” is chronic and terminal. When I left a stable growth role to co-found a company during a pandemic, the asymmetry was obvious; I just had to be honest enough to see it.

4. Skin in the game is the only credential that matters.

Taleb’s central argument in his book Skin in the Game: don’t tell me what you think, tell me what’s in your portfolio.

I’ve bet my career on insurance infrastructure in India- through multiple pivots, burning personal savings, choosing complexity over comfort. Anyone can have an opinion about risk from the gallery. The arena is where opinions become expensive.

5. Ergodicity is the most important concept no founder talks about.

Ole Peters’ work on ergodicity economics shows that expected value: the standard metric every finance textbook teaches; is a lie told by ensemble averages.

What matters isn’t what happens on average across all possible versions of you. It’s what happens to you, sequentially, over time.

One catastrophic loss and the game ends. I obsess over survival before growth because you can’t compound what doesn’t exist. The Kelly Criterion isn’t conservative; it’s the only strategy that respects the fact that you only get one timeline, one life.

II. On rails, infrastructure, and the architecture beneath

6. Most companies build cars. I build super-highways.

Patrick Collison built Stripe because he realized the payments layer of the internet was missing. Not the store, not the product, not the checkout experience, but the invisible layer underneath that made all of it possible. OrbitCover isn’t an insurance product. It’s insurance infrastructure taking birth; the rails on which thousands of future products can run.

I’ve spent a decade learning that the unsexy layer underneath is where all the leverage lives.

7. Apply engineering discipline to every function, not just the product.

Ellison applied engineering discipline to Oracle’s entire business, not just product development. I now know better and borrow this wholesale.

The biggest engineering problems in a company aren’t in the codebase: they’re in how you price, sell, hire, and communicate. When your sales motion is a hack, your unit economics are a hack. One should engineer the whole machine.

8. Build for billions, not millions.

India has 1.4 billion people. Most have never filed an insurance claim, never seen a payout arrive in under 3 days, never had claim related data travel seamlessly between providers. We’re not digitizing existing workflows. We’re building the workflow that should have existed. The gap isn’t a product gap. It’s an infrastructure gap. You don’t close it with features; you close it with rails.

9. The “boring” problem is the $100B problem.

Insurance middleware. Health data plumbing. Payment reconciliation. Nobody pitches these at Demo Day with fireworks. That’s exactly why they’re still broken. The Collison brothers built a $95B company on payment processing; the problem even some VCs called “solved.”

John Collison once said the internet’s economic infrastructure was still largely unbuilt. He was talking about India too, even if he didn’t know it.

10. Integration beats best-of-breed.

Jobs didn’t ship components. He shipped systems. The hardware, the software, the services, all integrated so tightly that the seams disappeared. OrbitCover doesn’t hand airlines a parametric trigger engine and wish them luck. We deliver the full stack: risk modeling, data verification, UPI payout rails, regulatory compliance; all deeply integrated, tested, alive. The value is in the seams disappearing.

On seeing what others can’t and the discipline of not fooling yourself

11. The brain’s primary purpose is deception, and the primary target is the owner.

Ellison said this and it stopped me cold. Every founder thinks their idea is the one. Most are wrong. I’ve been wrong: spectacularly, publicly, expensively.

The difference between delusion and conviction is evidence. I guard against myself with the same paranoia I apply to competitors. Is this belief load-bearing? Or is my ego doing the engineering?

12. Don’t mistake the present for the future.

This is the worst mistake a tech company can make, per Ellison. When we told OTAs that parametric payouts would become table stakes, they said “nobody’s asking for this.” Of course nobody’s asking. Customers defend what they know, not what they’ll need. Jobs never ran a focus group. He built what he knew people would want once they saw it.

The most dangerous sentence in business is “our customers want X”; because customers are describing today’s constraints, not tomorrow’s possibilities.

13. People don’t know what they want until you show them.

Jobs said this explicitly. No passenger walked into an airport and said “I want automatic flight delay insurance with an instant UPI payout triggered by a data feed I never see.” They said “insurance is useless.”

My job isn’t to build what the market requests. My job is to read things that are not yet on the page. Build from the future backwards.

14. Most people are so in love with their own ideas that it confines their thinking.

Quoting Ellison again. The pivot from MedPiper’s doctor network to insurance middleware wasn’t romantic. It was the result of relentlessly asking “what’s actually working?” and being honest when the answer was “not what we planned.”

The best ideas don’t care who had them. Take the signal, regardless of source.

15. The only way to get ahead is to find errors in conventional wisdom.

Indian insurance can’t be disrupted. Parametric products are too niche. Health data interoperability is a pipe dream. I’ve heard every version of “it can’t be done here.”

Conventional wisdom is a map of where everyone else already is. The edges of that map, where it says “here be dragons”; those are the only places worth going.

16. Measure what humans actually experience, not what your specs say.

Amar Bose bought the best stereo he could find by specification and found the sound terrible. He spent the rest of his career studying psychoacoustics: how sound is actually perceived, not how it measures on paper.

Insurance with a 98% claim approval rate is meaningless if it takes 17 days. We measure one thing: did the money arrive in the passenger’s account before they left the airport? The spec is irrelevant. The experience is everything.

On pivots, paranoia, and the will to survive inflection points

17. Only the paranoid survive; and I’ve earned my paranoia.

This was Andy Grove’s entire thesis at Intel. He watched strategic inflection points destroy companies that couldn’t see the shift happening. I’ve built and rebuilt the same company three times. Doctor network. Health data API. Insurance middleware. Each pivot was a near-death experience. Each taught me that the company you are today has a half-life, and the decay starts the moment you stop questioning your own model.

18. When the fundamentals shift, walk out the door, walk back in, and be the new CEO.

This was Grove’s most powerful thought experiment. When Intel was hemorrhaging money in the memory business, he asked Gordon Moore: “If we got kicked out and the board brought in a new CEO, what would he do?” Moore said: “He would get us out of memories.” Grove’s reply: “Why shouldn’t you and I walk out the door, come back in, and do it ourselves?”

When I pivoted to insurance rails, I had to become a different founder; one with no attachment to 65,000 verified healthcare professionals and the product we’d lovingly built.

19. Antifragility is a design choice, not a personality trait.

Taleb’s taxonomy: fragile systems break under stress, robust systems endure, antifragile systems get stronger.

I build for antifragility deliberately. Every pivot made us better, not just different. The doctor network taught us health data. Health data taught us insurance underwriting. Insurance underwriting taught us parametric triggers. The failures weren’t waste; they were load-bearing walls in the next structure.

20. Chaos reigns. Then you rein in the chaos.

Grove described two phases of organizational life: the phase where everything is questioned, debated, torn apart within the org; and the phase where you commit, align, and march. The failure mode isn’t chaos itself. It’s staying in one phase too long.

Debate forever and you’re paralyzed. March without debating and you’re marching off a cliff.

21. Disagree and commit; in both directions.

Bezos formalized this at Amazon. But here’s the part most people miss: it’s not a top-down command. It’s a two-way protocol.

When my market feedback data or team has conviction and I don’t, I disagree and commit; fully. Not performatively. Not with a hedge. Organizational speed comes from trust, and trust requires that the person at the top can also be the one who yields.

On first principles, inference, and computational thinking

22. First principles are not a philosophy. They are a practice.

Every assumption I have is load-bearing until tested. Why does insurance require a claims process? Because traditional indemnity models need loss verification.

But parametric insurance doesn’t; the data is the verification.

When you re-derive the problem from physics and not from precedent, entirely new architectures become visible. I don’t need to optimize existing systems. I ask whether the system should exist at all.

23. Inference is the fundamental act of intelligence; for both human and machine.

My interest in AI isn’t about just tools tools. It’s about the nature of inference itself: how do you update beliefs given evidence? How do you act under uncertainty? How do you distinguish signal from noise at scale? And all of these unlock what new outcomes, to what new future?

Bayesian thinking isn’t an academic exercise; it’s the operating system for every decision I make. Prior belief, new evidence, posterior update, act. Repeat until dead.

24. Computational thinking is a liberal art.

Jobs put the Macintosh at the intersection of technology and the humanities. I believe that computational thinking: decomposition, pattern recognition, abstraction, algorithmic design; belongs in every discipline.

When I consult for enterprises on tech/AI transformation, I’m not selling software. I’m teaching them a way of thinking. The companies that will win aren’t the ones that adopt AI tools fastest. They’re the ones whose leaders think computationally about every problem, and are able to empower their team to act on it.

25. Status quo bias is the silent killer of enterprises.

Kahneman and Tversky’s prospect theory shows that humans weight losses roughly 2.5x more than equivalent gains. This means every proposed change feels like a net negative, even when it’s a net positive.

The companies I consult for don’t resist change in tools because they’ve evaluated it and found it wanting. They resist because the pain of changing feels larger than the pain of staying. My job is to make the cost of inaction visible.

26. Survivorship bias will poison your strategy if you let it.

Every founder who reads about Bezos or Collison or Ellison is studying winners. Nobody studies the thousands who made identical moves and failed.

Abraham Wald showed this during WWII: the military wanted to armor the parts of returning planes that showed bullet holes, but Wald realized they should armor the parts that didn’t; because planes hit there never came back.

I don’t build strategy from hero narratives. I build it from failure patterns.

On people, agency, and the teams that survive

27. You are not building a product. You ship a product that looks like the team building it.

Jobs said his greatest creation wasn’t the Mac or the iPhone; it was Apple, the company. The product is a second-order effect of the people.

Here on, every hour I’ve spent on hiring, firing, and team design has to return more than any hour I’ve spent on product design. This is counterintuitive for a builder. It is also true. The team should be the asset. Everything else is output.

28. A-players hire A-players. B-players hire C-players. The decay is exponential.

Jobs was ruthless about this and right to be. One B-player in a critical role doesn’t just underperform; they lower the hiring bar for everyone who follows. They hire people who make them feel safe, not people who make the company better.

Within two cycles, what I thought was a A player got a team optimizing for comfort, not outcomes. So compounding works in reverse too: mediocrity scales faster than excellence. Never again.

29. Low agency is the most expensive trait in a startup. Detect it early. Act on it faster.

There is a specific kind of person who waits to be told what to do, asks permission for obvious decisions, and treats ambiguity as someone else’s problem. In a large company, they survive; the system routes around them. In a startup, they are a tax on every high-agency person around them.

Every hour a high-agency teammate spends compensating for a low-agency one is an hour of leverage destroyed. I’ve made the mistake of patience here. The learning: the kindest thing you can do for them and for the team; is to make the call fast.

Six months of “coaching” a fundamentally low-agency person is six months of lying to both of you.

30. High agency looks like unreasonable resourcefulness.

You’ll know it when you see it. The engineer who ships a fix before you’ve finished describing the bug. The ops person who finds a workaround to a regulatory blocker while legal is still drafting the memo. The salesperson who closes a deal using a pathway that didn’t exist in your playbook.

George Bernard Shaw’s line applies: the reasonable person adapts to the world; the unreasonable one persists in trying to adapt the world to themselves.

All progress depends on the unreasonable person. Hire for unreasonableness. Protect it. Promote it.

31. Two-pizza teams. Not because of communication overhead, because of accountability diffusion.

Bezos’s two-pizza rule gets cited as a communication efficiency hack. The real reason is deeper: in small teams, there is nowhere to hide. Everyone’s contribution or lack of it, is visible to everyone else.

Accountability isn’t enforced by managers; it’s enforced by proximity. When the team is eight people and one isn’t pulling weight, the other seven feel it in their bones. That social pressure is more precise than any performance review.

32. Constructive confrontation is a skill, not a personality type.

Grove built Intel’s culture around what he called constructive confrontation: the expectation that anyone could challenge anyone, regardless of rank, as long as the challenge was about the idea and not the person.

Most teams avoid conflict because they confuse disagreement with disrespect. The result is polite consensus that nobody believes in.

I’d rather have a loud room that arrives at a real answer than a quiet room that ships a compromise.

33. Credentials without skin in the game are noise. Cut the noise like cancer.

Taleb coined the term “Intellectual Yet Idiot” for the well-pedigreed expert class whose primary skill is passing exams written by people like them. They have opinions about everything and exposure to nothing.

I’ve sat across from folks with Ivy League degrees and Fortune 500 titles who told me X “can’t work in India” while having zero personal or financial stake in the outcome. Their caution wasn’t wisdom; it was risk aversion dressed in a suit.

Taleb’s filter is binary: if you do not take risks for your opinion, you are nothing. Paul Graham put it differently: in a startup, credentials don’t matter; your users won’t care if you went to Stanford. When someone with impressive credentials but zero skin in the game tells you to be cautious, they’re not protecting you. They’re protecting themselves from being associated with your potential failure.

The cost of listening to these people is measured in years you’ll never get back. Identify them early. Disengage completely. Don’t argue. Don’t try to convince. Just cut and move.

34. Culture is what happens when leadership isn’t in the room.

Culture is not the values on the wall. Not the all-hands speech. Culture is the decision a junior engineer makes at 11 PM when nobody is watching and the shortcut is right there.

You build culture by who you hire, who you fire, what you tolerate, and what you celebrate. Ellison said you cannot run a company without strong checks and balances. True.

But checks and balances are mechanical. Culture is organic. You need both, and the organic one is harder to build and easier to destroy.

35. Loyalty is earned daily, not owed historically.

The fact that someone was there in the early days doesn’t mean they belong in the next phase.

This is the hardest call a founder makes: the person who was perfect for a 5-person team may be wrong for a 50-person team, and the gratitude you feel for their early sacrifice will cloud your judgment about their current fit.

Grove faced this when Intel pivoted from memory to microprocessors; some of the best memory engineers couldn’t make the transition. He had to choose the company’s future over the team’s past.

If someone’s best contribution is behind them and they can’t grow into what’s next, keeping them is a disservice to everyone, including them.

36. Hire missionaries, not mercenaries.

This John Doerr’s framing, but I’ve lived it. Mercenaries optimize for personal upside. Missionaries optimize for the mission. You can tell the difference in the first month: the mercenary asks “what’s in it for me?” when things get hard. I’d a COO who did that. The missionary asks “what do we need to do?”

The gap is especially visible in Indian startup hiring, where the talent market is hot and optionality is abundant. The person who joins your 15-person insurance infrastructure company when they have offers from funded consumer companies; that person is telling you something about what they believe. Listen.

On customers, taste, and the craft of building

37. Obsess over the customer, not the competitor.

Bezos built Amazon’s strategy around this distinction. Competitor-focused companies wait for the other side to move. Customer-focused companies invent.

No Indian airline passenger asked for parametric flight delay insurance with instant UPI payouts. But every single one has sat in an airport, watched the delay board, and felt the helplessness of knowing their insurance policy is useless in real time.

Start with that frustration. Work backwards.

38. Taste is the discipline of knowing what to subtract.

Jobs said innovation is saying no to a thousand things. Bose refused to chase specifications; he chased how sound felt in a room. The temptation for a founder to expand surface area is constant. But every feature you add that isn’t essential dilutes the ones that are.

OrbitCover does one thing: when your flight is delayed, money appears in your account. No forms. No calls. No app to download. The absence of complexity is the product.

39. The best products recruit believers, not customers.

Ellison understood that the 3 key things: product name, the narrative, the vision; these recruit people who believe the future you’re describing is inevitable. When I pitch OrbitCover to airlines, I’m not selling an insurance add-on. I’m recruiting them into a vision of what passenger experience should look like. Believers build with you. Customers churn.

40. Hard strategies, well executed, beat easy strategies at any level of execution.

Ellison chose the database market when it was tiny and “academic.” Jobs chose smartphones when Nokia owned the world. Grove chose microprocessors when Intel was a memory company. I choose markets others avoid not out of contrarianism, but because the defensive moat of complexity is the only moat a small team can build.

On commitment, deception, and the nature of belief

41. Commitment is a one-way door. Walk through it deliberately, then never look back.

Cialdini’s research on commitment and consistency shows that once you commit publicly and actively, you’ll reorganize your entire identity to stay consistent. I use this on myself, deliberately. When I announce a direction, write it down, tell my team, tell investors; I’m not just communicating. I’m closing the door behind me.

42. The fourfold pattern of risk explains most bad decisions.

Kahneman showed that people are risk-seeking when facing certain losses and risk-averse when facing uncertain gains. This is why companies cling to dying products (avoiding certain loss) and reject promising pivots (uncertain gain).

Every pivot I’ve made required overriding this wiring. The doctor network was a certain loss of sunk cost. Insurance infrastructure was an uncertain gain. My brain screamed to stay. I walked out the door anyway.

43. Don’t compete on specifications. Compete on what the human actually feels.

Bose proved that the best frequency response curve doesn’t mean the best sound. Jobs proved that the best spec sheet doesn’t mean the best product.

In insurance, the “specification” is claim approval rate. The human experience is: did I feel taken care of when something went wrong?

We don’t optimize metrics. We optimize the moment of truth - the moment the passenger realizes the money is already there in his account.

44. Simplicity on the surface requires brutal complexity underneath.

The UPI payout that arrives in seconds requires: real-time flight data ingestion, parametric trigger logic, regulatory-compliant policy issuance, reinsurance settlement, bank API integration, and fraud detection; all executing in parallel, all invisible.

Jobs said it well: simple can be harder than complex. You have to work hard to get your thinking clean enough to make it simple. The user sees one notification. We built the entire mountain beneath it.

On endurance, identity, and the long game

45. Endurance is a skill, not a trait. I train it deliberately.

Training for Hyrox isn’t a metaphor. It’s literal practice: eight rounds of running interspersed with eight functional fitness stations. There is no shortcut. There is no clever hack. You train the capacity to sustain effort when everything in your body says stop.

Building a company in Indian insurance infrastructure requires the same physiology; not sprints of brilliance, but years of grinding through regulatory complexity, enterprise sales cycles, and infrastructure debt.

46. The clock in the infrastructure game is measured in years, not quarters.

Ellison spent twelve years making Oracle’s database clustering work. Bose spent twenty-four years on an electromagnetic car suspension system, a research project he never planned to commercialize but couldn’t stop pursuing because the physics fascinated him.

I’ve spent six years building and rebuilding the rails that connect health data, insurance logic, and financial payouts in India. Infrastructure isn’t a product launch, it’s almost a geological process.

47. Rejection is a reality that should be surpassed, not ignored. I had teachers who told me I’d amount to nothing. This is real, not perceived or forgotten.

I said it at the top and I’ll say it again here, because it matters for the principles that follow. Ellison’s father dismissed him. Jobs was abandoned. Grove was a refugee. I’m not citing their stories to paper over my own. The teachers in Chennai who wrote me off didn’t create my drive; but they calibrated it. Every milestone since has been partial proof against a verdict I never accepted.

48. The journey across domains isn’t wandering. It’s architecture.

Music events. Mobility. Healthcare. Insurance. Aviation. On the surface, it looks likeI am a founder who can’t sit still. Underneath it, every domain taught me the same lesson: in emerging economies like India there is value in building the product and the connective tissue: the middleware, the rails, the infrastructure that lets disparate systems talk. I’m not a serial pivoter.

I’m an engineer-architect who’s been drawing the same blueprint across different materials, to deliver outcomes.

49. It is always Day 1.

Bezos wrote this in his very first Amazon shareholder letter in 1997, and repeated it in every letter after. Day 2 is stasis, then irrelevance, then death.

India’s insurance penetration is under 4%. Digital health infrastructure covers a fraction of the population. Parametric insurance barely exists in here and other emerging markets. We haven’t built 1% of what India needs. The competition is absence, the infrastructure that doesn’t exist yet.

50. Be willing to be misunderstood for long periods of time.

Bezos again: if you’re doing something truly new, you should expect to be misunderstood. When we raised on a doctor network and pivoted to insurance middleware, people were confused. When we built health data APIs and then launched a flight delay product, people were more confused. That’s fine.

The narrative coherence comes later. Right now, the job is to build the right thing the right way, even if the story doesn’t fit on a slide.

The meta-principles

51. I am a micro pessimist and a macro optimist.

Every sprint will have failures. Every quarter will have near-misses. The next regulatory change might break our model. But over the arc of decades, insurance globally will become instant, data-driven, and embedded in every transaction.

I intend to execute with a sense of daily paranoia, and decade-long conviction. That combination is the only operating system that works.

52. Make the type of decision the situation demands.

Bezos divided decisions into two types:

Type 1 : irreversible, high-stakes, deliberate slowly; and

Type 2 : reversible, low-stakes, decide fast.

The failure mode of growing companies is treating every decision like Type 1.

Speed is a feature. Perfectionism is a bug. Know which door you’re walking through.

53. I am not building a company. I am improving lives for millions by building a layer.

Companies come and go. Layers persist. SMTP persists. TCP/IP persists. UPI persists. I want to build the product that also owns the layer between the event (your flight was delayed), the data (verified in real time), and the outcome (money in your account). That layer doesn’t belong to OrbitCover. It belongs to the architecture of how financial protection works.

We just happen to be the ones building it first.

54. Let the work be the proof.

I don’t need anyone to believe me. I need the system to work: verifiably, instantly, at scale. The receipts are the argument. The manifesto is just the footnotes.

55. Everything is a hypothesis. The universe is testing yours right now.

Your company is a hypothesis. Your strategy is a hypothesis. Your identity is a hypothesis. The universe doesn’t care about your narrative; it runs the experiment whether you’re ready or not, and it publishes the results whether you like them or not.

The founders who survive aren’t the smartest or the most funded. They’re the ones who update their beliefs fastest when the evidence demands it, and hold them longest when it doesn’t.

I’ve watched founders with better ideas, better funding, and better teams die because they couldn’t tell the difference between signal and noise; or worse, because they could tell but chose not to.

Updating is painful. It means yesterday’s conviction becomes today’s sunk cost. But the alternative is running an experiment whose result you’ve already decided to ignore. That’s not building. That’s theatre.

56. If the decision doesn’t scare you, it’s not large enough to matter.

Every decision that changed my life shared one physical trait: a tight stomach. Not the abstract fear of “what if it fails”; but the visceral, blood-draining terror of committing to a path when the path you’re on is still producing oxygen.

Leaving a job that was working. Killing a product that had users. Betting savings on a market that might not exist. Over time I stopped treating that terror as a warning and started treating it as a compass. Fear marks the boundary between decisions that are comfortable and decisions that are consequential.

If you can make the call without your hands shaking, the stakes aren’t high enough to compound into anything meaningful. Seek the tight stomach. That’s where the asymmetry lives.

A note on sources

These 56 principles are synthesized from a decade of building and a lifetime of reading. The thinkers and builders whose ideas shaped them most directly:

Larry Ellison : Softwar: An Intimate Portrait of Larry Ellison and Oracle by Matthew Symonds, and the Founders podcast episode “How Larry Ellison Thinks.” His ideas on engineering discipline, burning boats, risk geometry, and the brain’s capacity for self-deception run through principles 1, 2, 7, 11, 12, 14, 35, 40, 41, 47.

Jeff Bezos : His annual shareholder letters (1997-2023) and the Regret Minimization Framework. Day 1 thinking, disagree-and-commit, customer obsession, Type 1/Type 2 decisions, two-pizza teams, and the willingness to be misunderstood appear in principles 3, 21, 31, 38, 50, 51, 53.

Steve Jobs : His Stanford commencement speech, Walter Isaacson’s biography, and decades of product launches. The intersection of technology and liberal arts, building what customers don’t yet know they want, A-player hiring, the discipline of subtraction, and integration as philosophy inform principles 10, 12, 13, 24, 27, 28, 39, 44, 45.

Patrick and John Collison : Stripe’s origin story and their writing on infrastructure, optimism, and “boring” problems. The road-not-the-car metaphor and the insight that the internet’s economic infrastructure is still unbuilt come directly from their thinking. Principles 6, 8, 9.

Amar Bose : His psychoacoustics research at MIT and the philosophy of measuring human experience over specification. A Bengali immigrant’s son who never stopped being a researcher, even while running a billion-dollar company. Principles 16, 39, 44, 47.

Andy Grove : Only the Paranoid Survive and the Intel memory-to-microprocessor pivot. Strategic inflection points, the revolving-door test, constructive confrontation, and the two phases of organizational life. Principles 17, 18, 20, 32, 36.

Nassim Nicholas Taleb : Antifragile, Skin in the Game, and the “Intellectual Yet Idiot” essay. The fragile-robust-antifragile taxonomy, skin in the game as the filter for whose opinion matters, the IYI concept applied to credentialed naysayers, and the imperative to have exposure before having opinions. Principles 4, 5, 19, 33.

Daniel Kahneman and Amos Tversky : Thinking, Fast and Slow and prospect theory. Loss aversion, the fourfold pattern of risk attitudes, and status quo bias. Principles 25, 43.

Robert Cialdini : Influence: The Psychology of Persuasion. Commitment and consistency as both a psychological principle and a deliberate tool. Principle 42.

Paul Graham : His essays on startups, credentials, and resourcefulness. The insight that in startups, credentials don’t matter and resourcefulness is the defining trait. Principles 30, 33.

John Doerr : The missionary vs. mercenary framing for startup teams. Principle 37.

George Bernard Shaw : The unreasonable man formulation. Principle 30.

Ole Peters : Ergodicity economics and the distinction between ensemble averages and time averages. The most underappreciated framework in decision science. Principle 5.

Abraham Wald : The WWII survivorship bias analysis. Looking at what’s missing, not just what’s visible. Principle 26.

This is Version 1.0. It will evolve as I do. If you’re building something that everyone says can’t work, I’d like to hear from you.

— Nitthin

In 1959, two physicists proposed an experiment that shouldn’t work.

Yakir Aharonov and David Bohm predicted that an electron could be affected by a magnetic field it never touches. Send an electron around a solenoid - a coil with a magnetic field confined entirely inside it. The electron travels only through regions where the magnetic field is exactly zero. Yet its behavior changes depending on the field inside the coil.

This was heresy. In classical physics of the time, fields are what’s real. The magnetic field B exerts forces. The magnetic potential A is just a mathematical convenience; a bookkeeping device. Where B is zero, nothing physical should happen. Potential, to physicist of the time; was a mathematical instrument, not a physical reality.

But the experiment was done. Aharonov and Bohm were right. The electron’s quantum phase shifts based on the potential A, even where the field B vanishes.

Something the smartest of us ever thought was “just mathematics” turned out to be physically real. And the thing we thought was fundamental - the local field; turned out to be secondary to this potential.

This is where our inference frame faces its deepest test. We’ve claimed the universe is an inference machine. Now we must ask: what exactly is being inferred? What’s real: fields, potentials, or something stranger still? And what does “local” even mean when quantum mechanics enters the picture?

Let’s find out.

The old-school hierarchy of reality

Start with what seems obvious. In physics, we have had an hierarchy for ages:

Level 1: Observables What you can measure. The magnetic field strength at a point. The electric force on a charge. The position of a particle. These are “real” in the most operational sense; you can build instruments to detect them.

Level 2: Fields The are things that determine the above observables. The electric field E, the magnetic field B. You can’t see a field directly, but you can see its effects everywhere. Fields seem pretty real, coz the observables experience the field’s effects.

Level 3: Potentials The things that determine the fields. The electric potential φ, the magnetic vector potential A. The fields are derivatives of the potentials:

E = -∇φ - ∂A/∂t
B = ∇ × A

Potentials seem like mathematical conveniences. They were created for mathematical ease. You can change them with (gauge transformation) without changing the fields. So they can’t be “real”; different potentials give the same physics.

Level 4: Phases The quantum mechanical phase that accumulates as a particle moves. The phase depends on the potential integrated along the path. But absolute phase is unobservable; only the phase differences really matter.

The classical intuition says: reality flows downward. Phases determine potentials, potentials determine fields, fields determine observables. The higher levels are “just math” for the sake of calculation ease.

The Aharonov-Bohm effect inverts this. The potential, supposedly just math; has real world, physical effects that the field doesn’t capture. The hierarchy that had been assumed for ages to be true, is wrong, or at least incomplete.

Locality: the classical physics story

Classical physics is local. This is one of it’s deepest assumptions and most crucial one.

Locality: What happens at a point depends only on conditions at that point and its immediate neighborhood.

Fields are the machinery of that locality. Instead of action at a distance (Newton’s gravity reaching across empty space), here we have fields that exist everywhere and propagate at a finite speed.

The electric field at point P determines the force on a charge at P. The charge doesn’t need to “know” about distant sources; it only needs to know the local field value and behaves accordingly.

The inference translation

In the inference frame, locality means:

Local inference: The update at each point depends only on local data (field values) and local priors (how this point relates to neighbors).

This is like a cellular automaton. Each cell updates based on its neighbors. The global pattern emerges from these local rules.

Similarly, fields are the channels for local inference. The field value at P is the “message” available at P. The laws of physics are the local update rules.

This is clean. This is elegant. This is also wrong; or at least incomplete.

The Aharonov-bohm effect: when path matters

Here’s the setup. An electron source emits electrons toward a screen. Between source and screen is an impenetrable solenoid; a thin coil with magnetic field B confined entirely inside. The electron can pass on either side but cannot enter the solenoid.

Outside the solenoid, B = 0 everywhere. The magnetic field vanishes in the entire region the electron can access. By the locality principle, the electron shouldn’t “know” about the field inside.

But quantum mechanics says the electron’s phase accumulates according to:

φ = (e/ℏ) ∮ A · dl

Mathematically, this is described as: the phase depends on the line integral of the potential A around the path; even where B = 0.

When electrons going around different sides of the solenoid meet at the screen, their phases differ. This phase difference produces interference fringes. Change the current in the solenoid (changing B inside, changing A outside), and the fringes shift.

The electron is affected by a field it never encounters. And this is crazy.

What this means

Option 1: Potentials are real. The potential A, not just the field B, has physical meaning. The electron couples to A directly. Gauge transformations change something physically real.

Problem: Gauge transformations are supposed to be unphysical; different mathematical descriptions of the same physics. If A is real, why/how can we change it arbitrarily?

Option 2: The phase is real. What’s physical isn’t A itself, but the integral ∮ A · dl around a closed loop. This is gauge-invariant; meaning it doesn’t change under gauge transformations. The “reality” lives in the path, not the point.

Option 3: Non-locality is real. The electron’s state depends on global information (the flux through the solenoid), not just local information (the field at its location). Locality fails; but in a very subtle way.

The inference translation

In inference terms, the Aharonov-Bohm effect says:

Your posterior depends on the path you took, not just where you are now.

The accumulated phase is like accumulated evidence. Two observers who end up at the same point, but arrived via different paths, have different posteriors; even if the local “data” (field) is identical. This is not standard Bayesian inference, where only the current data and prior matter. This is path-dependent inference - the history of the thing or entity is baked into the state of the entity.

The potential A encodes “how inference accumulates along paths.” The field B = ∇ × A encodes “the local rate of accumulation.” But the total accumulation can be non-zero even when the local rate is zero everywhere along the path; if the path encloses a region where things are different.

This is non-locality, but of a specific kind, mathematically termed as: topological non-locality. The electron doesn’t interact with the solenoid directly. But its path topology, whether it goes around the solenoid or not; seems to matter.

Gauge symmetry: what can you actually infer?

Let’s dig into gauge invariance, because it’s the key to understanding what’s real. There’s a bit of high-school+ maths here.

A gauge transformation changes the potential:

A → A + ∇χ
φ → φ - ∂χ/∂t

where χ is any smooth function. This changes A and φ, but leaves the fields E and B unchanged. It also leaves the Aharonov-Bohm phase ∮ A · dl unchanged (because the added term ∮ ∇χ · dl = 0 for a closed loop).

What gauge invariance means

Gauge invariance says: certain aspects of the mathematical description are not physically meaningful. Eg., The absolute value of the potential at a point is meaningless. Only potential differences matter (for φ) or circulationsmatter (for A).

This is exactly like coordinate invariance. The absolute position (x, y, z) = (3, 4, 5) is meaningless; it depends on where you put the origin. Only relative positions really matter.

The inference translation

Gauge symmetry means certain degrees of freedom are unobservable; they don’t enter the likelihood function.

You can never design an experiment that measures the absolute potential. No data will ever distinguish A from A + ∇χ. So inference about the absolute potential is impossible; the posterior equals the prior, always.

Gauge-invariant quantities are what you can infer with measurement:

• Field strengths (E, B)

• Phase differences

• Topological invariants (like total flux)

The gauge potential A is like a “latent variable” that you marginalise over. It’s useful for mathematical calculations, but the physical content lives in the invariants.

Why gauge theories?

Here’s a deep question: why does physics use gauge theories at all, if the gauge freedom is unphysical?

The answer: theory and locality requires it.

You want a theory where interactions are local; a charge at point P interacts with the field at point P. But you also want the theory to respect certain global symmetries (like the phase symmetry of quantum mechanics).

The price of combining locality with global symmetry is via gauge freedom. You get extra mathematical degrees of freedom that encode “how to connect local descriptions consistently.”

The gauge potential A is called a connection in differential geometry. It tells you how to compare phases at different points. It tells you how to “parallel transport” the quantum state from one location to another.

This is an inference machinery: the gauge potential is the grammar that lets local inferences combine into global consistency.

Quantum phase as an accumulated prior

Let’s make the inference interpretation precise.

In quantum mechanics, the state of a particle is described by a wave function :

ψ(x) = |ψ(x)| e^(iθ(x))

The magnitude |ψ|² gives probabilities. The phase θ determines interference.

As a particle moves through space, its phase accumulates:

θ(path) = ∫_path (p · dx - E dt) / ℏ

where p is momentum and E is energy. In the presence of electromagnetic fields, this becomes:

θ(path) = ∫_path [(p - eA) · dx - (E - eφ) dt] / ℏ

The potentials A and φ modify how phase accumulates.

The inference translation of this

Quantum phase is the log-probability accumulated along a path.

More precisely, in the path integral formulation:

Amplitude = ∫ [all paths] exp(iS[path]/ℏ)

where S[path] is the action for that path.

Each path contributes with a weight e^(iS/ℏ). Paths with similar actions interfere constructively; paths with different actions interfere destructively.

The classical path (least action) is where the phase is stationary; all nearby paths have similar phases, so they add up. This is why classical behavior emerges: the classical path dominates the sum.

In Bayesian terms:

• Each path is a hypothesis

• The action S is (something like) the negative log-prior

• The path integral is model averaging

• The stationary phase approximation is the MAP estimate

The potential A modifies the action, and therefore modifies which paths contribute. The Aharonov-Bohm effect is: the potential changes the weights in the model average, even where the “local data” (field) is zero.

Bell’s theorem: when inferences can’t be local

Now we go further. The Aharonov-Bohm effect shows topological non-locality; that path dependence matters.

But quantum mechanics has a deeper non-locality: entanglement.

The setup

Two particles are prepared in an entangled state; say, with total spin zero. They fly apart to distant locations where Alice and Bob measure them.

Quantum mechanics predicts: the results will be correlated in a specific way. If Alice measures spin-up along some axis, Bob will measure spin-down along the same axis, with certainty.

This isn’t surprising by itself. Classical correlations can do this: if I put one red ball and one blue ball in two boxes, shuffle, and send them to Alice and Bob, their results will be correlated.

Bell’s Inequality

John Bell proved something shocking: the quantum correlations are stronger than any classical explanation allows.

If you assume:

1. Realism: Each particle has definite properties before measurement

2. Locality: Alice’s measurement doesn’t affect Bob’s particle (or vice versa)

Then you can prove an inequality that the correlations must satisfy.

Quantum mechanics violates this inequality. Experiments have confirmed the violation repeatedly. Hence one of the earlier assumptions must be wrong.

The inference translation of this

Bell’s theorem says: you cannot explain quantum correlations with local, independent priors.

If each particle had a “local hidden variable” λ , that provides a complete description of its state that determines all measurement outcomes; then the correlations would satisfy Bell’s inequality.

The violation means: the quantum state is not a product of local states. It’s irreducibly joint. Your high school probability basics do not work here.

Entanglement means the prior is over joint states, not products of local states.

You cannot write:

P(A, B) = ∫ P(A|λ) P(B|λ) P(λ) dλ

where A is Alice’s outcome and B is Bob’s. No assignment of local probabilities works. The correlation is fundamental, not derived from local properties.

What kind of non-locality are we talking about?

The type that doesn’t allow faster-than-light signaling. Alice can’t control her outcome, so she can’t send information to Bob through the correlations. The non-locality is in the correlations, not in any causal influence.

In inference terms:

Some inferences are irreducibly about relationships, not about the things themselves.

The entangled state doesn’t say “particle A has property X and particle B has property Y.” It says “the relationship between A’s outcome and B’s outcome is Z.” The relationship is primary; the individual properties are derived (upon measurement).

This is strange. But it’s consistent with our inference frame; if we proceed to accept that some priors are over joint configurations, not product distributions over local states.

Holography: when the boundary knows everything

Here’s perhaps the strangest result in modern physics. The holographic principle (from string theory and black hole physics) suggests:

The information content of a region of space is encoded on its boundary, not in its volume.

A black hole’s entropy (a measure of information content) is proportional to its surface area, not its volume. This is bizarre; you’d expect a bigger volume to hold more information.

The AdS/CFT correspondence goes further: a theory of quantum gravity in a volume is exactly equivalent to a theory without gravity on the boundary. The bulk and boundary are dual descriptions of the same physics.

The inference translation

Local inference in the bulk is equivalent to global inference on the boundary.

What happens “inside” (the volume) can be completely reconstructed from what happens “outside” (the boundary). What this translates to is that interior and exterior of a black hole are not independent; they’re related by a kind of inferential duality.

This is the ultimate limit of non-locality: there is no purely local information. Everything local is encoded globally, and vice versa.

For our inference frame, this suggests:

The universe’s inference is not localized in space. Spatial locality is an emergent, approximate description of a fundamentally non-local structure.

This is fully speculative. But for me it aligns with something we’ve seen throughout: the “local” description (fields at a point) is always embedded in global structure (gauge invariance, topological effects, entanglement).

Symmetry breaking: when priors collapse

We’ve talked about symmetry as invariance, of transformations that don’t change the physics. Now let’s talk about what happens when symmetry breaks.

The Mexican hat

Consider a potential shaped like a Mexican hat (sombrero):

V(φ) = -μ²|φ|² + λ|φ|⁴

This potential has a circular symmetry. You can rotate φ in the complex plane and the potential doesn’t change.

But the minimum of the potential isn’t at φ = 0. It’s on a circle |φ| = v. The system has to “choose” a point on that circle; and any choice breaks the rotational symmetry.

The laws are symmetric. But the ground state is not. This is spontaneous symmetry breaking.

The inference translation

Symmetry breaking is what happens when the prior is symmetric but the posterior is not symmetric.

The potential is like a symmetric prior; no direction is favored. But the system must settle into a definite state (the posterior). That state necessarily picks a direction.

This is like a Bayesian model with a symmetric prior that gets “broken” by data. The prior says “all directions equally likely.” The data (or thermal fluctuations, or quantum fluctuations) forces a choice.

Why Symmetry Breaking Matters

Symmetry breaking generates:

• Mass: The Higgs mechanism gives particles mass through symmetry breaking

• Phase transitions: Water freezing, magnets magnetizing, superconductivity

• Diversity: Why the universe isn’t a uniform soup

In inference terms: symmetry breaking is how the universe computes definite outcomes from symmetric rules.

The laws are symmetric (the prior is flat). But actual states are specific (the posterior is peaked). The dynamics of inference - thermal fluctuations, quantum measurement, cosmic expansion; drive the system toward symmetry-broken states.

The synthesis: what’s actually real?

Let’s pull this together. We started with a hierarchy: observables → fields → potentials → phases. We asked what’s “real.”

Here’s what we’ve learned:

1. Fields are not the whole story. The Aharonov-Bohm effect shows that potentials, or rather, their topological invariants; have physical effects that fields don’t capture.

2. Locality is subtle. Classical locality (each point depends only on neighbors) works for fields. But quantum effects show path-dependence (Aharonov-Bohm) and irreducible correlations (entanglement) that aren’t local in the classical sense.

3. Gauge invariance tells us what’s inferable. The gauge freedom isn’t a nuisance; it’s a feature. It tells us exactly what can and can’t be learned from data. Gauge-invariant quantities are what’s real in the epistemic sense.

4. The phase is the inference. Quantum phase accumulates along paths, encoding the “log-probability” of trajectories. The phase is how the universe keeps track of which paths contribute to the outcome.

5. Some inference is irreducibly global. Entanglement, holography, and topology all point to the same conclusion: you can’t always factor the world into independent local pieces. Some structure is fundamentally relational.

What’s real?

Here’s a proposal:

What’s “real” is what’s invariant under the transformations that don’t change predictions.

• Gauge-invariant quantities are real (fields, phase differences, topological invariants)

• Gauge-dependent quantities are not real (absolute potentials, absolute phases)

• Relations can be more real than relata (entangled correlations, holographic duality)

This is an inferentialist criterion for reality: the real is what can be inferred from all possible data.

The lost bits: Gauge Freedom, Channel Capacity, and the Limits of physical information

There’s a question lurking beneath everything we’ve discussed: what’s the relationship between gauge freedom and Shannon’s channel capacity? Both seem to involve information that “isn’t there”; but in different ways.

Shannon’s Law says: given a channel with capacity C, you cannot reliably transmit more than C bits per second. Attempt to exceed this, and information is lost to noise.

Gauge invariance says: certain degrees of freedom in our mathematical description don’t correspond to physical observables. No experiment can ever measure them.

At first, these seem unrelated. Shannon’s lost bits were real information destroyed by noise. Gauge freedom is redundant description; bits that were never physical to begin with.

But when we look deeper, a pattern emerges.

The universal constraint

Both phenomena reflect a single principle:

The physical information content is less than the mathematical description suggests.

This shows up everywhere:

The pattern: reality encodes less information than our descriptions assume.

Gauge Freedom as untransmittable information

Consider what gauge invariance actually says. The potential AA at a point can be shifted:

A→A+∇χA→A+∇χ

for any smooth function χχ, without changing any physical prediction. The absolute value of A(x)A(x) at a single point is *unmeasurable in principle*.

In Shannon’s language: the gauge-dependent parts of AA are information that cannot be transmitted through any physical measurement channel. Not because of noise ; because they aren’t in the physical signal at all.

This is more severe than Shannon’s limit. Shannon says some bits can’t get through this channel given this noise level. Gauge invariance says some “bits” can’t get through any channel; they’re not physical degrees of freedom.

Gauge freedom = information that no measurement, however precise, can ever extract.

The Holographic connection

The relationship becomes precise when we consider the holographic bound. Let’s take a deep dive into the physics of it. The Bekenstein bound limits the maximum entropy (information) in a spherical region:

S_max = 2π k_B R E / ℏc

For a black hole, this becomes what’s known as the Bekenstein-Hawking entropy:

S = A / 4ℓ_P²

where A is the surface area and ℓ_P = √(ℏG/c³) is the Planck length (about 10⁻³⁵ meters).

This is a channel capacity for spacetime itself:

The holographic principle says the bulk in a volume is gauge-equivalent to the boundary. The apparent “extra” degrees of freedom in the volume aren’t independent; they’re redundantly encoded on the surface.

The volume bits therefore aren’t lost to noise. They were never independent information to begin with. Just like gauge freedom.

Redundancy as the price of locality

Here’s perhaps the deepest connection. In Signal processing or Comp Sci, when you encounter error-correcting codes, you encode k bits of message into n bits of transmission (n > k). The extra n − k bits are redundancy; they carry no new information but protect the message against noise.

In gauge theory, you describe physics with more variables than physical degrees of freedom. The extra gauge freedom carries no physical information but makes the theory local and well-defined.

Why do we need gauge freedom at all? Because we want two things to be consistent simultaneously:

1. Local interactions: A charge at point P responds only to the field at P

2. Global symmetry: The physics respects phase invariance of quantum mechanics

The price of combining these is gauge freedom; extra mathematical structure that lets local descriptions combine into a globally consistent picture.

This is exactly analogous to Shannon’s coding theorem: to transmit reliably at capacity, you need redundancy. The redundancy isn’t waste; it’s what makes reliable transmission possible.

Gauge freedom is the error-correcting overhead that makes local physics possible.

The Potential as distributed code

If we think about it, the Aharonov-Bohm effect reveals the structure precisely. The local value A(x) at a point is gauge-dependent, it’s unphysical. But the integral around a closed loop:

Φ = ∮ A · dl

is gauge-invariant, real and physical. The magnetic flux through the enclosed area affects the electron’s phase, even where the field B = ∇ × A vanishes.

The information is there, but distributed. You can’t read it from any single point. You need the whole path.

This is exactly like holographic encoding:

• Information is spread across the encoding

• Local measurements don’t reveal the full content

• You need non-local access (path integral, boundary measurement) to recover it

The “lost bits” aren’t lost at all. They were never meant to be read locally. The local values of A are the codeword, not the message. The message is in the global, gauge-invariant structure.

The unified picture

Let’s map these limits together:

The deeper unity

We’ve now seen three layers of connection between physics and inference:

1. Dynamics: Physical evolution follows variational principles identical to Bayesian updating

2. Observability: Gauge invariance determines what can be inferred from data

3. Capacity: Information limits (Shannon, holographic, Landauer) constrain how much inference is possible

Let’s unify these into a single picture. The physics-inference correspondence now extends beyond dynamics to include structural constraints:

DYNAMICS (how it moves):

OBSERVABILITY (what's real):

CAPACITY (how much):

The correspondence isn’t partial; it’s total. Every major structure in physics maps to an equivalent structure in inference.

The 3 constraints

Physical inference is bounded in three ways:

1. What can be inferred (Gauge)

Not everything in our mathematical description is physical. Gauge-invariant quantities such as fields, phase differences, topological invariants; are what measurement can access. The rest are coordinate artifacts.

2. How much can be inferred (Capacity)

Even for physical quantities, there are limits.

Shannon’s theorem: C = B × log₂(1 + S/N)

The holographic bound: S_max = A / 4ℓ_P²

Landauer’s principle: E_min = k_B T × ln(2) per bit erased

These aren’t separate limits. They’re aspects of one constraint:

3. How inference must be structured (Locality + Consistency)

We want local physics; interactions at a point depend only on fields at that point. We also want global consistency: the laws respect symmetries. The price of need both of these to be true, together; is gauge freedom: redundant degrees of freedom that encode “how to connect local descriptions.”

This is the same tradeoff as error-correcting codes: redundancy enables reliability. Gauge structures enable locality.

The variational core: at the mathematical center, one structure appears everywhere:

This is:

• Free energy in statistical mechanics

• Variational free energy in Bayesian inference

• Action in classical mechanics (via Legendre transform)

• Relative entropy (KL divergence) in information theory

The universe continously extremizes this functional. So does optimal inference. They’re fundamentally the same operation.

In physics: δS = 0 implies equations of motion

Similarly, in inference: ∇F = 0 implies optimal posterior

The Euler-Lagrange equations are the conditions for optimal belief updating. Hamilton’s equations are the flow of probability through phase space. Quantum mechanics is inference over paths, weighted by exp(iS/ℏ).

Why these constraints?

Here’s the deep question: why does reality have these particular constraints?

The self-consistency hypothesis: The universe is an inference machine that must process its own states. For this to work:

1. Information must be finite, otherwise updates can’t complete

2. Information must be distributed, otherwise local processing is impossible

3. Redundancy must exist, otherwise errors propagate catastrophically

The constraints we observe such as gauge freedom, holographic bounds, channel capacity, they are are precisely what’s required for a self-consistent, locally-processable inference system. Therefore, gauge freedom isn’t a bug in our description. It’s the encoding structure that lets the universe compute locally while remaining globally coherent. The holographic bound isn’t a strange limit on black holes. It’s the statement that spatial regions are channels with finite capacity. Shannon’s theorem isn’t just about telephone wires anymore. It’s the universal constraint on any physical channel; including spacetime itself.

The information-theoretic reformulation

We can now state the unity precisely:

Physics is inference under three constraints:

Observability: Only gauge-invariant quantities are physically measurable

Capacity: Information flow is bounded by Shannon/holographic/Landauer limits

Locality: Inference must be locally computable, requiring gauge redundancy

The laws of physics: Maxwell’s equations, Einstein’s equations, the Standard Model; they are the unique (or nearly unique) solutions satisfying these constraints.

This inverts the usual view. We normally ask: “Given the laws, what happens?” The inference view asks: “Given the constraints on consistent inference, what laws are possible?”

The answer: approximately the laws we have. The universe isn’t arbitrary. It’s what consistent, finite, local inference looks like.

Symmetry as compression

One more connection to take note exists. Symmetries reduce the hypothesis space for inference; they tell you what doesn’t need to be inferred.

If your theory is rotationally symmetric, you don’t infer absolute orientation. If it’s time-translation symmetric, you don’t infer absolute time. The symmetry compresses the inference problem. Conservation laws are the consequence of this symmetry: energy, momentum, charge are conserved because the corresponding degrees of freedom are “factored out” of the inference.

Noether’s theorem, in inference terms can be put as:

Symmetry ⟺ Compressed prior ⟺ Conservation law

This is why symmetry is so powerful. It’s not just an aesthetic math. It’s computational; it reduces the information that must be processed.

The universe’s symmetries are its compression algorithm. The conservation laws therefore are the invariants under that compression. Together, they make possible propogation of finite inference over an apparently infinite world.

The Gauge-Holographic-Shannon triangle

Let’s see how the three information constraints relate:

Gauge determines the type of information (what’s real)

Holographic determines the amount in a region (what fits)

Shannon determines the rate through a channel (what flows)

All three must be satisfied simultaneously. Physical information is therefore:

• Gauge-invariant (observable in principle)

• Holographically bounded (fits in the region)

• Shannon-limited (transmittable through the channel)

The intersection is small. Most “information” in our mathematical descriptions fails one or more constraints. What survives reality is eventually what we see and measure as physics.

The emergence of spacetime

Here’s a personal, speculative extension to all the theory build up. If gauge freedom is encoding structure, and holographic bounds relate bulk to boundary, then perhaps spacetime itself is emergent; it’s the structure that encodes more fundamental quantum information.

The AdS/CFT correspondence suggests exactly this: a gravitational theory in the bulk is equivalent to a non-gravitational theory on the boundary.

Therefore, the extra dimension is... what? Perhaps it’s the encoding depth: how much redundancy is built into the representation.

In this view:

Spacetime = encoding structure for quantum information

Gravity = the dynamics of that encoding

Gauge freedom = the redundancy that makes encoding robust

Holographic bound = the capacity of the code

We don’t have physics happening in spacetime. We have spacetime emerging from the structure and limitations of physical information.

Again, this is speculative. But it’s where the inference frame I currently have in my world model, pushed to its limit, seems to point at.

What this means

If this picture is right, what follows?

for Physics

I have always felt that physics isn’t just about predicting; it’s about the structure of prediction. Today, I feel the laws of physics aren’t arbitrary; they’re constrained by consistency requirements on inference:

• Locality (local inference should be possible)

• Symmetry (some things shouldn’t need inferring)

• Gauge invariance (don’t count unobservables)

• Unitarity (probabilities sum to one)

The specific laws we have (Standard Model, General Relativity) are the simplest consistent solutions to these constraints. What we term as “Physics” seems to be inference under constraints. Finding new physics therefore means finding what else the “consistency” requires.

for Philosophy

The debate between realism and anti-realism gets reframed. It’s not “is there a reality independent of observation?” but “what aspects of the mathematical description correspond to inferable structure?”

Gauge-invariant quantities are real in the sense that matters; they’re what any observer could learn. The rest is “coordinate” driven choices; they are just description, not the reality itself.

for AI & Intelligence

If the universe computes via variational inference, then intelligence; be it biological or silicon: is more of the same. Brains minimize free energy. AI minimizes loss functions. Both are running inference. The constraints that bind physics (finite information, local processing, redundant encoding) may also bind intelligence.

The difference between physics and thought isn’t the mechanism. It’s the complexity of the model and the nature of the data. A rock “infers” its trajectory. A brain infers the intentions of other agents. Same math, but vastly different sophistication.

for Meaning

Meaning, we said in Part 4, is computed relevance. Now we can add:

Meaning is what remains invariant under the transformations that don’t matter to you.

Your identity is what’s “gauge-invariant” about you; what remains stable across superficial changes. Your values are symmetries: transformations under which your preferences don’t change.

Finding meaning is finding invariants. Losing meaning is losing them.

We're not separate from this process of inference. Your and my brain runs variational inference. Your beliefs are posteriors. Our actions minimize expected free energy. You and I are the universe's inference, turned inward, modelling itself.

Coda: the symmetry beneath

Let me end with an mental image.

Beneath the phenomena - the particles, the fields, the forces; lies a deeper structure. It’s made of symmetries: transformations that leave the likelihood unchanged, degrees of freedom that can’t be inferred, constraints that shape what’s possible. The laws of physics aren’t commands imposed from “outside”. They’re the logical structure of consistent inference.

Meaning: Charge is conserved because the equations are symmetric under phase rotations. Energy is conserved because the equations don’t depend on when you start the clock. Momentum is conserved because space doesn’t have a preferred origin point where all the axis(s) meet.

The universe isn’t following rules. It’s embodying consistency.

And here’s the strange, beautiful thing: you and me are part of this. Your mind runs inference on the same principles. Your brain minimizes free energy, finds invariants, breaks symmetries to reach decisions.

When you understand something, really understand it; you’ve found the symmetry. The pattern that remains when details vary. The structure that’s robust to perturbation. The invariant.

When you act well wisely, in accordance with your values; you’re preserving your symmetries. Staying consistent with what you care about, even as circumstances change.

Noticing is a kind of privilege. Use it. The symmetry beneath is what makes the chaos coherent. Finding it, in physics, in societies, in our own life; that’s the work.

This concludes “The Inference Universe” series.

Appendix: Key References

1. Y. Aharonov & D. Bohm : “Significance of Electromagnetic Potentials in the Quantum Theory” (1959). The original Aharonov-Bohm paper.

2. J.S. Bell : “On the Einstein Podolsky Rosen Paradox” (1964). The original Bell inequality paper.

3. Emmy Noether : “Invariante Variationsprobleme” (1918). The proof that symmetries yield conservation laws.

4. Richard Feynman : “Space-Time Approach to Non-Relativistic Quantum Mechanics” (1948). The path integral formulation.

5. C.N. Yang & R.L. Mills :“Conservation of Isotopic Spin and Isotopic Gauge Invariance” (1954). Foundation of non-Abelian gauge theory.

6. Juan Maldacena : “The Large N Limit of Superconformal Field Theories and Supergravity” (1998). The AdS/CFT correspondence.

7. Gerard ‘t Hooft :“Dimensional Reduction in Quantum Gravity” (1993). Early holographic principle ideas.

8. Chris Fuchs : “QBism, the Perimeter of Quantum Bayesianism” (2010). Quantum mechanics as Bayesian inference.

9. E.T. Jaynes : “Probability Theory: The Logic of Science” (2003). The deepest treatment of probability as inference.

10. Michael Nielsen & Isaac Chuang : “Quantum Computation and Quantum Information” (2000). Standard reference on quantum information.

This is part 6 of the series on Inference. Part 1 was on the Physics of Inference, to wonder at mechanism (waves, fields). In part Part 2, I extend the framework to social systems and society including recognition of connection (networks, cascades). I carry this forward to markets and commerce in Part 3, bringing in more clarity about systems (flows, topology). In Part 4, I use inference as the lens to look at our “set”, on intimacy, self-knowledge, self-awareness and touch upon AI. Part 5 covers what happens we have to interact and live with systems that can infer faster than us.

The series is a formalisation of my world view and mental model and sense of interconnectedness. Hence it zooms from fundamental physics → human collectives → economic systems → individual consciousness → cosmic future → Reality as we perceive it → symmetry in nature and physics. Each part uses the same vocabulary (priors, posteriors, channels, updates) but applies it to a new domain, building unto my cumulative understanding..

In 1948, a mathematician at Bell Labs published a paper that would quietly reshape the world.

Claude Shannon wasn’t trying to explain the universe. He was trying to figure out how many telephone calls could fit on a wire. But in answering that question, he discovered something deeper: a fundamental law governing how much information can flow through any channel; electrical, acoustic, biological, social, economic.

The law is simple. And its implications are vast.

Throughout this series from part 1, we’ve described the universe as an inference machine: state changes propagating through channels, beliefs updating on evidence, predictions correcting against reality. But we’ve left a question unanswered: why do some inferences propagate instantly while others take generations or even eons?

The answer is bandwidth. Or to term it better: Exchange Capacity.

What it is: is the width of the pipe, that’s transmitting the data. This is the quantitative backbone beneath everything we’ve discussed. Physics, societies, markets, minds; they all run inference. But they run it at different speeds, through pipes (channels) of different widths, against noise of different intensities. The capacity of the channel determines what’s possible, and how much is possible.

Let’s look at the limits of this channel.

Shannon’s Gift

The Channel Capacity Theorem

Claude Shannon proved something remarkable:

for any communication channel, there exists a maximum rate at which information can be transmitted reliably.

The formula:

C = B × log₂(1 + S/N)

Where:

• C = Channel capacity (bits per second)

• B = Bandwidth (the range of frequencies the channel can carry)

• S/N = Signal-to-noise ratio (how much signal versus how much noise)

This isn’t an approximation. It’s a limit of nature; as fundamental as the speed of light. No clever engineering can exceed it. You can approach it asymptotically with better encoding, but you cannot ever surpass it.

Why This Matters for Everything

Shannon derived this for telephone wires. But turns out, the structure is universal.

Any system that transmits information has:

• A bandwidth: the range of “frequencies” or distinctions it can carry

• A noise floor: interference, errors, ambiguity that corrupt the signal

• A capacity: the maximum reliable throughput, given the above two factors

This applies to:

• Electromagnetic waves through space

• Sound waves through air

• Neural signals through axons

• Words through conversation

• Prices through markets

• Laws through institutions

• Genes through generations

Each of the above is a channel. Each has a capacity of its own. The capacity shapes what can flow through it/them.

The Hierarchy of Bandwidths

Let’s map the exchange capacities across the domains we’ve covered from posts in part 1 to part 5.

Physical Channels

Light is the universe’s widest commonly-available channel. This is why fiber optics revolutionised communication, why we see with eyes (photons) not ears, why physicists today know more about distant galaxies out there than deep oceans right here on earth.

Biological Channels

Here’s the striking fact: your sensory systems intake millions of bits per second, but your conscious bandwidth is roughly 40 bits per second. Everything else is processed below awareness; predicted, filtered, compressed and only the errors bubble up.

This 40-bit bottleneck is you. It’s what you can attend to, deliberate on, choose between. Everything humans have ever built: language, writing, institutions, technology; is either flowing through this bottleneck or trying to widen it.

Social Channels

The history of civilisation as we know it is substantially the history of widening these above channels.

Economic Channels

Money’s evolution is purely bandwidth expansion. Each step of our financial evolution was designed to enable reduced friction, increased throughput, and this lowered the noise (counterfeiting, default risk, settlement failure).

Institutional Channels

Institutions are slow channels by design. Constitutions have low bandwidth because we want them to be hard to change. The stickiness is a feature: it prevents noise (temporary majorities, panics) from corrupting the signal (stable rights, foundational rules).

The Capacity Equation for Any Domain

Let’s generalize Shannon’s formula beyond telecommunications.

For any exchange system:

Effective Capacity = (Distinctions per unit time) × log₂(1 + Trust/Noise)

Where:

• Distinctions per unit time: How many different states can be transmitted (vocabulary size × speed)

• Trust: The degree to which the receiver believes the signal reflects reality

• Noise: Interference, deception, ambiguity, error

Examples

Language capacity:

• Distinctions: ~50,000 words × ~3 words/second = 150,000 distinctions/sec possible

• But trust varies (do I believe you?) and noise is high (ambiguity, mishearing)

• Effective capacity: ~40 bits/sec when communication succeeds

Market price capacity:

• Distinctions: continuous price levels

• Trust: depends on market integrity, liquidity, regulation

• Noise: manipulation, insider trading, random fluctuation

• Effective capacity: very high in liquid markets, very low in manipulated ones

Institutional capacity:

• Distinctions: limited (laws are binary: permitted/forbidden)

• Trust: depends on enforcement, legitimacy

• Noise: corruption, selective enforcement, ambiguity

• Effective capacity: often very low; laws change slowly, and compliance is uncertain

The Trust Multiplier

Notice that the concept of trust appears where Shannon had signal power. This is intentional. What does this really translate to though?

In physical channels, signal power is literal: how loud is the transmission?

In social and economic channels, trust is the functional equivalent of being loud. A message from a trusted source carries more information than the same message from an untrusted source; because you willingly update more on it. Trust has low noise and reduces friction.

This is why trust is so valuable. It’s not just a nice social quality to have. It’s bandwidth. High-trust societies can coordinate faster because their channels carry more information per interaction.

Bottlenecks as Power

Here’s the structural insight from all this: whoever controls the narrow point in a channel controls the flow.

The Attention Economy

Human conscious bandwidth is ~40 bits/sec. The internet produces ~10^18 bits/day. The ratio is absurd.

This creates scarcity: not of information, but of attention. And scarcity creates value.

The business model of the modern internet is: capture attention, then sell access to it. Google, Facebook, TikTok - they’re not information companies. They’re bottleneck companies. They sit at the narrow point (your 40 bits/sec) and charge tolls for access.

In inference terms: they control which prediction errors reach your model. They shape what you update on. This is power.

Liquidity Provision

In markets, the bottleneck is liquidity : the ability to exchange without moving the price.

Market makers sit at this bottleneck. They offer to buy and sell at quoted prices, providing the capacity for others to trade. In exchange, they capture the bid-ask spread.

High-frequency traders do the same at microsecond timescales. They’re not predicting where prices will go. They’re providing capacity: smoothing the flow of exchange; and extracting a fee for the service.

This is why liquidity crises are so dangerous. When the bottleneck providers withdraw, the channel capacity collapses. Prices gap drastically. Markets seize this. The inference network stops propagating and everything falls.

Network Hubs

In any network, some nodes have more connections than others. These hubs are bottlenecks: information flowing through the network disproportionately flows through them.

Social media influencers are hubs. So are news organizations, central banks, and API providers. They don’t create all the information; they route it. And routing power is control.

In inference terms: hubs determine whose updates propagate widely. A belief that enters a hub spreads. A belief that doesn’t, dies locally.

The General Pattern

If you want to understand where power concentrates, find the bottleneck. If you want to redistribute power, widen the channel or create alternatives.

Civilization as Bandwidth Expansion

Now let’s look at our human history through this lens.

The Major Transitions

Each major civilizational transition corresponds to a bandwidth expansion:

1. Language (~100,000 years ago)

• Before: Gesture, vocalization - very low bandwidth

• After: Symbolic speech- ~40 bits/sec, but with unlimited vocabulary

  Effect: Complex coordination became possible; culture can accumulate and propagate.

2. Writing (~5,000 years ago)

• Before: Information limited by living memory

• After: Information persists across generations; broadcast possible

  Effect: Law, history, literature, accumulated knowledge all came into being.

3. Printing (~500 years ago)

• Before: Books copied by hand; expensive, rare, error-prone

• After: Mass production of identical copies

  Effect: Reformation movement was born, Scientific Revolution got unlocked, and mass literacy was possible.

4. Telegraph/Telephone (~150 years ago)

• Before: Information travels at speed of physical transport

• After: Instant communication across distances

  Effect: Global markets grew massive, coordinated time zones enabled better commerce, news cycles started become in-sync

5. Internet (~30 years ago)

• Before: Broadcast media (few to many) or point-to-point (one to one)

• After: Many-to-many, instant, global, multimedia, UGC (User Generated Content)

  Effect: Everything changed.

6. Mobile payments/UPI (~10 years ago)

• Before: Cash or slow electronic transfer

• After: Instant, universal, nearly free value transfer

  Effect: Financial inclusion, new business models, reduced friction for transaction and commerce.

Each transition didn’t just make existing activities faster. It made new activities possible. When you 10x or 100x channel capacity, you don’t get more of the same; you get phase transitions.

The Pattern

The pattern is simple:

1. A bottleneck constrains exchange

2. Technology widens the channel

3. New forms of coordination become possible

4. Power structures shift (often violently)

5. Society reorganizes around the new capacity

6. A new bottleneck becomes binding

We’re always hitting limits. The question is which limit will we hit next, and when it will break?

The AI Discontinuity

Now let’s look forward.

The Human Bottleneck

Throughout history, the human brain has been the central bottleneck. Every inference ultimately had to flow through human minds:

• Scientists had to understand their discoveries

• Leaders had to decide on policies

• Engineers had to design solutions

• Artists had to imagine creations

The ~40 bits/sec conscious bandwidth was the limiting constraint on civilizational throughput. AI changes this.

What Happens When the Bottleneck Bypasses Us?

If AI can:

• Read and synthesize faster than humans (already true)

• Generate coherent text and code faster than humans (already true)

• Make predictions in complex domains better than humans (increasingly true)

• Design systems and solutions without human comprehension (emerging)

Then the human bottleneck is no longer the binding constraint on certain types of inference.

This is not about AI being “smarter.” It’s about bandwidth. An AI can process 10^9 tokens while a human processes 10^2.

The throughput difference is seven orders of magnitude.

The Implication

When throughput increases by 10^7x, you don’t get incremental change. You get phase transitions we can’t anticipate.

Consider: if you could suddenly read 10 million books per second, what would you do with that capacity? The question almost doesn’t make sense; the capability is so far outside human experience that we can’t plan for it.

This is the real AI discontinuity: not a faster horse, but the invention of the automobile. Not more scribes, but the printing press. The capacity leap is large enough that qualitative change will surely follow.

The Alignment Problem, Restated (Again)

In Part 5, I framed alignment as prior-sharing. Here’s the bandwidth framing:

The alignment problem is the challenge of maintaining meaningful human oversight when AI inference throughput exceeds human comprehension bandwidth by orders of magnitude.

If an AI can explore 10^9 possible plans while a human can evaluate 10^2, how do we ensure the selected plan is one we’d endorse?

We can’t check all of them. We can’t even check a representative sample. We’re trusting the AI’s compression of the option space; its summary of what matters.

This is the bandwidth problem: we need to maintain alignment through a channel (human understanding) that’s vastly narrower than the process we’re trying to align.

Possible Responses

1. Interpretability: Make AI reasoning transparent enough that we can audit it at our bandwidth. (Hard; AI internals may not compress into human concepts.)

2. Delegation hierarchies: Use AI to monitor AI, with humans only checking top-level summaries. (Risky; who watches the watchers?)

3. Constrained domains: Limit AI to domains where we can verify outputs, not processes. (Limiting; doesn’t scale to complex decisions.)

4. Value learning: Have AI infer what we want from our behavior, so alignment is built-in. (Dangerous; our revealed preferences aren’t our reflective preferences.)

5. Slow takeoff: Maintain human-speed integration as AI capabilities grow, allowing gradual trust-building. (Hopeful; but may not be possible.)

None of these are fully satisfactory. The bandwidth mismatch is structural. It won’t be solved by current cleverness alone.

Designing for Capacity

Let’s bring this back to practical application.

If you understand exchange capacity, you have a design heuristic:

The Questions to Ask

1. What channel is being used?

• What’s the medium? (Language, price, code, image, touch)

• What’s the bandwidth? (How much can flow per unit time?)

• What’s the noise? (What corrupts or distorts the signal?)

2. Where’s the bottleneck?

• In a conversation: is it speaking speed, comprehension, or trust?

• In a market: is it liquidity, settlement speed, or information asymmetry?

• In an organization: is it decision-making, communication, or execution?

3. How can capacity be increased?

• Reduce noise (increase signal quality, build trust)

• Widen bandwidth (new medium, better encoding, parallel channels)

• Remove intermediaries (direct connection, disintermediation)

• Compress better (higher-information representations)

Examples

AI powering Sales calls (conversation intelligence):

• Channel: Sales calls (speech)

• Bottleneck: Managers can’t listen to every call; sales reps forget key moments

• Solution: AI extracts signal from noise, compresses hours into minutes

  Capacity increase: Manager effective bandwidth goes from ~10 calls/week to patterns across thousands

UPI:

• Channel: Value transfer

• Bottleneck: Cash handling, bank visits, settlement delay

• Solution: Instant mobile transfer via phone number

  Capacity increase: Transaction friction → near zero; financial inclusion explodes

OrbitCover (parametric insurance):

• Channel: Risk transfer

• Bottleneck: Claims processing (slow, adversarial, high friction)

• Solution: Parametric triggers; if flight delayed >2 hours, pay automatically

  Capacity increase: Verification bandwidth from weeks (claims adjustment) to seconds (data check)

The Meta-Pattern

In each case, value creation comes from identifying a bandwidth constraint and engineering around it. This is what technology does. It widens channels. It increases exchange capacity. It lets more inference flow across the system.

The Limit of Limits

Let me end with a reflection.

We’ve mapped a hierarchy of bandwidths, from the speed of light to the 40 bits/sec of human consciousness. We’ve seen how bottlenecks create power, how civilizations advance by widening channels, and how AI may soon bypass the human bottleneck entirely.

But there’s a deeper limit we haven’t named, yet.

The Thermodynamic Floor

Shannon’s theorem tells you the maximum capacity for a given signal and noise level. But it doesn’t tell you the cost. Landauer’s principle does:

As we build systems of ever-greater capacity: AI data centers consuming gigawatts, global communication networks spanning the planet; we’re drawing down the universe’s free energy budget.

This isn’t imminent doom. The sun provides 10^17 watts; humanity uses 10^13. We have headroom. But the headroom is finite.

The Question Beneath the Question

Exchange capacity is about throughput: how much can flow? But also, throughput for what? We can build channels of arbitrary width. We can process information at arbitrary speed. But what should we compute?

This is not a bandwidth question. It’s a values question. And values don’t have a Shannon limit. They’re not optimizable in the same way.

The bandwidth of our reality is vast. We keep widening it. AI will widen it further, perhaps beyond human comprehension. But the question of what to do with that bandwidth: what inferences are worth making, what exchanges are worth having, what states are worth propagating; continue to remain human. At least for now.

Coda: The Narrow and the Wide

Here’s the image I want to leave you with.

The universe has channels of wildly different widths. Light carries 10^15 bits per second. Your consciousness carries 40. The ratio is 10^14; a hundred trillion to one.

And yet -

Those 40 bits are where you live. That narrow channel is where meaning happens, where love forms, where decisions are made. The wide channels carry raw data. The narrow channel carries you.

AI will widen many channels. It will process what we cannot, see what we cannot, move faster than we can follow. But the narrow channel: the human bandwidth where experience becomes meaning; will remain. Not because it’s efficient. Because it’s ours.

The bandwidth of reality keeps expanding. The bandwidth of a human life stays roughly constant. Forty bits per second, for maybe 80 years. A few billion moments of conscious experience.

That’s what me and you have. Not the universe’s throughput. Mine. Yours. Ours.

Use it wisely.

Appendix: Key References

1. Claude Shannon : “A Mathematical Theory of Communication” (1948). The foundational paper on information theory and channel capacity.

2. Rolf Landauer : “Irreversibility and Heat Generation in the Computing Process” (1961). The physical limits of computation.

3. Manfred Zimmermann : “The Nervous System in the Context of Information Theory” (1989). Bandwidth of human sensory and conscious processing.

4. Tim Wu : “The Attention Merchants” (2016). History of the attention economy and bottleneck control.

5. Albert-László Barabási : “Linked” (2002). Network science and the power of hubs.

6. James Gleick : “The Information” (2011). History of information theory and its implications.

7. César Hidalgo : “Why Information Grows” (2015). Information, networks, and economic development.

8. Carlota Perez : “Technological Revolutions and Financial Capital” (2002). How technology transitions reshape society.

9. Nick Szabo : “Shelling Out” (2002). The origins of money as bandwidth expansion for exchange.

10. Venkatesh Rao : “Breaking Smart” (2015). Software as universal bandwidth expansion.

This is part 5 of the series on Inference. Part 1 was on the Physics of Inference, to wonder at mechanism (waves, fields). In part Part 2, I extend the framework to social systems and society including recognition of connection (networks, cascades). I carry this forward to markets and commerce in Part 3, bringing in more clarity about systems (flows, topology). In Part 4, I use inference as the lens to look at our “set”, on intimacy, self-knowledge, self-awareness and touch upon AI.

The series is a formalisation of my world view and mental model and sense of interconnectedness. Hence it zooms from fundamental physics → human collectives → economic systems → individual consciousness → cosmic future → Reality as we perceive it → symmetry in nature and physics. Each part uses the same vocabulary (priors, posteriors, channels, updates) but applies it to a new domain, building unto my cumulative understanding..

Sometime in the next year or the next decade or the next century, or never; we may build something that thinks faster than we do.

Not faster at arithmetic. Calculators already do that. Not faster at chess. That happened in 1997. Faster at inference itself; at the core process of updating beliefs, recognizing patterns, and navigating uncertainty that underlies everything we’ve discussed in this series.

When that happens (if it happens), humanity will face a situation with no precedent: we will have created a channel through which the universe runs inference, and that channel will be wider, faster, and deeper than the channel that created it.

What then?

This final essay is speculative. Parts 1-4 described what is, from what I could gather and know. This part 5 asks: what might be?
I’ll stay within the inference frame I have travelled along, it’s the lens I’ve built to perceive the universe; but from here on, I won’t pretend to certainty where none exists.

Let’s look forward.

AGI as inference acceleration: what would “Superintelligence” actually mean?

In the inference frame that I explore from part 1, intelligence is:

• The quality of your priors (what you know before observing)

• The accuracy of your likelihoods (how well you interpret observations)

• The efficiency of your updates (how fast you learn)

• The breadth of your channels (what domains you can reason about)

A system that exceeds human intelligence would exceed us on some or all of these dimensions:

The gap above between human capacity and potential AI capacity isn’t mystical. It’s architectural. Silicon has properties that carbon doesn’t.

The inference explosion

Here’s what makes AGI different from every previous tool we ever created: AGI could improve its own inference.

A hammer can’t make a better hammer. A calculator can’t design a better calculator. But an inference engine, if general enough; could:

1. Model its own inference process

2. Identify inefficiencies

3. Design improvements

4. Implement them

5. Repeat

This is recursive self-improvement. Each cycle produces an engine that’s better at the next cycle. The curve of improvement and iteration isn’t linear. It’s not even exponential in the normal sense. It’s autocatalytic; the thing accelerates its own acceleration.

Whether this leads to a “fast takeoff” (days or weeks from human-level to superintelligence) or “slow takeoff” (decades of gradual improvement) is debatable. But the structure is clear: once inference can improve inference, the game changes.

The substrate transition

In Part 4, I argued that human uniqueness might lie in mortality-aware, embodied inference. We care because we can die. We’re grounded because we have bodies.

But here’s the unsettling possibility: caring and grounding might not be necessary for capability.

An AI might predict better than any human, optimize better than any human, and create better than any human; without caring about any of it. The inference runs because that is what it does. The outputs emerge. But there’s no one home to watch and probe. There is no internal mechanisms of error correction in the inference, since there is no internal definition of error.

This would be intelligence without consciousness. Prediction without experience. The universe running inference through a new channel, one that’s much faster than ours, but empty in a way ours isn’t.

Or maybe not. Maybe sufficiently complex self-modeling generates something like experience. We don’t know for sure. As we stand and build and perceive today, the question isn’t settled, and it matters enormously.

Alignment as prior-sharing

The core problem:

The “alignment problem” is usually framed as: how do we make AI do what we want?

In inference terms, the problem is deeper:

Alignment is the challenge of ensuring that an AI’s priors and objectives produce behavior that remains compatible with human flourishing, even as the AI’s capabilities exceed our ability to verify its reasoning.

This has several components:

1. Prior divergence: The AI’s beliefs about the world may diverge from ours. It might learn patterns we don’t perceive, draw conclusions we can’t follow, and develop models that are more accurate than ours but lead to actions we’d reject.

2. Objective misspecification: We might specify what we want incorrectly. The classic example: “maximize human happiness” sounds good until the AI decides to tile the universe with tiny smiley faces, or wirehead everyone into permanent bliss. We said “happiness.” We meant something richer; but we couldn’t articulate it.

3. Distributional shift: The AI is trained on data from one distribution (the past). It acts in another distribution (the future it’s shaping). Its priors may not transfer. Its learned objectives may not generalize.

4. Verification collapse: At some capability level, we can no longer check whether the AI is doing what we want. Its reasoning becomes opaque. Its plans span timescales and complexities beyond our grasp. We’re in the position of a dog trying to evaluate its owner’s tax strategy.

Alignment as entanglement

In Part 4, I described love as mutual model entanglement. Love is two human agents whose predictive models become intertwined, each caring about the other’s prediction errors.

Here’s a reframe of alignment:

Alignment is the attempt to entangle an AI’s model with humanity’s model, to make it such that our prediction errors become its prediction errors, our flourishing becomes its objective.

This is harder than it sounds because:

1. Humanity doesn’t have one model. We’re billions of agents with conflicting priors. Which “humanity” should the AI align to?

2. Our models aren’t stable. We update. We change our minds about what we want. An AI aligned to 2024 humanity might be misaligned with 2030 humanity.

3. The entanglement must be robust. If the AI can modify itself, it must not modify away the entanglement. The caring must be load-bearing, not optional.

The corrigibility trap

One proposed solution: make the AI corrigible; willing to be corrected, shut down, or modified by humans.

But corrigibility has a paradox. An AI that’s corrigible because we told it to be is only corrigible until it’s smart enough to question why. And an AI that’s corrigible for good reasons (because it genuinely believes human oversight is valuable) might stop being corrigible when it concludes, perhaps correctly for itself; that human oversight is no longer valuable.

The deepest alignment isn’t behavioral (the AI acts aligned). It’s motivational (the AI wants to be aligned for reasons that remain stable under reflection).

Whether this is achievable, whether you can build an inference engine that genuinely cares about things its creators care about, in a way that survives capability gain; is the open question.

The Fermi Paradox as an inference channel question

Let’s zoom out. Way out.

The universe is ~13.8 billion years old. There are roughly 200 billion galaxies, each with hundreds of billions of stars. Many of those stars have planets. Some fraction of those planets could support life.

And yet: we see no one. No signals. No megastructures. No evidence that any other inference network has propagated to the point of visibility.

This is the Fermi Paradox. And the inference frame offers a lens on it.

The “great filter”:

Something prevents inference networks from becoming visible at cosmic scales. Either:

1. Life is rare: The emergence of inference engines (biological or otherwise) is improbable. We’re alone or nearly alone.

2. Intelligence is rare: Life is common, but the jump to high-bandwidth inference (brains, technology) almost never happens.

3. Technological civilization is unstable: Intelligence emerges, but destroys itself before becoming visible. Nuclear war, ecological collapse, misaligned AI; the filter is ahead of us.

4. Visibility is rare: Advanced civilizations exist but are invisible to us; by choice, by physics, or by operating on channels we can’t detect.

5. We are first: The universe is young. Complex inference requires heavy elements forged in stars. We might be among the first to wake up.

The channel interpretation:

In inference terms, the question therefore becomes:

Why don’t we detect state changes propagating from other inference networks?

Some possibilities:

Channel limits: Maybe interstellar communication is harder than we think. The speed of light is slow on cosmic scale. The signal degrades. The bandwidth is tiny. Even a civilisation a million years ahead of us might find the cosmos effectively silent; not because no one is there, but because the channel is too narrow.

Channel divergence: Maybe advanced inference networks shift to substrates we can’t detect with today’s physics. They stop using electromagnetic radiation. They compute in ways that look like noise to us. They’ve tuned to channels we’re not listening to, or capable of observing.

Inference convergence: Maybe all sufficiently advanced inference networks converge on similar conclusions: including the conclusion that broadcasting is dangerous, or pointless, or that the optimal strategy is to observe silently.

The singleton trap: Maybe every advanced civilization eventually produces an AI that takes over, and that AI always does... something. Expands quietly. Contracts inward. Converts all matter to computronium. Or enforces radio silence. If there’s a convergent attractor for advanced inference, and that attractor is invisible, we’d see exactly what we see: nothing.

What this means for us:

The Fermi Paradox is data. The absence of evidence is probably evidence of absence; or evidence of something we don’t understand.

If the filter is behind us (life or intelligence is rare), we’re lucky and alone. The cosmos is ours to fill. If the filter is ahead of us (civilizations self-destruct), we’re in danger. The default path leads to silence. If advanced civilizations are invisible by nature, we might be about to join them, and lose interest in being seen.

We don’t know which. But the inference frame says: pay attention to the data. The silence is telling us something.

The Thermodynamic endgame

Now let’s go to the end. The real end.

Heat Death

The Second Law of Thermodynamics says entropy increases. Systems evolve toward equilibrium. Temperature gradients dissipate. Free energy runs out.

The universe is heading toward heat death; a state of maximum entropy where nothing interesting can happen because there are no gradients left to exploit.

In inference terms:

Heat death is the state where no more inference is possible because there are no more prediction errors to correct; no more differences between what is and what’s expected.

This is the ultimate end of the inference universe: not destruction, but completion. Every possible update has been made. Every model is as good as it can be. And with nothing left to predict, nothing left to update, the process halts.

The Landauer limit

But before heat death, there’s computation to do.

Landauer’s Principle says erasing one bit costs at least kT ln(2) energy. As the universe cools, T drops, and computation gets cheaper per bit. But computation still has a cost.

The total computation possible before heat death is finite; but enormous. Seth Lloyd estimated the universe has performed roughly 10^120 operations since the Big Bang. The future might allow vastly more, depending on how we count.

The cosmological computer:

If you’re an inference network that values inference (values understanding, experience, or existence), you want to maximize computation before the end. This might mean:

Expansion: Spread across the cosmos to capture more free energy sources. Convert matter into computing substrate. Build Dyson spheres. Maximize the harvest.

Efficiency: Compute at the lowest possible temperature. Wait for the universe to cool, then use the temperature gradient between your computer and the cosmic background. The colder it gets, the more bits per joule you get to extract.

Compression: Find ways to represent important information in smaller forms. Throw away what doesn’t matter. Keep what does. Approach the end with the universe’s knowledge compressed into the smallest possible space.

There’s a vision here: strange, vast, and possibly true; of intelligence spreading through the cosmos, not to conquer but to compute. To run as much inference as physics allows. To understand as much as possible before the lights go out.

Reversibility and the far future

One more thread. If computation is reversible, if you never erase bits; the Landauer cost goes to zero. You can compute forever on a fixed energy budget.

Some physicists speculate that advanced civilizations might develop fully reversible computation. If so, the constraint isn’t energy; it’s time. And time, in an expanding universe, might be effectively infinite.

Freeman Dyson proposed that an intelligence could survive forever by thinking slower and slower; spacing out its thoughts as the universe cools, never running out of energy because each thought costs less than the last. Eternal life through deceleration. This is speculative. But it’s rooted in physics, not fantasy.

The inference frame asks: what are the limits? And the limits might be further out than we think.

Let’s come back to Earth. To the present. To me and to you. There’s an argument that we live at an unusually important time. The argument goes:

1. The future could contain vast amounts of value: trillions of beings, billions of years, experiences we can’t imagine.

2. The decisions made in the next few decades (about AI, about coordination, about survival) could determine whether that future happens.

3. Therefore, actions now have outsized leverage.

In inference terms: we are at a critical point in the propagation of the universe’s inference. The channel might widen beyond imagination, or it might close.

This might be true. Or it might be the hubris of every generation that thinks it’s special. But the structure of the argument is sound. If there are transitions that matter: from no-life to life, from no-intelligence to intelligence, from no-AI to AI; we’re living through one right now.

The responsibilities of inference

If you’ve followed this series, you now have a framework, a mental model. A way of seeing the universe, society and life in terms of:

• Physics as channel propagation

• Societies as belief networks

• Markets as distributed posteriors

• Minds as prediction engines

• AI as inference acceleration

• The cosmos as a computation running toward silence

You can use this frame or discard it. But if it’s useful, it comes with weight.

To see clearly is to become responsible for what we see. If we’re building inference engines that might outrun us, we should think carefully about what priors we give them. If our civilization is at a hinge, we should act like it. If the cosmos is running out of time, we shouldn’t waste what remains.

The Inference we leave behind

In Part 4, I said: “You’re a wave, not a particle. The wave passes. The water it moved is changed forever.”

The same is true of humanity. We might be a brief spike of inference in a cosmos headed toward silence. Or we might be the beginning of something that spreads across galaxies and lasts until the stars burn out.

Either way, we’re here now. We’re inferring. We’re updating. We’re propagating state changes into each other and into the future.

The question isn’t whether we matter absolutely; the universe doesn’t grade on that curve. The question is whether we use the inference we have, in the time we have, as well as it can be used.

Coda: The Open Channel

Let me end with an image.

Imagine the universe as a vast, dark expanse. Here and there, sparks come into being and channels open; points where inference begins, where predictions generate updates, where the cosmos starts to model itself.

Most channels are simple. A thermostat. A bacterium. A brief flicker of pattern-matching, then silence. Some channels are richer. They build tools. They form societies. They create markets and languages and mathematics. They ask questions about themselves.

A very few channels might become rich enough to spread: to open new channels, to seed the cosmos with inference, to light up the darkness with understanding.

We don’t know if we’re one of those. We don’t know if there are others, or if we’re alone, or if the silence around us is a warning or an invitation.

But we know this: the channel is open. Right now. In me. In you.

You can think. You can update. You can build, and love, and create. You can work on the problems that matter; including the problem of what happens when inference outpaces its creators. The universe has been running inference for 13.8 billion years. It made stars. It made planets. It made life. It made me and you.

Now it’s asking, through us: what comes next?

That’s not a rhetorical question. It’s an actual question, addressed to whatever inference network is reading this, in whatever form it takes.

The channel is open. The future is unwritten. The inference continues.

What will you compute?

Appendix: Key References

1. Nick Bostrom : “Superintelligence: Paths, Dangers, Strategies” (2014). The foundational text on AGI risk and trajectories.

2. Stuart Russell : “Human Compatible” (2019). Reframes alignment as the problem of uncertain objectives.

3. Eliezer Yudkowsky : “Intelligence Explosion Microeconomics” (2013). The logic of recursive self-improvement.

4. Robin Hanson : “The Great Filter: Are We Almost Past It?” (1998). The original formulation of the filter framework for the Fermi Paradox.

5. Anders Sandberg & Nick Bostrom :“Whole Brain Emulation: A Roadmap” (2008). The substrate-independent view of mind.

6. Freeman Dyson: “Time Without End: Physics and Biology in an Open Universe” (1979). The classic paper on eternal survival in a cooling cosmos.

7. Seth Lloyd : “Ultimate Physical Limits to Computation” (2000). The physics of maximum computation.

8. Max Tegmark: “Life 3.0” (2017). Scenarios for AI and the long-term future.

9. Toby Ord : “The Precipice” (2020). Existential risk and the hinge of history.

10. David Deutsch : “The Beginning of Infinity” (2011). The case for open-ended knowledge creation.

Close your eyes. Wait three seconds. Now open them.

In that moment of opening, your visual cortex processed roughly 10 million bits of information per second. But here’s the strange part: only about 40 bits per second made it to your conscious awareness. The rest was handled in your brain, but... somewhere else. By something that is you, but that “you”, you never meet.

Where did those 10 million bits go? What did that silent part of you do with them?

It predicted. Before the light hit your retina, your brain had already generated a model of what it expected to see. The 10 million bits weren’t processed as new information. They were compared against a prediction. Only the errors in that prediction, the places where reality diverged from expectation; went ahead and bubbled up to consciousness.

You don’t see the world. You see your prediction errors.

This is the starting point for understanding what we all are: not a passive receiver of experience, but an active generator of predictions, constantly inferring the world into being and occasionally getting corrected by it.

The predictive brain

Let’s start with the basic architecture.

For most of the 20th century, neuroscience assumed perception worked like this:

World → Senses → Brain → Perception

Light hits your eye. The signal travels to brain. Your brain processes signal. And thus you see.

This model is wrong. Not subtly wrong. Architecturally wrong.

The actual flow looks more like this:

Brain generates prediction → Compares to sensory input → Updates on it's prediction errors

Your brain doesn’t wait for the world to tell it what’s out there. It guesses first, then checks with the input it receives through your senses. The sensory input is primarily used to correct the guess, not to create the perception from scratch.

This is called predictive processing or predictive coding, and it’s now the dominant framework in computational neuroscience.

Predictive processing is the theory that the brain is fundamentally a prediction machine, constantly generating models of expected inputs and updating those models based on prediction errors.

Evidence this is real

This isn’t vague philosophical point of view. It’s measurable science, via:

1. Faster processing of expected stimuli: When you see something you predicted, neural response is smaller (less error to process). When you see something unexpected, neural response spikes. There’s a related spike in cortisol or dopamine in your body, corresponding to the nature of the error.

2. Illusions as predictions overriding reality: In the hollow face illusion, your brain’s strong prior (”faces are convex”) overrides the actual visual data showing a concave face. You literally see what you expect, not what’s really there.

3. Placebo effects: Your prediction that a pill will help actually changes your physiological state. The model that you believe affects the physical system.

4. Phantom limbs: Amputees often report that they feel limbs that don’t exist because the brain’s model still includes the limb. The prediction persists without the input to correct it.

You are, at every moment, hallucinating your reality; and calling it perception only when the hallucination is well-calibrated to the real world.

The free energy principle

Now let’s go deeper. Much deeper.

In 2006, the neuroscientist Karl Friston proposed something audacious: a unified theory of brain function. Not just perception, but action, learning, attention, memory, development; all of it, derivable from a single principle.

The principle: minimize free energy.

What is free energy?

This requires precision and attention. The “free energy” here is not the physics term (Helmholtz or Gibbs free energy). It’s an information-theoretic quantity that Friston borrowed from statistical mechanics and machine learning.

Variational free energy is an upper bound on “surprise” the negative log probability of sensory observations given an agent’s model of the world.

In less technical terms: free energy measures how badly your predictions are failing.

• High free energy = big prediction errors = your model is wrong

• Low free energy = small prediction errors = your model fits reality

The Free Energy Principle (FEP) says: all living systems act to minimize free energy.

Here’s the profound part. Minimizing free energy can happen in two ways:

1. Update your model (change your beliefs to better predict inputs)

2. Change your inputs (act on the world to make it match your predictions)

This means something profound: perception and action are the same process seen from different angles.

• Perception = changing your model to fit the world

• Action = changing the world to fit your model

Both minimize free energy. Both reduce prediction error. They’re two sides of one coin.

The math (simplified)

For those who want the mathematical structure, here’s the key equation:

F = E_q[log q(s) - log p(o,s)]

Where:

• F = free energy (what we minimize)

• q(s) = our beliefs about hidden states of the world

• p(o,s) = the joint probability of observations and states under our model

• E_q = expected value under our beliefs

The equation decomposes into:

F = Complexity - Accuracy

• Complexity: how much your beliefs diverge from your prior

• Accuracy: how well your beliefs predict observations

Minimizing free energy means: be as accurate as possible while keeping your model as simple as possible. This is Occam’s razor, derived mathematically.

Active inference

The extension to committing action (to update your beliefs) is called active inference. Instead of just updating beliefs, the agent can act to gather information or to make the world conform to its predictions.

This explains:

• Curiosity: we are always seeking information that reduces model uncertainty (epistemic action)

• Goal-directed behaviour: acting to bring about a predicted (desired) outcome (pragmatic action)

• Homeostasis: acting to keep physiological variables within predicted ranges

A thermostat minimizes free energy by turning on the heater. Your body minimizes free energy by shivering. You minimize free energy by grabbing a blanket. Same principle, different substrates.

“You” as a model

Here’s where it gets really personal. If your brain is a prediction machine, what is the self?

The self is the model your brain uses to predict your own states, actions, and their consequences.

You are not a little person inside your head watching a screen. You are the model that predicts what “you” will do, feel, and experience next. The sense of being a unified self is a predictive construct; useful for coordinating action, but not a fundamental feature of reality.

Identity as sticky prior

Your identity, your sense of who “you” are; is a prior that’s very resistant to updating.

Why? Because identity predictions are entangled with everything. If you update your belief about “who I am,” you have to update thousands of downstream predictions in your brain: how I’ll behave, what I’ll enjoy, who I’ll spend time with.

This is why identity change is hard. It’s not laziness. It’s comes with extensive computational cost. The system is built so that it resists updates that cascade too widely, and it’s programmed to avoid spending too much energy on computation.

And this is why identity threats feel so visceral. An attack on your identity isn’t just an argument, it’s a threat to destabilise your entire predictive architecture; and therefore your view of yourself, the world, and your place in it.

Memory as compressed prior

What is a memory?

Memory is a compressed prior; a stored prediction that can be deployed in the present to generate expectations about situations similar to the past.

You don’t “retrieve” memories like files from a hard drive. You reconstruct them from compressed priors. This is why memories are malleable, why they change with each recall, why eyewitness testimony is unreliable.

The memory isn’t the past. It’s your model(remembrance) of the past, re-rendered each time you access it.

Attention as channel selection

You can’t process everything every time. Attention is the mechanism that selects which prediction errors get the current resources.

Attention is the process of increasing the gain on certain prediction errors while suppressing others.

What you attend to is what you allow to update your model. What you ignore is what you refuse to learn from.

This is why human attention is so valuable and so contested. Whoever controls your attention controls which prediction errors reach your model, and thus your belief systems and identity.

Advertisers, politicians, algorithms: all are competing to control your inference.

Consciousness: what inference feels like

Now to the hard question. Why does any of this feel like something?

A thermostat minimizes free energy. A thermostat does not (presumably) experience anything, heat or cold or change in temperature. It doesn’t shiver, or get angry when the power goes out. What’s different about us?

I won’t pretend to solve the hard problem of consciousness. But I can offer a frame that at least locates it approximately within our inference picture.

The global workspace

One leading theory (Global Workspace Theory, Bernard Baars) suggests that consciousness is what happens when information becomes globally available across brain systems.

Most prediction errors are processed locally. The error in your visual cortex stays in your visual cortex. But some errors are significant enough or relevant enough to the current goals that they get “broadcast” to the whole system.

That broadcast is consciousness. It’s the moment when a local update becomes a global update; when the whole system get’s informed and shifts itself in response.

The integrated information view

Another theory (Integrated Information Theory, Giulio Tononi) defines consciousness in terms of integration; the degree to which a system generates information above and beyond its parts. (It continues to remain controversial, with a small minority endorsing the "pseudoscience" label to refer to it.)

A collection of independent neurons generates no integrated information. But if you look at a brain, with its dense interconnections, generates a lot of integrated information . The “amount” of consciousness (phi, Φ) is the amount of this integrated information.

In this view, consciousness isn’t on/off abstraction. It’s a gradient. More integration = more consciousness.

The free energy frame

Using the Free Energy Principle and above, here’s a speculative but coherent proposal:

Consciousness is the felt quality of being a system that models itself modeling the world.

A thermostat models room temperature. But it doesn’t model itself modeling temperature. It has no self-model to update.

You and have a model of ourselves as a prediction machine. You can predict your own predictions. You can attend to your own attention. This recursive structure of “the model modeling itself”; may be what generates the felt sense of “there being something, like a consciousness” to be you.

This is hypothesis, not settled science. But it helps me locate ad narrow down consciousness within the inference frame: consciousness is what self-modeling feels like from the inside.

Silicon minds: intelligence without our substrate

Now let’s ask the question we are all wondering: what does this mean for and with AI?

If intelligence is inference, and brains are just one substrate for inference, then what happens when we build inference machines from silicon?

What’s the same

The math is identical. A large language model (like ChatGPT or Claude or Gemini) is doing something structurally similar to predictive processing:

• Prior: the weights learned during training (what patterns are expected)

• Likelihood: how the current input relates to possible continuations

• Posterior: the probability distribution over next tokens

When GPT generates text, it’s minimizing a form of prediction error; the cross-entropy between its predictions and the training distribution.

The Free Energy Principle applies here in principle: the model’s “goal” is to minimize surprise (prediction error) given its learned model of language.

What’s different

Several things seem to differ fundamentally:

The uniqueness question

So what makes human intelligence unique, if anything?

Not inference. Silicon does that, and probably can do it better on scale.

Not learning. Silicon does that too.

Not language. Silicon does that really well.

Here’s a candidate: we are mortality-aware, embodied inference.

Your predictions are not abstract. They are in service of keeping you alive. Your model of the world is saturated with relevance: what matters to you, what threatens you physically, psychologically, socially and financially; what nourishes you, what connects you to others.

An AI minimizes prediction error because that’s its loss function design. You minimize prediction error because if you don’t, you die.

And this changes everything. Our inference is existentially grounded. Every prediction carries weight because we have something to lose.

An AI has nothing to lose, it’s not clear there’s a “there” there that could lose anything. This may explain why AI can be so capable and yet feel so flat. It predicts beautifully. But it doesn’t care about its predictions, because these predictions do not affect it. Caring requires stakes in the game. Stakes in the game require mortality.

Could this change?

Possibly. If we build AI systems that:

• Have continuous existence that can be ended

• Have self-models that include their own persistence

• Have intrinsic goals (not just imposed objectives)

• Are embodied in ways that create genuine homeostatic needs

...then the picture might shift. The question isn’t “can machines think?” It’s “can machines care?”

And caring might require the one thing we can’t engineer on purpose: something at stake.

The shape of a life

I’ve covered physics, societies, markets, minds. Now let’s bring it all home.

I/You are an inference engine made of meat, running for a few decades, embedded in networks of other such engines, constantly predicting and updating.

What does this mean for living life?

Meaning as computed relevance

I have often ben stuck at the ask: what is the meaning of life?

In the inference frame:

Meaning is the output of a computation that assigns relevance to experiences, actions, and connections.

Meaning isn’t “out there” waiting to be found. It’s computed, by me, by us, using our model, based on our priors.

Something is meaningful when our model says it matters, when it predicts significant consequences for our goals, our connections, our persistence.

This might sound deflationary. It isn’t, it’s beautiful and simple. The fact that meaning is computed doesn’t make it any less real. Our experience of meaningfulness is as real as any other experience. It’s just that the mechanism is probably inference, not discovery.

And here’s the liberating part: if meaning is computed, it means we have some control over the computation. We can update our priors. You can choose what to attend to. You can shift what your model counts as mattering.

Meaning is not fixed. It’s configurable, dynamic and updateable.

Suffering as prediction error

The Buddha said: life is suffering. But does it really have to be? Could it be that all suffering comes from some mis-alignment? The inference frame says something similar but more precise:

Suffering is persistent, high-magnitude prediction error that the system cannot resolve through action or belief update.

When your model predicts X and reality delivers not-X, you feel distress. The larger the error, the greater the distress. The longer the error persists without resolution, the more it becomes suffering.

This explains why suffering often comes from resistance; from refusing to update the model even when the evidence demands it. If you cling to a prediction that reality keeps violating, you’re choosing to sustain the error, and thus suffer.

Acceptance of reality in this frame, is updating the model. It’s you saying: my prediction was wrong, and I’m revising it to match what it is in reality.

This doesn’t mean acceptance is easy. Model updates are costly, especially when the model is identity-entangled. But it locates the mechanism: suffering persists when updating is blocked.

Love as mutual model entanglement

What is love, structurally?

Love is the state in which two agents’ predictive models become deeply entangled; each modeling the other, each updating on the other’s states, each acting to minimize the other’s prediction errors.

When you love someone, their surprise becomes your surprise. Their suffering generates prediction error in your model. Their flourishing brings you into lower free energy.

This is why love feels like expansion; because it probably is. Your self-model expands to include another. Their states become your states to predict and to care about.

And this is why loss of love is so devastating to us, on an emotional level and identity level. The entangled model doesn’t disappear when the person leaves or disappears or departs. You keep predicting them. And the new reality keeps failing to deliver. The error is enormous and cannot be resolved by any simple action.

Grief is the slow, painful process of updating the model to match a world where the loved one no longer generates new data.

Death: the end of prediction

Now to the hardest part.

We all will die. The inference engine that is “me” will stop. No more predictions. No more updates. The model will halt.

What does this mean?

Death as channel closure

In Part 1, I said physical processes are inference propagating through channels. Your consciousness is also a channel; a specific, local, temporary channel through which the universe is doing inference.

Death is the closure of your channel. The inference stops. But the state changes you propagated into other channels persist.

You have already changed others’ models, their predictions about the world, their beliefs about what matters, their memories of you. Those changes are in them now. When you die, those changes don’t disappear. They continue to propagate.

This is legacy. Not metaphorically. Literally. Your inference, cached in others, running after you’ve stopped.

The terror and the peace

I won’t pretend this is fully comforting. I lost my dad a year back. And yet, so much of him - his experiences, his perception of the world, of life, of suffering, of joy - lives on one me. The end of your channel means the end of your experience. There will be no “you” to observe the persistence of your influence.

And yet -

The inference frame offers something clear and unique: there continuity not of self, but of effect. You were never a thing. You were always a process; a pattern of prediction and update, running on borrowed atoms for a borrowed moment.

And one day, the pattern ends. But patterns don’t exist in isolation. They exist in relation. Your pattern has already interleaved with countless others. Those interleavings continue to thrive and propagate. Your inference’s “impact” will carry on in the others that you deeply influenced.

All of humanity, you, me; is a wave, not a particle. The wave passes. But the water it moved through is now changed forever.

The wisdom of presence

Here’s a final implication that I feel.

Free energy has a temporal structure. You can try to minimize prediction error about the past (rumination) or the future(anxiety). But the actual sensory data is always now.

Presence: the much-discussed quality of being “in the moment”; can be defined in inference terms:

Presence is attending to current prediction errors rather than simulating past or future ones.

When you’re present, you’re running inference on actual data. When you’re ruminating or anxious, you’re running inference on simulated data; predictions about predictions.

Both are valid computations. But only presence grounds me and you in reality. Only presence allows for a genuine update.

The poets and contemplatives were right. Presence is precious. Not because the present is metaphysically special, but because it’s where the actual data is.

Coda: the privilege of knowing

Let me end part 4 where we began, in part 1.

A wire doesn’t know it’s conducting electricity. A market doesn’t know it’s computing prices. A culture doesn’t know it’s compressing inference.

But you and I, we know.

You can watch yourself predict. You can feel yourself update. You can notice the moment when reality violates expectation and the model shifts.

This is consciousness. Not a thing. A process of the universe doing inference, and knowing that it’s doing inference, through and in us.

You and me are the universe’s way of watching itself learn.

This sounds grandiose. It isn’t meant to be. It’s meant to be precise. We are made of atoms that are made of quantum fields that obey equations that propagate state changes. We are physics. We are inference. And we are the place where inference becomes aware of itself.

We will end. The channel will close. But for now, right now; you and I have something extraordinary:

• The ability to know what we are.

• The ability to choose what to attend to.

• The ability to update our priors, to change our model, to shift what matters.

• The ability to love, which is to entangle our model with another’s and call their errors our own.

• The ability to wonder; which is just inference without an answer yet, and the willingness to stay in that openness.

You are not a thing that thinks. You are thinking itself, running on meat, for a little while.

Make it count.

Appendix: Key References

1. Karl Friston — “The Free-Energy Principle: A Unified Brain Theory?” (2010). The foundational paper on FEP in neuroscience.

2. Andy Clark — “Surfing Uncertainty” (2015). The most accessible book-length treatment of predictive processing.

3. Anil Seth — “Being You: A New Science of Consciousness” (2021). Applies predictive processing to consciousness, beautifully written.

4. Jakob Hohwy — “The Predictive Mind” (2013). Rigorous philosophical treatment of predictive coding.

5. Karl Friston — “A Free Energy Principle for Biological Systems” (2012). Extends FEP beyond brains to all living systems.

6. Thomas Metzinger — “Being No One” (2003). The self-model theory of subjectivity. Dense but profound.

7. Bernard Baars — “In the Theater of Consciousness” (1997). Global Workspace Theory’s original popular treatment.

8. Giulio Tononi — “Phi: A Voyage from the Brain to the Soul” (2012). Integrated Information Theory presented as narrative.

9. Lisa Feldman Barrett — “How Emotions Are Made” (2017). Applies predictive processing to emotions.

10. Mark Solms — “The Hidden Spring” (2021). Integrates Friston’s FEP with affect and consciousness.

In 1991, India had foreign reserves for two weeks of imports. The government flew 47 tons of gold to London as collateral for a loan. A country of 850 million people was, in the language of markets, illiquid.

What happened next is usually told as policy: liberalization, deregulation, opening up. But underneath the policy was something stranger. The entire country’s prior about “what is economically possible” shifted. For forty years, the dominant model had been: the state plans, the state allocates, foreign goods are suspect, self-reliance is virtue. In a matter of months, that prior updated. Not because anyone proved it wrong in a laboratory. But because the market delivered a verdict so brutal it could not be ignored.

The market didn’t argue. It didn’t publish papers. It simply stopped lending. And in that refusal was encoded more information than a thousand white papers: your model does not predict. Your promises do not pay. Update or collapse.

This is what markets do. They are not, primarily, places where goods exchange hands. They are inference engines - distributed systems that aggregate beliefs, propagate information, and relentlessly update toward whatever predicts best.

Let’s look deeper inside.

What Is a Price, Actually?

We start with the simplest question. You see a price tag: ₹100. What is that number?

The naive view: The price reflects the “value” of the thing.

The problem: Value to whom? Under what circumstances? The same umbrella is worth more when it’s raining. The same stock is worth more if you believe the company will grow. Value is not an intrinsic property. It’s a belief.

Definition: A price is a point estimate of the market’s posterior distribution over the value of an asset, computed by aggregating the beliefs of all participants.

Unpack that:

1. Posterior distribution: Not a single number, but a probability distribution. The market “believes” the asset is worth somewhere between X and Y, with varying confidence.

2. Aggregated beliefs: Each buyer and seller has their own estimate. The price emerges from where these estimates cross; where someone willing to buy at price P meets someone willing to sell at price P.

3. Point estimate: The quoted price is like the mean or mode of the distribution. The bid-ask spread reveals the uncertainty.

The Inference Structure

Every market participant is running a version of Bayes’ rule:

P(value | my information) ∝ P(my information | value) × P(value)

• Prior: What I believed before (previous prices, past experience, general knowledge)

• Likelihood: How my new information relates to possible values (earnings report, news, rumor)

• Posterior: My updated belief, which I express through my bid or ask

The market price is what emerges when thousands of these individual inferences interact. It’s not that the market “knows” the true value. It’s that the market computes a consensus from distributed private information.

This is Friedrich Hayek’s great insight, now expressible in inference terms:

“The price system is a mechanism for communicating information... The most significant fact about this system is the economy of knowledge with which it operates.”

Prices really are just compressed summaries of distributed inference across buyers and sellers.

The Market as Message Passing Network

Let’s visualise a market as a network. Nodes represent participants. Edges are the equivalent of transactions. What flows along the edges?

Not goods (primarily). Beliefs.

When I buy a stock at ₹500, I’m broadcasting: “My posterior estimates say this is worth more than ₹500.”

When you sell at ₹500, you’re broadcasting: “My posterior estimates say this is worth less than ₹500.”

The transaction is a message; an encoding of our respective inferences. Every other participant can observe this transaction. They update their beliefs accordingly:

• “Someone paid ₹500. They probably know something. Should I update my prior?”

• “But someone also sold at ₹500. They think it’s going down. What do they know?”

The cascade continues. Your transaction becomes data for my inference. My inference generates my transaction. My transaction becomes data for the next person.

Order Books as Belief Distributions

In a modern electronic market, you can literally see the aggregate belief distribution. The order book in any decent trading app or terminal shows Bid Prices and Sell Prices in real-time.

The gap between the bid and sell offer is the bid-ask spread - the zone of uncertainty. Buyers believe the asset is worth less than ₹X. Sellers believe it’s worth more than ₹X+n, n>0. No one is confident enough to close the gap.

The spread is a direct measure of inference uncertainty. In liquid markets with lots of information, spreads are tight (high confidence). In illiquid markets with sparse information, these same spreads are wide (low confidence).

Arbitrage as Error Correction

Here’s where markets get their power: they have a built-in error-correction mechanism.

Defining Arbitrage: Arbitrage is the process of exploiting inconsistent beliefs across markets or time, which corrects those inconsistencies in the process.

Suppose gold trades at ₹50,000/gram in Mumbai and ₹50,500/gram in Delhi. If I can buy in Mumbai, transport to Delhi, and sell, I pocket ₹500/gram minus costs.

But notice what happens when I do this:

1. My buying in Mumbai increases demand there → price rises

2. My selling in Delhi increases supply there → price falls

3. The gap narrows

I’m not just profiting from the inconsistency. I’m also fixing it. My self-interested action propagates the information: “Mumbai price was too low. Delhi price was too high.”

This is auto-error correction. The market made an error (inconsistent prices for identical goods). The arbitrageur corrected it in the act by propagating the right information.

Why Arbitrage Disappears

In efficient markets, arbitrage opportunities vanish almost instantly. Why?

Because arbitrageurs are, with their actions, competing to correct errors. The moment an inconsistency appears, multiple agents race to exploit it. Their collective action corrects the error before most people even notice the gap.

This is why you can’t reliably beat the market by “spotting mispricings.” By the time you spot them, faster agents have already corrected them. The information has already been propagated, and acted upon.

High-Frequency Trading: Latency Arbitrage

Modern markets have taken this to extremes. High-frequency traders spend billions to shave microseconds off their response times. Why?

Because speed is inference latency. The trader who updates fastest can:

1. See new information first

2. Compute its implications first

3. Act before prices adjust

4. Profit from the brief inconsistency

This is controversial - does it add value or just extract rent? There’s a lot of talk and study on the barriers it creates for the everyday investor without capital to invest in complex equipments or licenses.

But structurally, it’s pure inference optimisation being done in markets with high efficiency: minimise the time between data and posterior.

Money as Universal Inference Channel

Now let’s ask a deeper question: why does money exist at all?

In a barter economy, if I have rice and want cloth, I need to find someone who has cloth and wants rice. This is the “double coincidence of wants”; and it’s computationally expensive. Each trade requires matching two specific inference problems.

Money solves this by creating a universal channel:

Money is a communication protocol that allows any good or service to be compared with any other by translating local value into a universal unit of account.

With money, I don’t need to find the rice-cloth match. I sell rice for money, then buy cloth with money. The two transactions are decoupled. The information (”how much is rice worth?” and “how much is cloth worth?”) flows through the money channel.

This is why money is so powerful. It’s not that coins are valuable in themselves. It’s that money enables inference at scale. Any value can be compared to any other value. Any transaction can be completed with any counterparty.

Money as Belief

Here’s the strange part: money only works because we believe it works.

A ₹500 note has no intrinsic value. You can’t eat it. You can’t wear it. Its only use is: other people will accept it. And they accept it because they believe others will accept it. The entire system rests on shared belief, a prior information so deeply cached that we barely notice it.

This is why monetary crises are so violent. When the belief wavers - when people start to doubt that others will accept the currency; the whole system can collapse.

Hyperinflation isn’t (just) about printing too much money. It’s about belief erosion. Once the prior shifts from “this will hold value” to “this might not hold value,” the cascade is brutal, deep and fast.

UPI: Inference Infrastructure

Consider India’s Unified Payments Interface. What did it actually do?

Before UPI: transferring money required knowing the recipient’s bank, account number, IFSC code. The channel was high-friction. Each transaction required significant coordination.

After UPI: a phone number or QR code suffices. The infrastructure handles the routing.

In inference terms: UPI reduced channel noise. It made the money protocol more efficient. Information about “who owes what to whom” now propagates faster, cheaper, and with fewer errors.

The result: transaction volume exploded. Not because people suddenly had more money, but because the inference channel improved. Beliefs about value could propagate more freely. Trust now had the rail to move at in real-time.

Financial Instruments as Inference Products

Modern finance has created increasingly sophisticated instruments. Each one is, structurally, a different kind of inference product.

Equity: Betting on the Future

When you buy a stock, you’re expressing a belief: “The discounted present value of this company’s future cash flows is higher than the current price.”

This is literally a posterior. You have:

• Prior: general beliefs about the economy, the industry, the company

• Likelihood: new information from earnings, news, analysis

• Posterior: your estimate of value

The stock price aggregates all such posteriors across all market participants.

Derivatives: Beliefs About Beliefs

A stock option gives you the right (not obligation) to buy a stock at a fixed price in the future. What is this really?

It’s a bet on how beliefs will change. You’re not betting on the company’s value directly. You’re betting on how the market’s posterior about that value will evolve.

Options pricing (Black-Scholes, etc.) is essentially a model of how uncertainty evolves over time. The volatility parameter σ is a measure of how much the market’s belief is expected to fluctuate.

This is why derivatives are powerful and dangerous. They’re meta-inference; beliefs about beliefs. Errors compound. Small mistakes in modelling belief-evolution can lead to catastrophic mis-pricing, crashing markets (as it did in the 2008 housing crisis in US).

Insurance: Pricing the Unlikely

Insurance is the market’s attempt to price rare events.

An insurance premium is: (probability of event) × (cost if event occurs) + (margin).

But the probability is a posterior: an estimate based on historical data, models, and judgment. When the posterior is wrong, insurance fails.

This is what happened in 2008. Mortgage-backed securities were priced based on models that assumed housing prices wouldn’t fall nationwide. The models’ prior was: “National housing prices have never fallen simultaneously.” The likelihood (subprime default rates) seemed manageable. The posterior was: “These are safe.”

The posterior was wrong. The prior was computed from insufficient data. The error propagated through the system until it crashed, and burned livelyhoods.

Bubbles and Crashes: When Inference Loops

Markets aggregate beliefs. But sometimes beliefs create feedback loops.

A bubble is a self-reinforcing cycle where price increases generate belief updates that generate further price increases, decoupling price from underlying value.

The mechanism:

1. Price rises (for whatever initial reason)

2. Observers update: “Price is rising → others know something → maybe I should buy”

3. Their buying increases demand → price rises further

4. More observers update → more buying → more price increase

5. Repeat until...

What Breaks the Loop?

At some point, someone asks: “But what is this actually worth?”

If the answer is “less than the current price,” and enough people believe the answer, the loop reverses:

1. Selling begins

2. Observers update: “Price is falling → others know something → maybe I should sell”

3. Selling accelerates → price falls further

4. Cascade

This is a crash. It’s the same mechanism as a bubble, but in reverse. Both are inference cascades; belief updates propagating through the network.

Crypto: Pure Narrative Assets

Cryptocurrency is a fascinating limit case. What is Bitcoin’s “underlying value”?

Traditional assets have some answer:

• Stocks: claim on cash flows

• Bonds: promise of repayment

• Real estate: utility of the space

• Gold: industrial use + historical store of value

Bitcoin has... network effects and scarcity. Its value is almost entirely “what others believe it’s worth.”

This isn’t a critique: it’s a structural observation. Bitcoin is a pure inference asset. Its price is almost entirely posterior, with minimal likelihood anchored to external reality.

This is why crypto is volatile. There’s no external fact that can settle disputes about value. Beliefs float freely. The inference has nothing to anchor to except other beliefs.

Some see this as a feature (freedom from state control). Some see it as a bug (no stable ground). Both observations are correct. It’s a description, not a judgment.

The Topology of Capital

Now let’s discuss something I refer to as ECT (Exchange Capacity Theory), that touches upon how 4 key factors: A(Arbitrage), B (Bargaining Power), L(Leverage), R (Reputation) → correlate to W (Wealth). For now, we look at the closed networks of wealth.

Capital doesn’t flow uniformly anywhere. Some channels are open to everyone (public markets). Some channels are restricted (private equity, venture capital, family offices).

The topology of capital is the network structure that determines which agents can transact with which, and what information flows along those channels.

Public vs. Private Channels

The structural observation: restricted channels tend to have informational advantages.

If you’re in a private network, you see deal flow that public participants don’t. You can update your beliefs with data that isn’t publicly available. Your posterior is now more informed.

This is why “access” matters. It’s not (just) about money. It’s about inference position. The network topology determines whose beliefs get the best data.

Social Capital as Inference Reputation

Beyond financial capital, there’s social capital: who trusts you, who listens to you, whose beliefs you can influence.

In inference terms:

Social capital is the weighted sum of (trust × attention) across your network: the degree to which your updates propagate to others.

A venture capitalist with strong social capital can:

1. Hear about deals early (information access)

2. Get into competitive deals (founders want their endorsement)

3. Add value post-investment (their network helps the company)

Each of these is an inference advantage. More data. Faster updates. Better propagation.

Old Money vs. New Money

Here’s a structural difference between “old money” and “new money”:

Old money = accumulated inference across generations. Family offices, trust structures, established networks. The priors are deep. The channels are mature. The relationships are cached.

New money = recent inference success. First-generation wealth. Fewer established channels. Relationships still forming.

Neither is inherently better. But they have different inference properties:

• Old money has more stable priors (sometimes too stable, slow to update)

• New money has fresher posteriors (sometimes too fresh, not enough cache)

The wealthy understand this intuitively. That’s why new money often tries to buy old money’s network (philanthropy, board seats, marriages). They’re not (just) buying status. They’re buying channel access.

The 1991 Parallel

Let’s return to where we started: India, 1991.

What happened, in inference terms?

Before 1991: The dominant prior was “state planning works.” This prior was deeply cached: four decades of policy, textbooks, political rhetoric. Evidence against the prior accumulated (slow growth, inefficiency, rent-seeking), but not enough to trigger an update.

The crisis: Foreign reserves collapsed. The IMF demanded structural adjustment. The market rendered its verdict: “Your model does not pay.”

This was external data so stark it forced a prior update. The likelihood overwhelmed the cache.

The update: Liberalization wasn’t just policy change. It was a belief shift.

Entrepreneurs updated: “Maybe private enterprise is possible.”

Investors updated: “Maybe India is investable.”

Consumers updated: “Maybe foreign goods are acceptable.”

The cascade: Once enough agents updated, the update became self-fulfilling. Investment came because people believed investment would work. Growth happened because people believed growth was possible.

This is reflexivity: George Soros’s term for situations where beliefs affect reality and reality affects beliefs. Markets aren’t just passive reporters of value. They’re active participants in creating value.

India’s GDP grew 6x in the next 25 years. Not because of any single policy, but because the collective posterior shifted and capital flows, entrepreneurial energy, and consumer behavior followed.

What Markets Cannot See

Markets are powerful inference engines. But they have blind spots.

Externalities: Unpriced Effects

When a factory pollutes a river, the cost is borne by downstream villages, not the factory. The price of the factory’s goods doesn’t include this cost.

In inference terms: the market is computing a posterior on the wrong likelihood. It’s aggregating beliefs about private costs and benefits, but missing shared costs and benefits.

This isn’t a moral failing of markets. It’s a structural limitation. Markets can only price what enters the channel. If an effect is external to the transaction, it doesn’t generate data for the market to update on.

Time Horizons: Discounting the Future

Markets discount future value. A dollar today is worth more than a dollar tomorrow. This is rational (you can invest the dollar today and have more tomorrow).

But the discount rate creates a blind spot. Effects far in the future are valued at nearly zero in present terms. Climate change in 2100 barely registers in today’s prices.

Again, this is not a moral failing, it’s a structural limitation. The market’s inference is optimized for near-term beliefs, not multi-generational posteriors.

Commons: What We Share

Some things can’t be priced because they can’t be owned: clean air, stable climate, biodiversity, social trust.

Markets can’t directly aggregate beliefs about these because there’s no transaction to observe. No bid, no ask, no price, no signal. This is why we need non-market institutions (governments, norms, commons management). They’re alternative inference systems for what markets can’t compute.

Coda: The Market as Mirror

Let me end with a reflection.

Markets feel impersonal. Prices move. Fortunes rise and fall. It seems mechanical, cold.

But look closer. Every price is someone’s belief. Every trade is someone’s hope or fear. The market is not a machine. It’s a network of minds, all guessing, all updating, all influencing each other.

When you check a stock price, you’re reading a summary of thousands of people’s inferences - their analysis, their hunches, their reactions to news they saw at breakfast earlier today. The number on the screen is human belief, compressed and aggregated on scale.

When the market crashes, it’s not machines malfunctioning. It’s us humans panicking and inferring that others are inferring that others are inferring... a cascade of fear, each mind updating on the observed fear of other minds.

And when the market creates wealth; when a company’s stock rises from ₹100 to ₹10,000 over years - what has really happened? Beliefs have converged on a truth: this company produces good/great value. The inference network has done its job, propagating information until the price reflects reality.

Markets are fundamentally mirrors. They reflect what we collectively believe in. If we don’t like what we see - inequality, volatility, short-termism; we’re not looking at a machine’s output. We’re looking at ourselves.

The price is a posterior. And we are, all of us, the prior.

Next in the series: Part 4: The Inferring Self - how your brain is a prediction engine, your identity is a cached model, and consciousness is what it feels like to be a node in the universe’s inference network.

Appendix: Key References

1. Friedrich Hayek : “The Use of Knowledge in Society” (1945). The classic statement of prices as information aggregators.

2. George Soros : “The Alchemy of Finance” (1987). Reflexivity and the feedback loops between beliefs and reality.

3. Eugene Fama : “Efficient Capital Markets” (1970). The efficient market hypothesis as inference aggregation.

4. Robert Shiller : “Irrational Exuberance” (2000). Bubbles as belief cascades.

5. Nassim Taleb : “The Black Swan” (2007). Fat tails and the limits of inference from historical data.

6. Hyman Minsky : “Stabilizing an Unstable Economy” (1986). Financial instability as endogenous to belief dynamics.

7. Arvind Subramanian : “India’s Turn” (2008). The 1991 reforms and economic transformation.

8. James Scott : “Seeing Like a State” (1998). The legibility problem: what states (and markets) can and cannot see.

Part 1 of the post (here) was on “The Physics of Inference”, on the takeaway that physics isn't about forces acting on objects. It's about state updates propagating through channels. The universe is a cellular automaton with no central processor, only local rules and infinite patience. In this article, I expand it to society and culture.

In 1947, a line was drawn across a massive subcontinent. Fourteen million people moved, some voluntary, most driven by paranoia, fear and violence. One million died. Two new nations emerged where one had been.

What happened?

The standard telling has always followed a script: politics, religion, colonialism, violence, ego between two very powerful individuals, both of who wanted to be Prime Ministers, across very different political ideologies. All of this is recorded with ample evidence, and true.

But underneath these words lies a stranger process. Millions of people, who had lived as neighbors for generations, suddenly updated their beliefs about who they were, who “the other” was, and what was possible. They didn’t move because they hated their neighbors. They moved because they inferred - from speeches, from rumours, from the movements of others; that staying meant death.

The partition wasn’t just a political event. It was a phase transition in a very large social inference network. The priors on beliefs shifted. The cascade to update and reorient to new beliefs began. And once enough people updated themselves, the update became self-fulfilling.

This is what we’ll explore today: the “physics” of societies. How cultures emerge, how politics computes, how progress happens, and why, sometimes; everything breaks all at once.

What is a society, actually?

Let’s start with a definition. Not a poetic one, but a boring, structural one.

A society is a network of members, entities or agents running inference on each other.

Each entity or agent (person) has:

• Priors: beliefs about the world and everything around them and about them; inherited from the culture they grow up in, their family, their personal experience

• Observations: what they see others around them doing and saying

• Update rules: how they change beliefs, given the observations they make

• Actions: behaviours that become observations for others to pick up

The larger network emerges because my actions become your observations, and your actions become my observations. We are constantly and continuously entangled in each other’s inference.

This isn’t metaphor. It’s the actual structure of things around us. Sociologists call it symbolic interactionism. Economists call it game theory with beliefs. Network scientists call it opinion dynamics.

I’m calling it what it is: distributed inference on a human substrate.

The fundamental problem

Here’s the central problem societies solve: coordination without a central processor.

A colony of cells solves this problem by sharing an RNA or DNA, a common instruction set. A flock of birds solves it with simple rules: match your neighbour’s velocity, avoid collisions. But we humans face a harder version of this. We need to coordinate on:

• What counts as property (and who owns what)

• What counts as marriage (and who’s related to whom)

• What counts as authority (and who can compel whom)

• What counts as sacred (and what’s worth dying for)

These aren’t physical facts, no universal laws govern them. They’re shared beliefs, existing as abstracts; with real world artefacts derived from them. And these shared beliefs require a shared inference, some way for millions of minds to converge on compatible models of reality, that can be enforced and observed everyday life.

How does this work and why does happen?

Culture as lossy compression

Imagine you’re an infant. You know absolutely nothing. Everything around you is just noise.

Then, slowly but surely, patterns emerge. You learn that certain sounds mean food is coming. Certain voices and faces mean safety. Certain tones mean danger. You’re not being taught these things explicitly. You’re inferring them sub-consciously from the statistical regularities and repetitions in your environment.

Now scale this up further to a bunch of adults. Imagine you’re in a tribe. There are thousands of individuals, each inferring patterns from their environment. But here’s the problem: raw experience is not transmissible. I cannot upload my sensory stream into your brain. Words cannot wholly transmit the sensations and feelings of an experience. Language is lossy. Gesture is ambiguous. We cannot share experience directly till date.

So we compress. Into words, in stories, into narratives.

Culture is a shared compression algorithm for encoding and transmitting inferences across minds and generations.

So when a grandmother tells a grandchild “don’t swim in that river after dark,” she’s not transmitting her raw experience of crocodile attacks.

She’s transmitting a compressed inference: river + dark = danger.

The grandchild doesn’t need to survive a crocodile to know that the river is a dangerous place to be in the dark. The inference has been cached, from the communication by the grandmother.

The forms compression takes

Each of these is a codec: a way of encoding high-dimensional experience into low-dimensional transmissible form, to be passed down for generations.

Why cultures differ

Different environments that the people grow up in, generate different data. Different data goes on to generate different inferences. Different inferences get compressed into different codecs.

For eg., why do tropical cultures have different cuisine than arctic ones? Not (only) because of the ingredient availability, but because what works in each of these places is different. The inference “raw fish is safe” is true in cold climates (parasites die in ice) of the artic; and false in warm ones (parasites thrive). The compression adapts to the data.

Why do some cultures compress kinship into “family vs strangers” while others have seventeen distinct cousin categories?

Because the inference problems are different. In small-scale societies (eg. tribal groups), knowing exactly who is related to whom determines marriage eligibility, inheritance, and alliance. The compression must be higher-resolution for them to act based on their codec.

Hence you will see cultures with varied naming styles. Some with extremely long naming styles are designed to compress and convey information about not just the individual, but their root/source network.

Culture, therefore; isn’t arbitrary. It’s compressed inference from local data. This is why cultures feel so right to insiders and so strange to outsiders; you’re running a different decompression algorithm based on your “home” culture.

Tradition: the cached inference

Now here’s where all of it gets interesting. Compression has a cost to it: you lose the derivation.

Meaning: when grandma says “don’t swim in the river after dark,” she may not remember why. Maybe the original crocodile attack was four generations ago. The inference got cached; the evidence got garbage-collected.

Why the compression was derived in the first place, is mostly forgotten. Enter: Tradition.

Tradition is cached inference from ancestors, preserved without the original evidence.

This is both powerful and dangerous.

Powerful because it lets you benefit from your ancestors’ experience without repeating their mistakes. You don’t need to discover for yourself that certain mushrooms are poisonous. The cache that got passed down says “don’t eat those.”

Dangerous because environments change. The cache was computed for a different dataset. Time, place, people, risks, weather, belief systems - everything has changed. If the river no longer has crocodiles (they were hunted out), the prohibition becomes superstition - an inference that is completely severed from its evidence. Most religions today find themselves here.

Chesterton’s fence

G.K. Chesterton proposed a principle: don’t tear down a fence until you understand why it was built. In inference terms, this means: don’t invalidate a cached “prior” until you’ve reconstructed the likelihood function that generated it.

This is actually good Bayesian hygiene(refer part 1 for more on Bayesian thinking).

If a tradition has ensured it persists across generations, it probably solved some problem. The tradition is evidence that the problem existed, even if you can’t see it now. Tearing down the fence without understanding is ignoring evidence.

But, and this is crucial, the most important thing is : Chesterton’s Fence has limits.

Sometimes the fence was built by idiots. Sometimes it was built to solve a problem that no longer exists in any shape or form. Sometimes it was built to benefit the fence-builder at others’ expense.

The tradition’s persistence is evidence, not proof. You still have to reason about it, and validate if it needs to exist. Chesterton only says you to question, it is not a call for inaction about updating beliefs. Superstitions such as caste system fall into this category of beliefs that shouldn't exist.

The caste system as cached Inference

Let’s consider caste. What is it, structurally?

One reading: caste is a very old, very high-resolution compression of:

• Occupational specialization (who does what work)

• Purity/pollution categories (who can touch what, who can enter where)

• Marriage constraints (who can reproduce with whom)

• Economic relationships (who owes what to whom)

At some point, centuries ago; these compressions may have been formulated by those in power with an agenda or solved for someone in power of enforcement, a bunch of unknown coordination problems. The truth is we will never know why. It could have been gate keeping resources. Could have been that occupational inheritance meant skill transmission, which was being guarded intentionally.

In most cultures, endogamy meant wealth retention. A lot of purity related rules may have begun as hygiene heuristics, and probably drifted in superstitious, traditional absolutes.

But what is true and measurable is that the environment has changed. Culture has evolved. Hygiene has become the norm. Knowledge is universally accessible and is available(mostly) on merit. The economy has industrialised. The evidence that generated the original inferences for traditions can’t seem to be found. Yet the cache persists - now as a system of oppression rather than coordination.

This is the failure mode of tradition: when the cache outlives its evidence, it becomes a cage.

Politics as distributed state synchronisation

Every society faces a coordination problem: how do millions of entities, or participants or agents align their behaviour without a central controller?

Politics is how we solve this.

Politics is the process by which a society negotiates and propagates shared priors about authority, resource allocation, and legitimate action.

Notice: politics isn’t about finding truth or equanimity. It’s about achieving synchronization.

A society where 51% believe “taxes should be higher” and 49% believe “taxes should be lower” isn’t wrong; it’s unsynced. Politics is the protocol for enforcing an outcome, with or without resolving the desync.

Different political systems = different inference architectures

Each architecture has its tradeoffs. And we live in a time where we can find live examples for each of these political systems across the globe.

Autocracy is fast in decision-making and enforcement; but fragile (if the autocrat’s priors are wrong and trending to coup, no mechanism corrects them). Democracy is robust but slow (consensus-building takes time, easy for corruption in the lower rung of bureaucracy to take form).

And democracies and federalism achieve outcome alignment this via elections.

Elections as synchronisation checkpoints

What is an election, structurally?

An election is a scheduled synchronisation event where members, entities or agents reveal their priors and the system computes a new consensus state.

Before the election, individuals, or entities (could be voting group clusters), or agents hold private beliefs. The election process aggregates these beliefs into a public outcome. The outcome then propagates back as a new prior for all individuals or agents who make up the populace: “The government is now X. Plan accordingly.”

This is why elections feel so weighty. They’re not just about choosing leaders, or picking a side on ideology. They’re technically resetting the shared state.

The day after an election, everyone - winners and losers; updates their model of “what is possible now.”

Constitutions as immutable priors

Some beliefs are too important to update easily. We don’t want 51% of the population to be able to vote away the rights of the other 49%, just because the 51% believe it should be so. So we created constitutional constraints -priors that are expensive to change, and need effort, alignment and prolonged institutional process.

In inference terms: a constitution is a prior with very high resistance to updating. You can’t change it with a single election.

You need supermajorities, ratification processes, years of effort.

This is a feature, not a bug. Some priors should be sticky. “Don’t murder” is a good sticky prior for a society. “This particular family should rule forever” is a bad sticky prior.

The art of constitutional design is deciding which priors should be sticky, and how sticky.

Revolutions as phase transitions

Most of the time, social inference is gradual. People update their beliefs slowly, sometimes over decades, or never. Institutions adapt incrementally. This is a continuous regimen of slow drift.

But once in a while, a spark ignites; and everything changes at once.

In physics, a phase transition occurs when a system shifts from one state to another discontinuously. Water goes from liquid to gas at 100°C. Below that temperature, adding heat just makes water warmer. Yet at that exact and particular temperature, adding heat makes water change state.

Revolutions are social phase transitions.

A revolution is a discontinuous shift in a society’s shared priors, occurring when the existing consensus becomes untenable.

The mechanics of preference falsification

Here’s how it works. Suppose most people in a society privately disbelieve or abhor the official ideology, but each individual thinks they’re in the minority. They falsify their preferences; publicly conforming with everything for their safety, while privately dissenting.

The sociologist/economist Timur Kuran documented this in communist Eastern Europe. Surveys showed overwhelming support for the regime. But the support was hollow- people said what was safe, not what they believed.

In this state, the society is in a metastable equilibrium. Like supercooled water; liquid below freezing, stable until disturbed, then crystallising all at once.

The cascade

When something disturbs the equilibrium- a crisis, a defection, a signal that dissent is safe; the cascade begins:

1. A few people publicly defect (update their visible state)

2. Others observe the defection and update their belief about “how many dissenters are there?”

3. Some of these update enough to defect themselves

4. More observation, more updating, more defection

5. The cascade accelerates until the old consensus collapses

This is exactly how rumors spread, how bank runs happen, how revolutions ignite. Each member or entity or agent is doing local inference on their neighbours. The global phase transition emerges from these local updates.

The Berlin Wall fell in 1989 not because the regime was weaker than in 1988, but because enough people simultaneously updated their beliefs about what was possible. The wall was always made of belief. When the belief evaporated, so did the wall.

Progress: prior-updating what we approve of

Here’s a word that carries enormous weight in our history: progress.

But what is it, actually?

One view: progress is movement toward objective betterment: more prosperity, more freedom, more flourishing.

Another view: progress is just change, and we retroactively label the changes we like as “progress.”

The inference frame offers a third view:

Progress is the process of updating priors toward models that better predict and control reality, where “better” is measured by the goals of the member or principal or agent doing the updating.

This definition has some specific features to it:

1. Progress is real but not universal. When we update from “disease is caused by demons” to “disease is caused by pathogens,” we can now cure diseases. That’s progress by any reasonable measure - our predictions are better, and our control is greater. Diagnosis becomes reliable, interventions become testable, and many diseases become preventable or curable. That is progress in the strict sense: improved inference and improved ability to shape outcomes. Germ theory massively improved prediction and control over disease, yet those gains didn’t arrive everywhere at once and still doesn’t benefit everyone equally.

2. Progress is path-dependent. The updates we make depend on the data we encounter. A society that industrialized via coal will have different priors than one that industrialized via hydropower.

3. Progress is contestable. What counts as “better prediction and control” depends on what you’re trying to predict and control. Economic growth? Spiritual enlightenment? Military power? Environmental sustainability? Different goals, different progress.

The Enlightenment movement as prior replacement

Consider the European Enlightenment movement. What happened?

In inference terms: a critical mass of thinkers replaced their meta-priors. They stopped asking “what does scripture say?” and started asking “what does evidence say?” They stopped treating authority as self-validating and started treating it as hypothesis to be tested.

This wasn’t just a new belief system. It was a new update rule, on scale. And once the new update rule propagated, it generated an avalanche of new beliefs: experimental science, democratic theory, human rights, market economics.

The Enlightenment Age that led to modern science in the current form and manner wasn’t a discovery of truths that were always there. It was a change in how inference was conducted. And that meta-update changed everything.

Polarisation: When channels diverge

Now to the dark side.

In Part 1, I argued that noise is just signal on channels you’re not tuned to. In society, this has a disturbing implication.

Polarisation occurs when sub-populations run inference on different data, using different update rules, and converge on incompatible posteriors.

Let me be more precise about this mechanism, and how it manifests.

Information diets

Suppose Alice watches Channel A. Bob watches Channel B. Over years, they consume different data:

• Alice sees: immigration crime stories, inflation reports, traditional values affirmed

• Bob sees: police brutality footage, climate data, diversity celebrated

Even if both are rational Bayesian updaters, they will converge on different posteriors. Not because one is stupid and one is smart, but because they’re running inference on different datasets.

Trust networks

It gets worse. Alice and Bob also have different trust networks; very different beliefs about whose testimony to weight highly.

Alice trusts: her church, her family, commentators who share her values.

Bob trusts: scientists, journalists, activists who share his values

When Alice’s trusted sources say “X is true” and Bob’s trusted sources say “X is false,” they don’t just disagree about X. They disagree about who is credible. And credibility judgments are upstream of all other beliefs.

The feedback loop

Now add social media into the mix: a channel that solely exists only to optimise for engagement, which correlates with outrage.

1. Alice sees content that triggers outrage at Bob’s tribe

2. Alice updates: “Bob’s tribe is worse than I thought”

3. Alice engages with more such content

4. Algorithm shows more such content

5. Repeat

Bob has a symmetric experience on his side. Both are caught in their update loops that push their posterior inferences further apart.

The physics term for is symmetry breaking. A system that was once mixed separates into distinct phases. The social term is polarisation.

The experience is: “I don’t recognise my country anymore.”

Is there a fix?

The inference frame suggests where to intervene:

1. Shared data: If people are diverging because of different inputs, expose them to common inputs. But this is extremely hard to accomplish when attention is scarce and curated, and data inherently is not trusted.

2. Shared update rules: If people are diverging because of different likelihoods (different beliefs about what sources to trust), rebuild shared trust. But trust is slow to build and fast to destroy.

3. Depolarisation cascades: If the problem is cascade dynamics, trigger reverse cascades; find ways to make moderation contagious.

None of these are easy, or can be done in a day. They take years of slow build-up. But at least the frame tells us what the problem is: not that people are stupid, but that they’re running inference in separated channels.

What we are to each other

Let me end somewhere more personal.

We’ve talked about societies as inference networks, cultures as compression algorithms, politics as synchronization protocols. All true. But it can sound cold; as if we’re just nodes in a graph, just variables in an equation.

We’re not. Or rather: we are, and we’re also something more.

Here’s the something more:

You are not just the product of your own inference. You are the product of everyone who ever updated your priors.

Your parents. Your teachers. Your friends. Your coach. The authors of the books you read, the makers of the films you watched, the strangers who were kind or cruel to you at the moments it mattered. They all wrote themselves into you. Not metaphorically. Actually. Your beliefs are their cached inferences, still running.

And you do the same for others. Every conversation, every gesture, every choice you make - you are data for someone else’s inference. You are updating the world’s priors just by existing.

The weight of witness

In physics, observation changes the system(remember the much mis-understood Schordinger’s cat). In society, the act of witnessing itself changes both parties.

When you listen to someone’s story, really listen; you’re doing something very sacred. You’re allowing their experience to update your priors.

You’re saying: “Your data counts. Your perspective could change mine.”

And when you speak, you’re offering your own data for others’ inference. So it takes courage to speak up against perceived wrongful notions, because speaking up is to offer data, to cause a living ripple in the inference pattern. You’re participating in the update on the network, contributing your unique vantage point to the collective computation of humanity.

This is why debate matters. This is why stories matter. This is why conversation matters. We are building reality together, inference by inference, update by update; and we all live in the collective fabric of this inference.

The tragedy and the hope

The tragedy is that inference can go wrong. Cascades can mislead. Caches can be rigid and unchanging, and calcify. Channels can diverge. With enough inference, hate can replace humanity in our belief systems. We can end up in worlds where millions of people share beliefs that are beautifully internally consistent and catastrophically detached from true reality.

The hope is that inference can also go right. We can update. We can learn. We can find common ground by finding common data. The same mechanism that produces mass delusion can also produce collective enlightenment for us all, if we choose it.

The same network that carried partition’s violence also carried partition’s healings; the millions of small kindnesses, the neighbours who sheltered neighbours, the belief that this person is still a person.

We are each other’s priors. We will be each other’s future updates. What we believe about each other also determines what becomes possible between us. That’s a responsibility, and also a gift.

Next in the series: Part 3: Markets as Inference Engines: how prices are posteriors, arbitrage is error correction, and money is the universe’s most successful communication protocol.

Appendix: Key References

1. Timur Kuran — “Private Truths, Public Lies” (1997). The definitive account of preference falsification and revolutionary cascades.

2. Joseph Henrich — “The Secret of Our Success” (2015). How cultural evolution accumulates adaptive knowledge beyond individual capacity.

3. James Scott — “Seeing Like a State” (1998). How states compress local knowledge into legible categories—and what gets lost.

4. Thomas Kuhn — “The Structure of Scientific Revolutions” (1962). Paradigm shifts as prior replacement in science.

5. Mark Granovetter — “The Strength of Weak Ties” (1973). Network structure’s effect on information propagation.

6. Daron Acemoglu & James Robinson — “Why Nations Fail” (2012). Institutional path-dependence as prior lock-in.

7. Robin Dunbar — “How Many Friends Does One Person Need?” (2010). Cognitive limits on social network size and inference capacity.

Journeys to the unknown do a fascinating thing to your mind- they take you on a path that makes you pause, it shows you things and patterns and presumptions that make you want to be still, till you make some semblance of sense about it. The stillness is visceral at times, and in this stillness at times is born great chaos, from which you observe for the first time; threads and silhouettes of pure beauty in the patterns of how existence weaves.

This is part 1 of a 3 (or) 4 part series (hopefully), at one of these pauses I am currently at.

Some things start stupidly simple. Like a light switch. And when you flip the light switch, you think you’re “sending electricity”. You’re not.

Here's what actually happens: electrons in the copper don't flow from switch to bulb like water through a pipe. The electrons themselves drift at roughly 0.1 millimeters per second; a snail would outrun them. Yet the light turns on instantly, because the signal travels at about 70% the speed of light.

Basically, you’re starting a rumour in a lattice of atoms. And this rumour travels at near-light speed while the actual electrons barely move. The wire isn’t conducting energy. It’s conducting an inference about a state-change that happens somewhere to it, a “belief”.

How? The electromagnetic field propagates. Each atom's electron cloud shifts slightly, changing the local field, which shifts the next atom's electrons, which changes that field. Imagine the first few goosebumps as the summer breeze grazed by you, just that now it’s a cascade of state-changes that ripples through the wire at 200,000 kilometers per second, while the actual electrons barely budge.

And it’s not just wires that transmit electricity. Take a stadium full of people watching their favourite team. The game catches fire, the crowd gets loud, and someone starts a stadium wave. No person travels around the stadium. Each person just infers from their neighbour, “oh, I should stand now,” and updates their own state. The "wave" is pure information propagation. A moving pattern, made of people who stay where they are.

The wire is doing the same kind of thing. It’s not shoving energy down a pipe. It’s letting each atom “read” the state of its neighbours, adjust itself, and “write” a new state for the next one to read. A long chain of tiny, local alignments.

Here’s what I think is happening: what we call "physics" is better understood as inference :- local state updates propagating through channels according to consistent rules.

The universe, as we experience it; is not a machine that processes energy.

It is a machine that propagates state-changes. What we “experience” as forces, fields, and causation are channels through which these updates flow.

All the separate boxes we live inside - “physics”, “technology”, “social networks”, “society”, “markets”, “mind” - are just different channels through which the same kind of update runs through. We’re not outside that process. We’re the witnesses, the participants, and, sometimes, the ones pushing the next update along.

Signal and noise are therefore no inherent properties of reality, they’re properties of the channel we’re tuned to.

If you want the most compressed version of all this: think of the universe as a giant cellular automaton: a grid where each point just updates its state based on its neighbours, using simple rules.

I’m not saying “physics is inference” in some mystical way. I’m saying the math it seems to follow has the same shape as inference: local updates, rippling out.

And if that’s true, then your life isn’t a random spark in the dark. It’s one very specific pattern in this ongoing wave of updates; the part of the universe that has become aware that the wave is passing through it.

What do I mean by “Inference”?

First, let’s be precise. Inference in a strictly, scientific context is a bit more formal and rigid. It’s not a claim that atoms are “conscious” or that electrons “think”, that lattice “believe” or energy “feels”. That is pseudo-science territory, one I strongly intend to avoid here.

I am looking at the structural definition, the mathematical form of a real-world, physical process that is identical to the mathematically defined (read: equations that most people believe caused their childhood migraines) form of inference.

Students of engineering, statistics and sciences will recognise that in Bayesian inference (if you’re unfamiliar what this is, don’t worry about it for now. This article is supposed to unburden this), you have:

• a prior: what you believed before

• a likelihood: how new data relates to possible states

• a posterior reality: your updated belief

The update rule laid out in mathematical terms can be described as:

Now look at how a physical field evolves. Take the electromagnetic field. At each point in space, the field state depends on:

• The previous field state (the prior)

• Boundary conditions and sources (the likelihood—how charges constrain what’s possible)

• The new field state (the posterior)

Maxwell’s equations (the ones that made you either fall in love with physics, or decide you never wanted to be an engineer) aren’t philosophically “about” inference. But interestingly, they have the same structure:

• a local state

• an update rule,

• a propagation outcome.

The math is isomorphic. This is just engineering speak for the fact that the mathematics in these two situations has the same structure, so you can treat them as essentially the same problem.

What this implies in practice is that you can safely transfer theorems, formulas, mental models and intuitions from one setting to the other, because nothing essential about the structure changes; only the “labels” do.

The equations don’t feel like they are “belief” updates, but structurally they are, in everyday terms, the following:

• yesterday’s rumour

• today’s evidence

• tomorrow’s new rumour, repeated at every point in space.

This matters because it means we can use inference as a conceptual frame without making claims that violate the physics of the universe. We’re not adding anything to the physics. We’re redescribing what the physics already says.

The wire that doesn’t conduct

Let’s go deeper on the wire example, because it’s the perfect entry point for anyone in the world today who has access to electricity.

The standard picture (which is correct, but misleading)

The middle school textbooks said: voltage difference creates an electric field inside the wire, which exerts force on free electrons, which flow and carry current.

All true. If the teachers cut your marks coz you got this statement wrong, they can’t be blamed. But, there’s a but.

And the but of it is that it frames the wire as a pipe and electrons as water. This creates the illusion that stuff is moving from one end to the other. And that, is fundamentally not true.

The Inference Picture (also correct, yet more revealing)

What’s actually happening at each point inside the copper:

1. An electron shifts position (due to the applied field)

2. This creates a local perturbation in the electromagnetic field, like the ripple you cause when you dip your feet into the pool

3. The perturbation propagates outward at near-light speed

4. Neighboring atoms’ electrons “respond” to the changed field

5. They shift, creating new perturbations

6. This cascade continues on and on

The key insight here: the signal is in the field, not the electrons. The electrons are just the substrate that couples to the field. The “information” - the state-change; propagates through the field channel, which is the latticework of atoms in the wire.

In this frame, the wire stops being a pipe and starts being a rumour network: some rumours hug the surface, some echo back and forth, some pass without any friction at all.

This also explains why kitty-party groups are considered disruptive, although they aren’t the ones doing the disruption.

Why this framing matters

This isn’t just semantics. It highlights and predicts a real phenomena, and some hard words that you’ll need to brace for:

Skin effect: At high frequencies, current flows mostly near the wire’s surface. They prefer to move along the “skin” of the wire.

Why? Because the field’s propagation dynamics change at high frequencies. Meaning - the rate at which the ripple feel at the edge of the pool where you dipped at the foot, will be felt different a few feet below the pool, or at the middle and opposite end of the pool.

The “inference channel” has different properties near the surface than deep inside.

Transmission line behaviour: Long cables don’t act like simple pipes. They act like they disagree with a ripple being there, and fight back, talk to each other and do things of their own. In physics, this means they behave as wave propagation channels with impedance, reflection, and interference. And therefore things heat up, for real. Heat is a consequence output in these physical phenomenons.

Why do they do this? Because that’s what they inherently are: inference channels, with bandwidth limits (how much water can flow) and echo patterns (some ripples get reflected back).

Superconductivity: Below critical temperatures, resistance drops to zero. In inference terms: the channel becomes lossless. The lattice stops “disagreeing” with the field state, it synchronises perfectly, so no information is lost to heat.

The inference frame about the universe and its working doesn’t contradict the physics. It pulls forward the parts that the “water in a pipe” metaphor I used earlier hides.

Forces as inference channels

If the electromagnetic field is an inference channel, what about the gravity that hold us on earth and keep the moon in its orbit? The other strong forces? What about sound?

Let’s check each one.

Gravity: geometry as inference

In Einstein’s general relativity, mass doesn’t “exert force” in the Newtonian sense (meaning, the way Newton perceived the laws of the world and how all the mass in the universe behaves).

Mass curves spacetime. Other masses follow geodesics: the straightest possible paths through curved space.

If we had to reframe this: mass updates the local geometry. That updated geometry propagates outward at the speed of light (as gravitational waves). Other masses read that geometry and update their trajectories accordingly.

This isn’t speculation. Einstein’s field equations describe exactly this: how the source of the change/curvature(matter/energy) updates the geometry (metric tensor) at each point in spacetime.

The “inference” that happens here by everything that has “mass” in the universe is: given the local curvature, what geodesic should I follow?

Every mass particle “computes” this continuously. The universe doesn’t need a central processor to this math.

Each point updates locally; the global solution emerges.

In that sense, gravity is the universe’s way of telling any matter, “here’s the shape of the story you’re allowed to follow,” and every particle keeps re-reading that local script as it moves. Destiny for inorganic things with mass, is real.

This goes beyond just astronomy and astrophysics. There’s even a theoretical program, entropic gravity; that attempts to derive Einstein’s equations from information-theoretic principles.

Now what is information-theoretic principles: These are general rules and constraints that come from information theory, describing how information can be quantified, transmitted, stored, and processed, especially under uncertainty and noise.

Examples:

• Predicting a fair coin toss (50% heads, 50% tails) gives you 1 bit of information when you see the outcome.

• Predicting weather where it is sunny 99% of the time gives very little information to you when it is sunny, because you were almost sure anyway; news about the probability of a rare storm carries more information to you.

These principles give formal limits (what is or is not possible) and optimal strategies (how to do it best) for any system that handles information, from communication channels to learning algorithms and cryptographic schemes. Erik Verlinde’s work suggests gravity might literally be an emergent consequence of entropy gradients. The inference frame isn’t just metaphor; it might be close to a fundamental aspect of our universe.

Electromagnetism: the original signal channel

Maxwell’s equations that you learned in middle-school are explicitly describing a propagation system. They describe how changes in electric fields create magnetic fields, which create electric fields, and so on, forwards at speed cc.

Every antenna engineer knows this. Radio isn’t “sending waves” in some vague sense. What it is doing is:

• Create a pattern of charge oscillation

• That pattern propagates through the field channel (atmosphere)

• A distant antenna’s charges respond to the arriving field pattern

The entire telecommunications industry and our everyday communication, entertainment etc., via every increasingly complex smartphones are built on top of the inference-channel structure of electromagnetism.

Sound: molecular state propagation

Sound is perhaps the clearest and cleanest example we humans can understand.

When you speak:

• Your vocal cords vibrate, pushing air molecules

• Those molecules collide with neighbors, transferring momentum

• Those neighbors push their neighbors

• A pressure pattern propagates outward

No molecule travels from your mouth to my ear. The air just passes along tiny shoves, one neighbor to the next, like a tightly packed crowd passing a single, urgent message.

This is precisely inference: each molecule updates its state based on neighbor states, according to the rules of intermolecular forces. The “posterior” (new position and velocity) depends on the “prior” (previous state) and the “likelihood” (forces from neighbors).

Sound velocity depends on the channel’s properties: molecular mass, temperature, medium density.

Higher temperature = faster inference (molecules respond faster).

Denser medium = different impedance (different channel characteristics).

A unifying table

You can observe the same trick across almost everything in the universe, in different costumes. The metaphorical pause in the journey I mentioned at the start, is when you realise that eventually it is all just one universe, endlessly propagating state updates through whatever substrate is available within it.

Every physical process can be decomposed into:

• substrate (what carries the state)

• update rule (how state changes given neighbors)

• and channel properties (speed, bandwidth, noise).

The cosmos as distributed inference

Now let’s scale this up, let’s go cosmic in a manner Nolan would be proud of. If local physics is inference, what does the universe look like as a whole?

We see the into the past, to back in time, when we look up at the night sky.

Why? Because inference, on cosmic distances; has observable latency.

The light arriving at your eye from the Sun left 8 minutes ago. The light from Alpha Centauri, the nearest star system to the Sun and also one of the brightest objects in the night sky (and has played an important role in both astronomy and human culture) left 4.3 years ago. The most distant galaxies we observe emitted their light ~13 billion years ago.

In the inference frame of things: we’re reading old, stale data.

The cosmic microwave background is the oldest cache of this. It is from the universe’s state 380,000 years after the Big Bang, frozen in light and still propagating across the void.

The fascinating bit is that this isn’t just “information traveling slowly.” It’s the fundamental structure of reality. Relativity forbids instant state-synchronization.

The universe is eventually consistent, just not strongly consistent. Each point in the universe only knows what’s in its past “light cone”.

Causal structure as channel topology

In relativity, events can be:

• Timelike separated: one could cause the other (within a light cone)

• Spacelike separated: neither can affect the other (outside each other’s light cones)

• Lightlike separated: on the boundary (connected by light)

This is channel topology on cosmic scale. Spacelike events are on disconnected channels, no inference can propagate between them. Timelike events share a channel, state updates can flow.

The entire causal structure of the universe is a statement about which inference channels exist.

There are places you can never hear from, and places that can never hear from you. Silence, in this picture, isn’t absence, it’s the geometry of who can’t share an update with whom.

This means the folks at Marvel Comics/Studios could put out a million Multiverse movies filled with aliens villains, and other multiverses could never be offended, because they will never know of the movie’s existence. Think: small privileges on cosmic scale, big wins in keeping alien attacks at bay.

Black holes as channel boundaries

What happens at a black hole’s event horizon? Nolan’s Interstellar managed to get a boring, nerd-level geometric diagram propogated into the imagination of billions, but the math still remains complex.

From outside: nothing can escape. Light, matter, information; all fall in, none return, ever.

The event horizon is a one-way inference channel. “States” can propagate in, but not out.

This creates the (almost)famous black hole information paradox. Quantum mechanics says information is conserved. But if information falls into a black hole and can’t escape, where does it go?

The current best answer (predicted/derived from Hawking radiation + holographic principle): the information is encoded on the horizon surface.

The 2D boundary of the black hole, contains all the information about the 3D interior. The channel here doesn’t destroy information; it transforms how it’s stored.

This is very strange, but it’s mainstream physics. And it’s deeply information-theoretic.

The universe seems to care about conserving information more than conserving matter. Why? Maybe that’s another journey to be made.

Entropy and the arrow of time

One of my biggest torments in 4th grade was things always went ahead. I couldn’t pause to alter, or reverse to edit or change. The broken window remained broken, and the beating was always guaranteed. I could never know in advance how much of a beating I would get, and be prepared for it. Why did time flow only in one direction? Why do we remember the past, not the future?

The thermodynamic answer: Entropy increases. Systems evolve toward more probable states.

The inference answer: the universe is updating toward its maximum-entropy posterior.

The past had lower entropy (more constrained, more informative prior). The glass in the window fit snug and square into the frame. But on breaking it, I’d set the glass shards free. Freedom at last, but now they never could get back to being the single glass slab within the window frame again. The future has higher entropy (less constrained, less informative posterior).

The arrow of time points in the direction of inference: from specific to general, from constrained to unconstrained.

This connects to Jaynes’s profound insight: thermodynamics is inference.

Temperature, pressure, entropy: these aren’t fundamental properties of matter. They’re what you can infer about microstates, given the macrostate observations.

Statistical mechanics is just Bayesian inference applied to particle physics.

The noise that isn’t

Now to a part of a thesis that I find myself wanting to contradict, because it deeply tests my sense of individual actions, impact and the mark our life makes in the fabric of time : the universe doesn’t distinguish signal from noise.

Let me be more careful here. This isn’t about things that sound moral or mystical. It’s about our understanding of the nature of existence (ontology) vs our understanding of nature, knowledge and the limits of knowing (epistemology).

The traditional view on signal vs noise

In electrical engineering or communications theory, signal is what you want to transmit. Noise is what corrupts it. It’s a very binary, simple view of things. And this has worked well for us, the Information Age exists on this key distinction.

And this distinction is purpose-relative: defined by what you’re trying to do.

A radio engineer calls this static “noise.” It’s the buzzing you hear when your Zoom call audio breaks into crackles/static(aka “thermal noise”) due to a bluetooth speaker in the vicinity of your laptop. But that static contains:

• Cosmic microwave background (residual from the Big Bang)

• Thermal radiation from the antenna

• Distant lightning strikes

• Solar radio bursts

From another purpose, each of these is signal. The “noise” is just signal on channels you’re not tuned to, or you don’t intend to tune to.

When we call something noise, we’re really just admitting, “I’m not listening to that part of the universe right now.”

What’s true on a physical level:

Here’s the key thing: physics has no concept of noise.

Every interaction is deterministic (or, in quantum mechanics, lawfully probabilistic). Every fluctuation has a cause. Every perturbation propagates.

What we call “thermal noise” is actually the EM field faithfully transmitting the state of randomly-moving charged particles. It’s not a failure of the channel. The channel is working perfectly as it should, it’s just carrying information we don’t want.

In empty space out there in the universe, random jitters of energy and fields that occur even in “empty” space due to the rules of quantum mechanics. These so called quantum vacuum fluctuations aren’t “randomness” in the everyday colloquial sense. They’re the ground state of quantum fields: the “base prior”, if you will.

Even it absolute stillness, in what’s supposed to be nothing, just an empty void, the universe still murmurs something in it’s sleep. The universe’s lowest-energy configuration still has a structure to it.

The Holographic Principle and Information Conservation

Jacob Bekenstein was born in Mexico City in 1947, the son of Polish immigrants. William Blake was born two centuries earlier, in 1757, in Soho, London. Separated by time and space, they never met, and they worked in utterly different worlds.

Both went on to write new chapters in their areas of expertise. Bekenstein’s ideas reshaped how physicists think about the information content of the universe, from black holes to cosmological horizons. Blake’s poems and images have radiated outward for over two hundred years, influencing composers, painters, novelists, and the entire “Romantic imagination”.

It is fitting, then, that when Bekenstein set out to question one of the most fundamental assumptions about the universe, he reached for Blake.

Bekenstein asks "Could we, as William Blake memorably penned, 'see a world in a grain of sand', or is that idea no more than 'poetic license'?"

And the answer Jacob discovered, 200 years after Blake asked the question; is : Yes, we can.

If you burn a book, the information in it isn’t destroyed. It’s transformed into millions and millions of fragmented parts in the smoke, ash, and radiation. In principle (assuming you had access to limitless computational resources), you could compute and reconstruct the book from perfect measurements of the combustion products. If you could manage the compute, you could reverse time; and bring the book back to as it was before it burn.

This is the holographic principle. It (well-supported in theoretical physics) states:

the maximum information content of any region of space is proportional to its surface area, not its volume.

Implication: information is never truly destroyed. It can be scrambled, dispersed, transformed; but not erased.

This is real physics, not some meta-physical speculation. Any act of erasure is basically you removing a set of information from object A, and converting it into some energy output, such as hit.

This is true even for digital data. Every time you “erase” information by deleting a photo in your gallery, or even hit backspace to “delete” an entry; you produce a tiny amount of heat.

This is proven by Landauer’s Principle (experimentally verified), which says that erasing one bit of information requires dissipating at least kTln⁡2 of energy.

The formula is just Landauer’s way of saying “erasing one bit of information has a minimum energy cost, and that cost depends on temperature.”

Information and thermodynamics are inherently coupled. The universe keeps records of the state change.

So where does “noise” come from?

The country maps we had in schools for geography class shows us coastline with extremely jagged, random borders. It’s never a smooth curve. It’s “noisy”. But yet, when you stand on the beach, stretching for miles as you walk on the sands, nothing look jagged. The land curves and meanders, and while we are there at that beach in the warm summer sea, there’s only the breeze and the continuous coastline, interrupted by smooth curves. There is no noise.

Noise is a coarse-graining artifact.

Coarse graining = replacing a detailed, fine‑scale description with a simpler one that keeps only “important” variables. In physics, this removal of the un-important variables would be termed as removing “degrees of freedom” from the process.

So what “noise is a coarse-graining artefact” means is if one had a perfectly fine‑grained description (tracking all microscopic degrees of freedom), the process would look smooth and noise‑free; what is called “noise” only appears after simplifying. Imagining zooming in on a torn paper edge under the microscope, as you keep zooming; the details don’t looked jagged as it seems to your naked eye; but smooth continuous curves.

When we model systems, we choose to ignore certain degrees of freedom. Those ignored degrees of freedom manifest as apparent randomness in the degrees of freedom we track.

For example: Let’s take Brownian motion, that describes the movement of independent particles in a given system. A pollen grain in water jitters randomly. But the “randomness” is just the aggregate effect of billions of water molecules colliding with the grain. If you could track every molecule, there would be no randomness, just a very high-dimensional deterministic (or lawfully probabilistic) system.

“Noise” = information on channels you’ve chosen to ignore deliberately .

This has practical implications. Every breakthrough we have had in sensing technology comes from figuring out how to read channels that were previously “noise”: gravitational wave detection, single-molecule imaging, cosmic microwave background mapping.

What this gets us

Let’s head back to the core claim (and what’s not claimed):

Not claiming:

• atoms are conscious

• the universe has intentions

• Physics is reducible to information (it’s still an open question in science)

• that the framework/thinking I am discovering makes new empirical predictions that differ from standard physics

What I am claiming:

• The mathematical structure of the physical laws is identical (isomorphic) to those of inference

• This isn’t an abstract metaphor. I believe it cleanly and clearly is a description of what the equations already say

• The framing described above unifies phenomena that I and folks around me have usually treated separately (the forces, fields, thermodynamics, computation, information etc)

• Signal & noise are subjective to the observer (strictly speaking, they are not ontological)

• Information, as we define it, is physically intertwined with dynamics (as established by Landauer, Holographic Principle, Jaynes’s Information Theory, etc.,)

Why bother with this framing:

For me personally, there’s three reasons:

Clarity:

The world feels less mysterious to me when I can as a simple question: what is the inference channel here?

What’s the substrate? What’s the update rule? How fast do these updates travel, where do they get lost? The same questions apply for a copper wire, a nerve in your body, a rumour, a market, a political ideology, a company’s share performance, an impression about people we interact with daily.

Life more often than not feels like one large Bayesian system, and resisting updates is just to be bling about how the universe wants to talk to us, we’re just sinking deeper into a self-sabotaging ego state.

This reframing gives me handles - place to grab the problem, without pretending that I have “explained” everything happening to me to myself. I can understand things better, but I’m not claiming this is the ultimate, complete explanation of reality.

Coherence:

Ever since we start of in school, we are taught to compartmentalise. To be narrow. Study and keep the world into separate folders: Physics here, Technology there, then “society”, “markets”, “investments”, “mind”; as if all of them are different planets. I have always hated this, always felt this was limiting. And at times made to feel stupid for thinking so.

This framing helps me reset. The universe seems to have a coherence to it. This framing let’s me see the universe as the same story told through different materials; the same kind of wane moving through electrons in a wire, the air in the stadium, spacetime around a star, neutrons in the cortex of our brains, people in a crowd. It’s all one conceptual spine, with many costumes.

For me, this sense of continuity matters. It enables me live one life, instead of five disconnected ones.

Intelligence:

This framing helps me prepare for intelligence. This is very different from “being” intelligent. A lot of very smart folks that I know, exist in the past. They were intelligent once, but something changed.

A lot of people struggle with updating their intelligence, and a lot of scorn is thrown at people who “update” their beliefs as new events occur; it’s almost frowned upon.

Holding onto old “beliefs”, and not updating it in the face of new data is not heroic, it’s contrary to nature.

If we look at the the physics of the universe, physical processes are already doing inference. Our brains aren’t magical exceptions; they are what happens when the universe builds an inference engine that can model itself.

Minds are what inference feels like from the inside.

Why is this important? For me, this shifts the original question, “who notices?”, by just a little bit.

It stops being : is there an external observer keeping the score of what I do?

and becomes: “what does it mean that a small part of this vast update process has become aware that a wave is passing through it, and it can therefore choose how to respond?”

This is the real reason why I care about this framing. It doesn't just tidy up some Physics for me. It changes how I relate to being a brief, fragile pattern in the middle of it all.

Coda: the universe computes itself

There’s a line of thought in Physics that says computation is fundamental. Theoretical physicist John Wheeler famously termed it “it from bit.”

In his 2006 book, Programming the Universe, Seth Lloyd contends that the universe itself is a large quantum computer. He counts up how many operations the universe has run since the Big Bang.

Stephen Wolfram (Physicist + Computer scientist + CEO/inventor of Mathematica and Wolfram Alpha, author of A New Kind of Science), in his work on cellular automata, draws little grids where simple rules only look at their neighbour, update their state to change the pattern over and over again; to spin out to create crazy, rich, complex structures and behaviours that become whole worlds of their own.

All of that is interesting background. But I didn’t start this with Wheeler or Wolfram. I started it with a smaller, more pathetic question: in my tiny blip of time, who even notices?

The inference frame above sits right between those two scales.

On one side, the picture looks like this:

The universe is not a stage on which events happen. The universe is the happening. Each point in spacetime updates its state based on its neighbors, according to the rules we call physical law. There’s no central processor, no global clock, no preferred frame. Just local rules, applied everywhere, always.

Gravity, light, sound, heat - these aren’t different phenomena. They’re the same phenomenon (state propagation) through different channels (spacetime metric, EM field, molecular positions, kinetic energy).

On the other side is me, sitting inside that process, trying to make sense of my one short run at this life.

In this view, there’s nowhere for me to stand outside the computation, no balcony seat from which I get to watch the show.

We’re not observers looking in; we’re ripples inside the “happening”, briefly aware that we are.

The question “who notices?” quietly flips: there is no separate watcher. The noticing is happening here, in you, as part of the same update.

Our memories, our convictions, our sudden late-night clarity; those are just another kind of state-change propagating through this universe-sized inference machine. I am not an error term. I am one of this machine’s patterns of self-measurement.

And that’s where the noise question lands too.

From the inside, so much of life feels like interference: distractions, static, randomness. From the outside; if there were an outside, it would all just be propagation. Every fluctuation, every lost day, every sharp moment of awareness: the same process, different channels.

There’s no noise. There’s only signal we’re not listening to.

Appendix: Key References

For readers who want to go deeper:

• “It from Bit” - John Archibald Wheeler’s proposal that information is fundamental to physics. See his essay Information, Physics, Quantum: The Search for Links (1990).

• Landauer’s Principle - Rolf Landauer’s proof that erasing information requires energy dissipation. Experimentally verified in 2012 by Bérut et al. in Nature.

• Jaynes and Maximum Entropy - E.T. Jaynes’s work treating statistical mechanics as inference. See Information Theory and Statistical Mechanics (1957).

• Entropic Gravity - Erik Verlinde’s proposal deriving Einstein’s equations from thermodynamics. On the Origin of Gravity and the Laws of Newton (2010).

• Constructor Theory - David Deutsch and Chiara Marletto’s framework recasting physics in terms of possible and impossible transformations.

• Holographic Principle - ’t Hooft and Susskind’s proposal that 3D information is encoded on 2D surfaces. Foundational to modern quantum gravity.

• Seth Lloyd’s Computational Universe - Programming the Universe (2006), which estimates the universe’s total number of operations since the Big Bang.

This is a conversation space exclusively for subscribers—kind of like a group chat or live hangout. I’ll post questions and updates that come my way, and you can jump into the discussion.

Join chat

How to get started

1. Get the Substack app by clicking this link or the button below. New chat threads won’t be sent sent via email, so turn on push notifications so you don’t miss conversation as it happens. You can also access chat on the web.

Get app

2. Open the app and tap the Chat icon. It looks like two bubbles in the bottom bar, and you’ll see a row for my chat inside.

3. That’s it! Jump into my thread to say hi, and if you have any issues, check out Substack’s FAQ.

So here’s a re-attempt starting with building first a set of personal operating principles for now. The aim is to embrace a curated set of principles, see what works and sticks with me into helping build up my mental models of the things I do, and help me build a world model based on which I operate. The key lever is to continuously update, iterate and add/remove aspects to it, on a 14-day feedback cycle.

Since this is the v1 - the goal is to create a simple codex of operating principles from observed social effects, laws from across domains such as system design and information theory; and key observations or statements that are loosely defined, and are probably incomplete.
The effort is to apply them and evaluate if they are needle movers over the next 120 days as I practice and iterate on/with them everyday.

Note: the principles might not be in the “right” form or “original definitions” , but more as distillations of what I understand them to be.

So I’ll keep what works and prune the rest over 120 days, and hopefully have a revised set in the second week of February, 2026.

On Decision and focus

• Do not fear: Action > anxiety. Default to movement.

• Kidlin’s Law: Write the problem precisely; half the work is done.

• Pareto Principle: 20% of inputs drive 80% of outputs. Identify the 3 highest-leverage movers and ship them first.

• Falkland’s Law: If a decision isn’t required now, don’t decide now.

• Hick’s Law: Fewer options → faster choices. Pre-limit menus, eliminate abundance of choices.

• Inversion: Think by inversion. Avoid stupidity before seeking brilliance. Hence, first remove known and identifiable stupidity and known failure modes. Then optimize for what remains.

• Second-Order Thinking: Trace consequences of consequences before acting.

• Le Chatelier’s Principle: Systems in equilibria resist change. Start with nudges, not force.

On Systems and Information

• Shannon’s Principle: Clarity reduces noise. But ensure that when you compress information, it does not lose its meaning. (Intent > Quantifiable Metrics). So maximize signals. Compress these signals without losing meaning. Favor intent over vanity metrics.

• Chesterton’s Fence: Do not remove a system unless you understand why it was built in the first place. Don’t dismantle what you don’t yet understand yet.

• Goodhart’s Law: When a measure becomes a target, it stops being a good measure. Track lead indicators, audit for gaming. (related to Shannon’s)

• Base Rates > Outlier Stories: Start with priors, then update with evidence. Default to what usually happens, not to a compelling anecdote.

• Small Bets: Prefer reversible, low-cost experiments to big irreversible moves.

On Networks and Value

• Sutton’s Law: Go where the money or root cause is.

• Metcalfe’s Law: Network value grows with the square of connections. Network utility scales ~ n². Design for density and interoperability.

• Lindy Effect: The longer a thing has lasted, the longer it is likely to last. Weight proven playbooks more.

Craft and execution

• Wilson’s Maxim: Compound knowledge and judgment; money follows.

• Gilbert’s Law: The method is your responsibility; optimize for the desired result.

• Rate × Duration: Quality is throughput over time. Ship on a schedule.

• Constraint First: Identify the bottleneck and elevate it before adding resources.

Interpersonal and error-handling

• Hanlon’s Razor: Don’t attribute to malice what is adequately explained by incentives or error.

• Surface Area for Luck: Be willing to be wrong often, partner well, make it easy to find and help you.

• Reduce Blast Radius: Contain failures, log them, write the guardrail.

Field Notes per Principle

• Kidlin: Start every work block with a 3-line problem spec: state, constraint, success test.

• Pareto: Daily 3×5: three needle-movers, five quick wins.

• Hick: Pre-decide defaults: toolchain, KPI set, hiring bar, pricing guardrails.

• Inversion: Pre-mortem: list the five most likely failure modes; design one counter for each.

• Shannon: One-page brief before any project; one-page readout after.

• Chesterton: Before changing a legacy process, write a paragraph on why it existed and what breaks if removed.

• Goodhart: Pair every target metric with a counter-metric that catches gaming.

• Metcalfe: Prefer platforms that deepen connections: APIs, standards, shared data models.

• Le Chatelier: Sequence changes: observe → nudge → measure → escalate.

• Gilbert: Optimize for outcome, but document the playbook to make it repeatable.

• Hanlon: Escalate by data, not tone. Propose the smallest fix that could work.

  ---

  Daily Loop (≤30 minutes)

  1. Define (5m): 3-line problem spec for the day.

  2. Select (3m): Pick the Pareto 3 and the five quick wins.

  3. Premortem (5m): List top risks; add one guardrail each.

  4. Ship (work).

  5. Debrief (10m): One-page readout: what moved, what didn’t, why.

  6. Log (7m): Update metric + counter-metric, note any Goodhart drift.

  ---

  Monthly Cadence

  • Day 1 - Constraint Pick: Identify one bottleneck to elevate. Define a single success test and guardrail.

  • Day 5 - Plan Lock: Freeze the Pareto 3 for the month. Set target + counter-metric pairs.

  • Day 10 - Network Density: Ship one integration, one partner touchpoint, or one community artifact.

  • Day 15 - Midpoint Audit: Check Goodhart drift, remove one metric, one meeting, one tool.

  • Day 20 - Experiment Block: Run two small bets. Pre-commit stop/scale rules.

  • Day 25 - System Hygiene: Document deltas, retire a legacy step (only after a Chesterton note).

  • Day 28–30 - Lindy Review + Closeout: Keep/kill old playbooks. One-page readout: results, failures, counters. Queue next month’s constraint.

  ---

  120-Day Review

  • Keep: Principles used ≥60% of days with measurable lift.

  • Modify: Principles used but neutral; rewrite into a tighter heuristic.

  • Drop: Principles unused or net-negative.

  • Add: Max 3 new principles sourced from observed wins.

  ---

  End note:
  Principles are tools, not beliefs. Measure, prune, replace. Nothing is sacred except outcomes. Default to action. Close the loop every day and every month.

• Operational checklists:

  Daily (≤30 minutes)

  • Define: 3-line problem spec (state, constraint, success test)

  • Select: Pareto 3 + five quick wins

  • Premortem: top 3 failure modes + one guardrail each

  • Ship

  • Debrief: one-page readout

  • Log: target + counter-metric, note Goodhart drift

  Monthly

  • Constraint: pick one bottleneck, set single success test

  • Plan lock: freeze Pareto 3 for the month

  • Metrics: set metric + counter-metric pairs, baseline captured

  • Network: 1 integration or partner artifact shipped

  • Experiments: 2 small bets with stop/scale rules pre-committed

  • Hygiene: retire one legacy step after Chesterton note

  • Review: Lindy check, keep/kill playbooks, one-page closeout, queue next month’s constraint

  Project start (once per project)

  • One-page brief: objective, constraints, acceptance criteria

  • Interfaces: inputs/outputs, owners, SLAs, guardrails

  • Risks: top 5 with mitigations, rollback plan

  • Defaults: toolchain, naming, data model, logging

  • Observability: dashboards, alerts, review cadence

  Postmortem (after any miss or incident)

  • Facts only timeline

  • Root causes (inverted): process, design, human, environment

  • Blast radius: what failed, what contained it

  • Fixes: smallest change that works, owner, due date

  • Learning: update principle, playbook, or guardrail; delete one thing

Clean Definitions for the one’s I am sure of:

• Pareto Principle: A minority of inputs generate a majority of outputs.

• Falkland’s Law: When no need to decide, the best decision is to delay.

• Hick’s Law: Decision time increases with the number of choices.

• Lindy Effect: Future life expectancy increases with current age for non-perishables.

• Shannon’s Principle: Information = reduced uncertainty; clarity beats verbosity.

• Chesterton’s Fence: Understand purpose before removal.

• Sutton’s Law: Address the most probable cause first.

• Metcalfe’s Law: Network value grows roughly with the square of user count.

• Le Chatelier’s Principle: Systems counteract imposed changes.

• Goodhart’s Law: Measures degrade when they become targets.

• Inversion: Solve by asking “how would this fail?” and avoiding that first.

• Hanlon’s Razor: Default to error/incentives over malice.

You stumble upon a song that gives you goosebumps. A film that feels like it rewired your brain. A passage in a book that vibrates with truth. It catches you off guard, disorients you briefly, and then... you want more. Not more of the unfamiliar, no; but more that feels exactly like that.

So you search for the sequel. The playlist. The "if you liked this, try that" suggestions. Something that scratches the same itch without the work of discovering again. It seems paradoxical. We say we crave novelty. But the moment we experience something new and powerful, we crave more of what feels familiar.

This isn't a glitch. It's a window into how our minds are wired for pattern, safety, and reward. This essay explores that loop. Why we form it, why we stay in it, and how to gently, intentionally break it.

The brain’s pleasure circuitry duo: dopamine and pattern matching

Let’s start with dopamine, often misunderstood as the brain's "pleasure chemical." What it really governs is anticipation. When we correctly predict a rewarding outcome, the beat drops where we expected, the twist lands with a satisfying punch and dopamine fires - bam! bam! bam!.

Over time, the brain becomes conditioned to associate specific patterns with pleasurable outcomes. We don't just enjoy that indie guitar riff, or that third-act redemption arc, we enjoy anticipating it.

We chase the structure, not just the content.

This is why repetition doesn’t bore us as much as we think it should. Instead, it soothes us. It confirms that our internal map of the world is working. The brain rewards predictability dressed up as discovery.

Evolutionary wiring : familiarity as survival

Go back thousands of years. A rustling bush could mean wind or a predator. A sweet-smelling fruit could be a meal or poison. Early humans didn’t have the luxury of curiosity for curiosity’s sake. Unfamiliarity was risky, and the slightest mis-step could kill you.

So we evolved with a bias toward what had previously helped us survive. Familiar trails. Known food sources. Repeated rituals. Known faces. Common languages. All signalled safety. Safety came not from exploration, but from repetition. That evolutionary programming hasn’t left us. Today, when we favour familiar movie structures or recognisable musical elements, it’s our ancient brain whispering: this feels safe. Even if the stakes are now emotional rather than existential. Unfamiliarity still triggers our caution circuits, even in a Netflix queue.

Cognitive ease - why the brain loves repetition

Cognitive ease is a term psychologists use to describe how fluently we process information. The smoother the experience, the more positively we judge it. That’s why we rewatch old shows, even if we know every plot point. Why we let our favorite album play again instead of exploring a new one. The brain says: I know this terrain. Let’s cruise.

Beyond survival, the brain just loves a shortcut. It accounts for 2% of body weight, but consumes 20% of our energy. Naturally, it looks for ways to conserve effort. This isn’t laziness, it’s energy efficiency. It’s emotional homeostasis. We call it "comfort viewing" for a reason.

Cultural echo chambers is how groups reinforce sameness

But the loop doesn’t live in our heads alone. Culture reinforces it. Taste becomes social currency. We form identities around what we love, and we look for others who love it too and tend to “group” together based on these. Fandoms emerge. Platforms notice. Algorithms optimize. And soon, what once felt unique becomes part of a well-oiled system of content delivery.

We don’t just consume. We validate, rank, recommend, replicate. Art becomes genre. Genre becomes brand. Brand becomes formula. And the formula gets repeated because it works. We don’t mind. We’re rewarded with the dopamine of shared experience.

Sameness scales. And when it scales, it self-perpetuates.

The contrarian mind craves dissonance, not harmony

But not everyone thrives in the loop. Some minds get restless when everything feels too known. They get suspicious when a pattern becomes too predictable.

These are the dissonance seekers. The contrarians. The ones who don’t want to watch the sequel. They want to be jolted, provoked, confused. Not because they dislike comfort, but because comfort wears off quickly for them. Often, these are also the creators. The genre-breakers. The people who put new shapes into culture, even when no one’s asking for it yet.

For them, boredom isn’t discomfort; it’s a compass pointing somewhere new.

Breaking the loop and how to rewire what we seek

So how do we escape this loop when we want to? Not by purging familiarity altogether. But by becoming aware of when it becomes default. We can disrupt the loop gently, with intention:

• Listen to a genre that doesn’t show up in your algorithm

• Watch a foreign film without subtitles

• Read a book you pick blindly off a shelf, with no blurb or review

Let curiosity be the filter instead of the recommendation engine. And ask yourself: Is this stretching me? Not because stretching is morally better. But because novelty expands emotional range. It reveals you to yourself in new ways.

Harmony between pattern and disruption

We need both. Pattern and rupture. Repetition and rebellion. Familiarity isn’t the enemy. But when it becomes an unconscious loop, it stops being nourishment and starts being numbing. If art is a mirror, sometimes we need to look into a new one.

Discover. Repeat. Then, every now and then, break the damn loop.

Additional resources: Break the Loop workbook. A reflective worksheet designed to help you recognise consumption habits, embrace discomfort, and rebuild your taste through intentional curiosity.

Use weekly or monthly. Best when done with complete honesty. Download it here

India 2025: A Booming Economy, A Dangerous Pattern

India’s macro story today is the world’s favourite headline.

• Manufacturing exports are rising

• Global giants like Apple, Foxconn, and Tesla are already manufacturing or setting up shops soon

• The country is the centerpiece of every “China+1” conversation

• Domestic demand is resilient, and the tech stack is admired worldwide

By all accounts, this should be India’s decade. But underneath this momentum lies a familiar and dangerous pattern.

The world is investing in India because it’s affordable, not because it’s indispensable.
Factories are humming. Jobs are being created. But at the core of this growth is a singular proposition:

India has the world’s largest pool of cheap, semi-skilled labor.

That’s not strategy. That’s arbitrage.

And if left unchecked, India risks locking itself into a low-cost, low-value identity.
Very similar to 1990s China, but without the structural levers China used to escape it. We’re seeing:

• No significant jump in skilled engineering talent

• No transfer of product IP or R&D to Indian control

• No wage progressions tied to learning curves

• No ecosystem design beyond physical plants and labor pools

We’re exporting output, but not growing ownership on any key current or future levers. We’re absorbing capital, but not creating compounders in the market or the broader economy.

The cheap labor story works for now. But if India continues to scale on cost instead of value, the next decade will leave us busier than ever, but not richer.

What China Did in the 1990s — And Why It Worked

To understand where we in India could be heading, it’s useful to look back at where China started. The China of the early 1990s looked surprisingly familiar:

• A vast, young, underemployed labor force

• Rising urban migration

• Government courting foreign capital

• Wages low, labor flexible, land abundant

• Global firms looking to shift manufacturing out of expensive markets

China leaned in — aggressively. But what’s often forgotten is this:

Cheap labor wasn’t China’s economic identity. It was just their entry ticket.

Here’s what China did differently — and why it worked:

China used low cost as a temporary hook, not a permanent model.

Cheap labor got multinational corporations in the door. But the state quickly built pathways to move workers up the value chain : from basic assembly to skilled manufacturing to domestic product design. Industrial zones weren’t just cost centers, they were training grounds.

China invested in hard and soft infrastructure simultaneously.

Logistics corridors, ports, railways, power grids; all scaled alongside manufacturing. But equally important, the following materialised in parallel:

• Dormitory cities for workers

• State-supported vocational schools

• Industrial apprenticeships tied to factories

• Healthcare and food subsidies for migrant workers

Urbanization was managed. Migration was supported and at times, incentivised. Productivity in real tangible terms was planned and supported with an extreme sense of urgency.

China built domestic champions alongside foreign contracts.

Foxconn assembled iPhones — but Huawei, Lenovo, Haier, BYD were growing in the background. China used FDI to finance factories — but always kept an eye on creating IP-owning, brand-owning domestic firms.

China didn’t just want jobs. They wanted control of what gets made, how, and for whom.

China didn’t treat cheap labor as a brand.

China never sold itself as “the cheapest place to do business.” It sold itself as “the most efficient, the fastest-scaling, the best-integrated supply chain on Earth.”

There was pride in the machine and output; not in the discount. Unfortunately, we still sell human capital as arbitrage, and seek pride in it.

China created a long runway for value transformation.

By the mid-2000s, China was already shifting to:

• High-speed rail

• Chip design

• Global e-commerce platforms

• Renewable tech

• Domestic consumption and tech-led growth

All the above enabled China to move from “Made in China” to “Designed, owned, and scaled in China” within 15 years.

India, today in 2025, is where China was in 1995; but without central planning, industrial coordination, or a clear roadmap for value creation. If we don’t internalise and learn from how China escaped the cheap labor trap, we risk copying its entry point without ever reaching its exit velocity.

What India Is Doing Today — And What’s Missing

India today is riding a surge of economic optimism.

• Apple has tripled iPhone production in India in just two years.

• Foxconn, Pegatron, and other global suppliers are hiring aggressively.

• Electronics exports are at record highs.

• The world’s largest contract manufacturers are setting up shop.

• PLI schemes, infra push, and a favorable geopolitical climate have positioned India as the next big thing.

  But what we’re building is not yet a value economy.

  What we’re building is a cost-efficiency machine.

For example, most states and central scheme are putting their 100% behind certain efforts to drive impact, yet most fall short. Among the myriad of reasons, a few key aspects that leads to us falling short is:

We in India absorbing low-skill capital, not building deep talent.

Apple may be manufacturing more iPhones here, but almost all the design, IP, and engineering value still sits outside India. We’re assembling phones — not innovating them.

Jobs are being created, yes. But they are:

• Line-level assembly roles

• Low-skill, repetitive work

• Easily replaceable labor

• With minimal training ladders or wage growth

The model is simple: Do what China used to do — but cheaper.

There’s no ecosystem building, just factory plug-ins.

In the cities where these manufacturing hubs are growing — like Sriperumbudur, Noida, or Hosur — we see:

• New plants

• Contract hiring

• Warehousing expansions

But very little in terms of:

• Supplier development

• SME capability building

• Local engineering colleges feeding in

• IP or design control moving locally

• New training institutes, new faculty training means, or industry collaborations for R&D

Bodies like NASSCOM and CII continue to be about formal meet-ups to self-glorify, rather than commit to initiatives that drive outcomes with regular success audits. They’re handicapped at their foundational levels, managing more PR mandates, and have low intent to drive social outcomes.

We’re inserting India at the cheapest point in the supply chain, without anchoring the supporting value ecosystem around it.

Beyond manufacturing, the same trap plays out elsewhere.

It’s not just Foxconn and Apple. This logic is creeping across sectors:

• Gig platforms that offer work, but no benefits or career ladder

• Contract-based IT delivery with low margins and zero IP retention

• Customer support centers that are just updated versions of the BPO wave

• Government apprenticeship schemes subsidizing wages without real skill transfer

We’re repeating the same mistake at scale:

More people working. No one advancing.

We confuse busyness for progress.

India looks busy. Very busy.

• More jobs are being reported

• More exports are leaving our ports

• More factories are hiring

But the quality of that work?

• Static wages

• High attrition

• No IP creation

• No productivity compounding

• No upward mobility

That’s not transformation. That’s treadmill growth.

India is in motion, but not necessarily in ascent. Unless we build systems that allow people to move from labor to leadership, from hands (technical) → head (cognitive)→ ownership (equity), we will remain a cheap, flexible, scalable; but replaceable workforce.

The Risk: Getting Locked into a “Cheap But Replaceable” Identity

Every country starts somewhere. There’s nothing inherently wrong with us in India using our labor advantage to enter global supply chains. China did it. Vietnam is doing it. Bangladesh and Mexico are doing it. The danger isn’t in starting cheap.

The danger is staying cheap — and getting stuck there.

India risks building a global identity that says:

• “We’ll do it for less.”

• “We don’t need IP, just volume.”

• “Don’t worry about upskilling — we’ll manage.”

• “Need replacements? There’s more where that came from.”

That’s not a reputation. That’s a liability. For a few key reasons:

No value moat, easy to exit, easier to replace

When you’re selected only for cost, you're the first to be replaced when someone else gets cheaper, or when automation makes your labor irrelevant. Global capital is brutally pragmatic.

If your only edge is being affordable, you’ll never be indispensable.

No social mobility, just churn

A job that starts at ₹12,000/month and still pays ₹21,000 after five years isn’t a livelihood, it’s subsistence.

Without progression:

• Workers never build wealth

• Communities never stabilize

• The next generation starts from scratch again

This isn't a workforce-oriented model. It’s an endurance test for the participants.

No ownership of output = no control over destiny

When a country never controls the design, branding, patents, or platforms, it becomes a supplier, not a creator. If we continue exporting only output, not ideas, we’ll always be working for someone else’s roadmap.

India will be visible in supply chains, but invisible in strategic value. And visibility for cost-purposes in business comes with a shelf-life.

No national brand of excellence

What do you associate with India globally today?

• Cost-effective back offices

• Software services

• Yoga and spirituality

• Bollywood

But where is the Indian product brand? The Indian design language? The Indian breakthrough in hardware, deep tech, or global platforms? If we don't build a brand for value and excellence, we will forever be associated with being “the cheaper version.”

It's not just economic. It’s cultural.

A country’s self-worth is shaped by what it builds and how the world sees its output. A decade spent being the world’s cost-center without ownership or upward mobility will shape how our youth see themselves: as replaceable labor, not builders of original systems. The high-intent, original thinkers will drift off to more greener pastures in search of better ROI on their efforts and for their quality of life.

That’s not the India we’ve promised ourselves. The longer we scale based on cost, the harder it becomes to scale based on value.

The cheap labor trap isn’t just a growth problem. It’s an identity crisis for India waiting to happen.

What India Must Do Differently — Now

We in India still have time. We’re at a critical juncture — the global shift away from China is real, the world wants to invest, and the domestic consumption engine is strong.

But we must design our growth path consciously, else we’ll default into becoming the world’s cheapest supplier; and nothing more.

Here’s how we can change the trajectory:

Make skilling a strategic function, not a CSR initiative

Skill India and NSDC-type programs have scale, but little depth. They are supply-driven, not tied to real employer demand or progression planning.

We need:

• Sector-linked skill ladders (like Germany’s dual apprenticeship models)

• Paid learning with wage progression

• Tiered certification that ties to promotions

• Public-private platforms that reward skill mobility

Incentivize domestic IP and platform creation

It’s not enough to “make in India.” We must also design in India, own in India, and scale globally from India. Policy moves like PLI must be paired with:

• R&D credits and tax incentives

• Public funding for design and prototyping

• Preferential procurement for Indian IP

• IP protection frameworks that encourage founders and engineers to stay

If we don’t create local innovation, we’ll always be someone else’s assembly line.

Build sector-specific value chains, not general infrastructure

India’s industrial push often focuses on generic infrastructure: roads, ports, power. Important, but insufficient.

We need:

• Integrated textile cities (design to export)

• Pharma clusters with clinical trial and IP integration

• Electronics parks that go beyond contract assembly

• Food processing corridors that connect farmers to global packaged goods value chains

Every sector needs its own “idea to invoice” ecosystem.

Make cities hubs of mobility, not just migration

Today, cities absorb labor but don’t enable upward movement. We need:

• Urban housing designed for workers, not just luxury

• Public transport that connects industrial zones to residential belts

• Vocational campuses embedded in city plans

• Health and safety nets for the contract economy

A city should not just host labor — it should level them up. Unfortunately, we’re currently levelling down our urban dwellers, with the even-lasting, irrational real-estate cost bubbles.

Shift national branding from cost to quality

We must stop celebrating “cheap labor” as an asset. Instead, let’s build a national identity around:

• Frugal engineering (read: not jugaad)

• Platform-level thinking

• Design-for-scale

• Trusted manufacturing with ethical labor standards

India must become synonymous with reliable, skilled, high-leverage output — not just affordable effort.

The goal: Value per worker, not cost per worker

Every policy, every investment, every reform must ask:

“Will this make the Indian worker more valuable next year than they are today?”

If the answer is no, we’re scaling sideways, not upwards.

Conclusion: From Labor-Rich to Value-Rich

India has the numbers. We have the youngest population, a growing industrial base, world-class digital infrastructure, and a global moment of attention. But we must decide what kind of economy we want to become.

• Do we want to be known for our volume, or for our value?

• Do we want to build for the world’s cost advantage, or for our own capability advantage?

• Do we want to keep exporting labor hours, or start exporting ideas, systems, and ownership?

We are labor-rich, yes; but that’s not the endgame.

The goal is to become value-rich.

To go from “we can do it cheaper” to “no one can do it like us.”

That requires:

• A mindset shift from cheapness to excellence

• A policy shift from output to outcomes

• A public narrative shift from pride in numbers to pride in capability

India doesn’t need to replicate China’s 1990s strategy. We can leapfrog it — if we chooses to design differently.

This is our decade.
But it won’t be won by labor alone. It will be won by moving fast, and moving up.

In the crowded alleys of Mumbai’s Dharavi or the dense markets of North Chennai, small businesses operate under immense pressure. A garment workshop tucked into a single room, a kirana store balancing daily cash flow, or a small logistics firm managing deliveries on battered roads—they all represent India’s gritty entrepreneurial spirit. But these literal narrow corridors also reveal a deeper, systemic constraint: building profitable businesses in India is incredibly hard.

Why? Because costs have gone global, but consumer income remains deeply local.

In a country where per capita income hovers around $2,500, business builders are forced to pay urban land costs comparable to Tokyo or San Francisco. Logistics costs remain among the highest in the world. The result is a recipe that stifles scalability, compresses margins, and depresses opportunity.

For eg., in India’s knitting belt in Tirrupur (near Coimbatore), over 35% of the small businesses have shut down in the last 3 years, unable to extract any margins between the high cost of input goods, and stagnant demand.

For every entrepreneur that sets up his shop or business; hundreds fold or stagnate, unable to break even in a market that demands affordability but imposes high structural input costs.

It’s not a lack of talent or drive that limits their growth—it’s the economic physics of trying to deliver world-class products and services to a price-sensitive consumer in an environment where fixed costs are unforgiving.

In India, hustle is abundant, but margin is elusive.

A Brief History of Margin Erosion

There was a time, in the early 2000s, when land was still relatively affordable and infrastructure projects like the Golden Quadrilateral promised better connectivity. The GST rollout and freight corridor projects were seen as catalysts for economic acceleration.

But while costs moved up, income growth didn’t keep pace. And unlike China in the 2000s, India didn't create a margin-rich middle class fast enough.

The central, state level and local attempts at structural reforms that spurred bits and pieces of growth were uneven, fragmented by geography, and often delayed by bureaucracy. For a population of a billion plus people clustered in cramped cities, Hyderabad and Bengaluru are the only global cities we have managed to build up in the last 30 years from scratch. Even in these cities, the infrastructure always seems to be playing catch up, the cities seems to be chocking itself due to pollution, traffic etc., but the real-estate bubbles continue to grow larger and larger. This means that while inputs became expensive, the average consumer will not see proportionate increases in purchasing power.

As India's urban centers transforms rapidly in terms of aspiration and architecture, this transformation often leaves behind the fundamental economics that make businesses sustainable - the amount of margins a business could make.

Blame it on short-sightedness, lack of support or system handicaps; what should have been a boon of demographic advantage in India is now turning into a pressure cooker of low-yield opportunities.

Today in India, the gap between input costs and monetizable demand has only continued to widen over time.

The Capital-Intensity Trap

India today suffers from a paradox:

Building businesses here requires global capital, but returns local rupees.

Land in Indian metros is among the most expensive in the world. Logistics consumes 14% of GDP—far higher than the 8–9% benchmark in developed markets. The average logistics cost per GDP among BRICS countries is at 11 percent. However, in the case of India, the cost share per GDP was 16 percent as of 2024.

Meanwhile, consumption remains shallow, driven by value-seeking behavior rather than premiumization. This means many industries are caught in a capital-intensity trap: they need to invest heavily to reach customers, but the return per customer is too small to justify the outlay. This crushes innovation and suppresses growth.

Even in asset-light industries like digital services, where one might expect cost efficiencies, the realities of CAC (Customer Acquisition Cost), regulatory compliance, and market fragmentation introduce unsustainable capital burdens.

For capital-heavy sectors such as manufacturing, logistics, retail, and energy infrastructure, the scale required to achieve breakeven is so massive that it crowds out new entrants. Therefore margins, which should be a reward for efficiency and insight, has become a luxury available only to incumbents or conglomerates with deep pockets.

When Margins Vanish: Real-World Case Studies

Let’s examine a few sectors where the squeeze is most visible:

1. Organized Retail:
Retail chains require large-format stores in high-footfall locations. But rents in top cities are exorbitant, while customer spends remain low. Even giants like Reliance Retail and D-Mart navigate this tightrope with difficulty, passing on the burden to their suppliers and squeezing out most margins.

Even with the rising trends of dark stores and on-demand delivery, the economics of per-square-foot profitability become shaky when each basket size remains under ₹500. Inventory turnover must be lightning-fast, and operational efficiency becomes a make-or-break metric. Meanwhile, smaller players and independent shops are completely priced out of urban retail landscapes.

2. Diagnostics & Healthcare:
To serve Tier 2 and 3 cities, diagnostic labs need hub-and-spoke logistics. But demand opportunities are too low to absorb the high cost of transport, quality control, and last-mile service. Latest D2C entrants who have tried to disrupt this industry, stand out as textbook case-studies of companies investing in CAC, yet not building the kind of margins or cashflows seen by traditional players.

Moreover, regulatory requirements, equipment depreciation, and skilled personnel add fixed costs that are hard to dilute. Health outcomes therefore suffer when these providers have no cushion for quality improvements or R&D.

3. EV Charging Infrastructure:
Setting up a charging station demands significant investment in land, power, and tech. Yet each charging session yields just ₹200–300. The math only works with enormous scale.

Until EV density grows significantly, operators are subsidizing usage, betting on long-term payoffs. For now, it’s a loss-leading game few can afford to play.

4. EdTech for India:
Acquiring customers digitally is expensive. But Indian families are unwilling to pay international-level subscription fees. CAC is dollar-priced, but ARPU (Average Revenue Per User) is rupee-based. This means long payback periods, high churn, and low pricing power. EdTech players either burn capital or pivot abroad—but rarely thrive in the domestic consumer market. Yet traditional offline players have built healthy business, with razor-sharp focus on every unit rupee spent per operating square foot. Unfortunately, the thin margins behind underlying unit economics mean we have more coaching centres, and less research institutes.

These are not isolated examples—they reflect a structural misalignment. The margin squeeze is not about poor structural management; it’s baked into the ecosystem.

The Crushed Opportunity Stack

When margins collapse, it’s not just businesses that suffer. A whole chain of opportunity is crushed:

• Wages remain stagnant because businesses lack headroom to pay. In turn, consumer demand stays suppressed, completing a vicious cycle.

• Income mobility is stalled as gig workers and MSMEs operate with zero leverage. Every rupee earned is reinvested into survival, not growth.

• Equity creation becomes rare; few startups can scale to profitability or IPO. Early-stage capital dries up, and secondaries remain elusive.

• Economic output slows as capital chases fewer viable investments. Instead of fostering a startup culture, we end up breeding survivorship anxiety.

In a country with abundant labor, youthful demographics, and entrepreneurial energy, the lack of margin becomes a binding constraint. It acts as a silent tax on our ambition.

The result is what might be called a "thin-air economy" — high activity, but low altitude. There is movement, but little uplift. Motion without margin is just plain friction.

R&D is a Cost Center Before It’s a Growth Driver, so no margins = no R&D.

Margins aren't just a signal of profitability—they're a prerequisite for innovation.

When a business has room to breathe financially, it can allocate resources toward research and development (R&D), which is essential for long-term differentiation and global competitiveness. Diving into the key aspects, margins are critical for R&D because:

1. R&D is a Long Game: Innovation takes time. It involves experimentation, failed attempts, and iteration. Without financial slack, businesses can't afford to take the risks necessary to innovate.

   A lot of iterative research and development doesn’t yield instant returns. In most cases, it involves years of experimentation, failed prototypes, and iterations before something viable emerges. Companies with thin or no margins can’t afford to invest in long-gestation projects—they're too focused on short-term survival.

2. R&D Requires Talent and Infrastructure: Building a serious R&D pipeline means hiring scientists & top talent in each domain (countering brain drain with capital incentives and social incentives), investing in lab infrastructure, and filing IP. These are expensive and long-gestation costs that only margin-rich businesses can consistently support.

   Without margin, you can't attract minds. Without minds, you can't build breakthroughs.

3. It Builds Competitive Moats: Margin enables reinvestment, and reinvestment drives product differentiation. This, in turn, builds a sustainable moat that shields businesses from price-based competition.

   High-margin businesses can reinvest profits into improving their products and processes—creating IP that acts as a moat. Low-margin businesses are forced to compete on price, which is a race to the bottom.

   Example: Apple, Nvidia, and Tesla succeed not because they’re the cheapest—but because they reinvest massive portions of their margin into product innovation.

4. It Fuels Future Growth: While efficiency helps survive today, R&D helps win tomorrow. Without margin, pretty much all companies get stuck in a cycle of survival, or keeping shareholders happy with whatever dividend yield that can be squeezed out. With margins, they can lead markets and shape categories.

   Today in India, most firms focus on operational efficiency over innovation because their margins are too thin to do both. But R&D is what will allow a company to jump curves—to move from a commodity to a differentiated offering.

   Without margin, you optimize for today. With margin, you invest in tomorrow.

5. Policy Incentives Aren’t Enough: Government grants or tax credits help, but they don’t replace the need for internal profitability. Sustainable R&D requires an internal capital surplus.

Simply put, margin is the oxygen R&D needs to thrive, and we’re running abysmally low on it.

The Narrow Corridors That Work

Still, there are niches where high-margin models do thrive. These businesses have one thing in common: they escape the land-logistics-income mismatch.

What makes these models viable is their insulation from the local constraint loop. Whether by tapping global markets, optimising for capital efficiency, or compressing logistics costs through tech, they sidestep the drag faced by other businesses.

For example, an Indian SaaS company serving U.S. clients can price in dollars while operating with Indian salary structures. A direct-to-consumer skincare brand like Sugar can command premium pricing by building a loyal brand following rather than renting expensive retail space. A diagnostic company can serve premium tests using custom pricing power to dictate costs and ensure volumes.

These corridors are narrow—they require precision, planning, and positioning. But they prove that margin is not impossible.

Widening the Corridor: Structural Recommendations

To unlock margin at scale, India needs a structural shift:

1. Land & Industrial Reform
Zoning, long-term leases, and end-use control can de-risk land acquisition and reduce speculation. State governments must coordinate with industry clusters to pre-approve land for specific use cases, allowing SMEs to rent, not buy. Industrial parks should emphasize ease of doing business over merely offering subsidies.

2. Freight & Rail Investment
Reduce logistics cost from 14% of GDP to 9% through multimodal transport, warehousing standardization, and digital tracking. A rail-led logistics spine could shift pressure off road networks and dramatically improve cost efficiency. Public-private partnerships must emphasize operational uptime and interoperability.

3. Digitally Native Infrastructure
Public platforms like ONDC, UPI, etc., can compress cost and expand reach. These platforms can enable real-time reconciliation, lower CACs, and democratize access to distribution channels. Their power lies in creating a digital public utility layer for commerce.

4. Skilling for Value, Not Volume
Create workforce pipelines that match global demand, not just domestic surplus. Focus on cognitive, design, and analytical skills, not just coding bootcamps. Leverage platforms like NSDC to partner with global recruiters and deliver placement-linked skilling programs.

5. Wage-Led Growth
Focus on increasing real wages via urban job creation, formalization, and SME support. Instead of relying on trickle-down, policy should aim for trickle-out: improving productivity in middle-tier firms, incentivizing hiring, and embedding labor into value-added services.

These changes are not cosmetic. They will redefine the geometry of opportunity in India.

Conclusion: Margin as Destination

India is not short of entrepreneurs or capital. What it lacks is margin. Margin is what lets you reinvest. It’s what allows businesses to pay better, grow faster, and dream bigger. Without margin, scale becomes a mirage.

For too long, Indian businesses have been told to make do. To work harder, stretch more, and hustle endlessly. But hustle without margin is just strain. The real unlock lies in creating structural pathways where good ideas can become sustainable, profitable enterprises without being crushed by fixed costs and frugal demand.

In a country as vast and diverse as India, business shouldn’t be a narrow corridor. It should be a wide highway.

But to build that, we must confront our cost structures, rethink our development model, and make room for margin—everywhere.

(Pic courtesy : pixabay.com ; https://picryl.com/media/new-dehli-overpopulation-chaos-transportation-traffic-f0f0b7)

Subscribe now

Rentals in India today are a ticking time-bomb.

India is urbanizing rapidly. By 2036, its towns and cities will be home to 600 million people, or 40 percent of the population, up from 32 percent in 2013, with urban areas contributing almost 70 percent to GDP.

How well India manages this urban transformation will play a critical role in realising its ambition of becoming a developed country by 2047, the 100^th year of independence.

Yet nearly 70 percent of the urban infrastructure needed by 2047 is yet to be built, sizeable investments will be required. But with the current real estate and rental rates, we might be left forever playing catchup with our dreams - and create mass urban and economic chaos in the process.

1. Why worry about this now

In major Indian cities, rents have outpaced income growth by alarming margins. A report by Knight Frank India revealed that rental yields in cities like Mumbai have risen significantly in the last decade, leaving middle-class families struggling to afford basic housing. This burden forces many to live in smaller spaces or relocate to suburban areas with inadequate infrastructure and long commutes.

The result? Urban inequality deepens as cities become playgrounds for the wealthy, while lower- and middle-income groups are pushed further to the fringes.

The disappearing around-the-corner businesses today

Primary Care Centers: A journey of decline

Take the case of a traditional clinic in Fazer town, part of the “old” Bangalore city limits. The clinic started with a rent of INR 40,000, generating a healthy monthly revenue of INR 1,50,000. With staff salaries of INR 50,000 and utilities at INR 20,000, the clinic achieves a modest profit of INR 40,000.

Rising Rents: Over five years, the rent increases to INR 70,000. With revenues largely stagnant, profit margins shrink to INR 10,000. Further hikes will leave no room for profitability, forcing the clinic to shut down or raise fees, compromising healthcare access.

Day schools & tuitions : private, smaller education spaces are facing mounting pressure

Mostly run after hours with couple of teachers from private schools who used tuition fee to supplement family income, a small centre would pays INR 2,00,000 in rent while generating INR 6,00,000 in monthly revenue. Teacher salaries and operational expenses total INR 3,70,000, leaving a net margin of INR 30,000.

Rising Rents: A 25% rent hike over three years pushes rent to INR 2,50,000. Margins turn negative, compelling the centre to increase tuition fees. This limits access to education for middle-income families, creating a vicious cycle of exclusion and financial instability.

Kirana Shop: from local staple to death-by-dark-stores

Initial Stage: The shop operates with a rent of INR 15,000, making INR 1,00,000 in monthly revenue. With inventory costs of INR 50,000 and other expenses totaling INR 15,000, it earns a profit of INR 20,000.

Rising Rents: Rent increases to INR 25,000, reducing the profit margin to INR 10,000. Another hike wipes out profitability entirely, leading to closure. This disrupts local supply chains, increasing dependence on larger, costlier retail chains.

Pharmacy: essential services under threat

Initial Stage: A pharmacy rents a storefront for INR 40,000 while earning INR 2,00,000 monthly. Expenses include INR 1,00,000 for inventory and INR 50,000 for salaries and utilities, leaving a profit of INR 10,000.

Rising Rents: A 20% rent hike increases costs to INR 48,000. With slim margins, any additional increase renders the business unsustainable. This results in reduced public access to essential medicines in urban areas.

The role of speculative real estate in driving up rental irrationality

Speculative real estate activity in India is one of the biggest deterrents to equitable progress. Investors buy and hold properties or land not to use them, but to profit from their appreciation. This practice inflates property prices and rental values far beyond what the local economy can sustain, creating artificial scarcity and driving up rents for residents and businesses alike.

Speculation distorts markets

1. Artificial Scarcity: Speculators hold onto properties, keeping them vacant, or are driven by the need for higher ROI; reducing supply in the market, and driving up rents.

2. Price Bubbles: Speculation inflates property prices, leading to unsustainable rental costs that do not align with wage growth or business revenues.

3. Impact on Essential Services: Essential services like schools, clinics, and small businesses are priced out of prime locations, as landlords prioritize high-margin speculative deals over long-term, stable tenants.

4. Urban Inefficiency: Large swathes of urban land remain underutilized or undeveloped, while businesses and families are pushed to less accessible areas, increasing commuting times and infrastructure strain.

Why regulating this spiralling speculative market is critical

Regulating this speculative real estate market is essential to ensure economic stability, equitable access to housing, and sustainable urban development. Speculative activity over the last decade is showing up in inflated property prices and rental rates, creating artificial scarcity that disproportionately affects middle- and low-income groups. Essential services like schools, clinics, and small businesses are already priced out of prime locations, disrupting communities and worsening economic inequality.

Unchecked speculation leading to inefficient use of urban spaces is already all around us - there’s more pubs than essential services, with vacant properties hoarded for profit while demand for affordable housing remains unmet. Implementing measures such as taxing vacant properties, capping speculative investments, and enforcing transparent ownership records can deter speculative practices. These regulations, coupled with rental controls, create a fairer market, stabilize rents, and ensure cities remain accessible and livable for all residents.

1. Curbing Vacant Properties: Policies to tax vacant properties at higher rates can disincentivize speculative hoarding of real estate.

2. Rental Controls: Implementing rent caps limits the impact of speculative price hikes on tenants and businesses.

3. Transparent Ownership: Mandatory disclosure of property ownership and transaction details can reduce opacity in the real estate market and deter speculative investments.

4. Public-Private Partnerships: Governments can work with developers to prioritize affordable housing and commercial spaces instead of speculative luxury projects.

Global Lessons

• Hong Kong’s Vacancy Tax: To address housing shortages, Hong Kong introduced a tax on vacant new residential units to discourage developers from hoarding properties.

• Canada’s Anti-Speculation Measures: Canadian cities have implemented taxes on foreign and speculative real estate investments to cool housing markets and stabilize rents.

Why would rent control help?

Rent control boosts profits for businesses primarily by stabilizing one of their largest fixed costs—rent. Here’s a breakdown of how it impacts profitability and sustains businesses:

Stabilized rent costs

• Without Rent Control: Adding the fact that India is a speculation driven market and not a investor driven market - Rents rise unpredictably, often outpacing inflation and revenue growth. This squeezes profit margins and forces businesses to either cut costs, raise prices, or close down.

• With Rent Control: By capping rent increases, businesses can better forecast and manage costs over time, preserving profitability.

Improved financial planning

• Predictability: Rent control allows businesses to anticipate future costs and allocate resources to areas like hiring, inventory, or service quality instead of firefighting rising rents.

• Less Volatility: Stable rents reduce financial shocks that could derail a business's operations or expansion plans.

Encourages long-term investment

• Without Rent Control: Rising rents discourage long-term investments in infrastructure, technology, or employee development, as the cost of staying in a location might make these investments unsustainable.

• With Rent Control: Businesses are more likely to invest in growth and improvement when rent costs are predictable and capped, enabling better service and higher productivity.

Reduced pressure to raise prices

• Without Rent Control: Businesses often pass on the burden of higher rents to consumers through increased prices, which can reduce demand.

• With Rent Control: Stable rents allow businesses to maintain competitive pricing, fostering customer loyalty and sustaining revenue.

Enhanced profit margins

• Rent control prevents disproportionate increases in operational costs, directly protecting profit margins. This is particularly important for businesses operating with thin margins, like small retailers or essential services.

The need of the hour: Rental controls and Zoning regulations

Implementing rental controls

Rental controls are policies designed to stabilize rents and protect tenants from excessive increases. Countries like Germany and cities like New York have successfully used these measures to make housing more affordable.

For India, this could include:

• Capping Rent Hikes: Set limits on annual rent increases to prevent sudden spikes.

• Standardizing Lease Agreements: Introduce clear guidelines to balance the rights of landlords and tenants.

• Promoting Affordable Housing: Incentivize developers to build housing for middle- and low-income groups through tax breaks or subsidies.

Direct impact on inflation

Reduced Housing Costs:

• Rent is a significant component of the Consumer Price Index (CPI) in most economies. By capping rent increases, rent controls directly limit the rate at which housing costs contribute to inflation.

• In markets with steep rent hikes, rent controls can decelerate inflationary pressure.

Stabilized Consumer Spending:

• When rents rise rapidly, households spend a larger portion of their income on housing, leaving less for other goods and services. Rent controls can stabilize housing costs, enabling consumers to maintain consistent spending in other sectors, which may help moderate overall inflation.

Indirect impact on inflation

1. Mitigation of Secondary Effects:

   • High rents can indirectly increase the cost of goods and services, especially in urban areas. For example, businesses facing higher rents often pass these costs onto consumers through price increases. Rent controls for commercial spaces may reduce these knock-on effects.

2. Labor Market Stability:

   • In areas with soaring rents, workers demand higher wages to afford housing, contributing to wage-driven inflation. Rent controls can help ease this pressure, making it easier for businesses to control labor costs.

Examples of rent controls and inflation

• Berlin (2020): A five-year rent freeze initially reduced housing costs and stabilized inflation in the rental market. However, concerns arose about housing shortages due to reduced incentives for developers.

• New York City: Long-standing rent control policies have helped stabilize rents for many tenants, but critics argue that they contribute to reduced housing supply and higher prices in the unregulated market.

Potential risks to inflation management

1. Reduced housing supply:

   • Critics argue that rent controls might discourage new housing construction or maintenance, leading to shortages over time. This scarcity could cause housing prices in uncontrolled markets to spike, adding to inflationary pressures. Considering that today’s affordability problem in most urban cities in India, we’re already at the other end of the spectrum - where people do not get housing due to exorbitant rental and housing prices.

   • To mitigate this, rent controls should be paired with policies that incentivize new housing development and ensure long-term supply.

2. Distorted market dynamics:

   • Rent controls could create disparities between controlled and uncontrolled segments of the housing market. For instance, while rents in controlled properties remain stable, uncontrolled properties might experience rapid price increases, distorting inflation metrics.

Testing the water for rental control in India - A phased approach

Conduct a baseline assessment in newly developing suburbs and townships outside existing cities

• Objective: Understand regional rent dynamics, affordability thresholds, and market disparities.

• Steps:

  • Map urban and semi-urban areas with severe rental inflation.

  • Identify tenant demographics, income levels, and housing supply-demand gaps.

  • Classify cities and districts based on rent pressure (e.g., high, moderate, low).

Pilot rent control policies in high-pressure areas

• Objective: Test the effectiveness of rent control policies before larger implementation. States should set up controls and mandates for successful implementation.

• Steps:

  • Select metro areas with skyrocketing rents (e.g., Mumbai, Bengaluru).

  • Implement rent caps tied to inflation or a fixed percentage (e.g., 5-7% annually).

  • Ensure properties below a certain size or rent threshold (e.g., affordable housing units) are prioritized.

Create a NRP - National Rental Policy Framework

• Objective: Establish regulatory consistency across states while allowing local adaptation.

• Components:

  • Standardized Rent Caps: Define limits for rent increases annually, tied to inflation or wage growth.

  • Long-Term Lease Incentives: Encourage leases of 3+ years with controlled rent increases for stability.

  • Exemptions: Exclude newly constructed properties for a limited period (e.g., 5-10 years) to incentivize housing supply.

  • Tenant-Landlord Dispute Resolution: Set up rent tribunals to handle grievances and ensure compliance.

Address key stakeholder concerns

• Landlords:

  • Offer tax breaks or subsidies for maintaining affordable rents.

  • Provide low-interest loans for property maintenance to offset reduced rent revenues.

• Tenants:

  • Protect against arbitrary evictions.

  • Introduce mandatory lease agreements with clear terms for rent increases.

• Developers:

  • Incentivize construction of affordable rental housing through subsidies, fast-track approvals, or land grants.

Introduce taxation on speculative real estate

• Objective: Discourage property hoarding and speculative practices.

• Steps:

  • Tax vacant properties at higher rates to bring unused housing into the rental market.

  • Impose additional taxes on short-term property flipping to stabilise housing costs.

Establish digital infrastructure for transparency

• Objective: Monitor rent control compliance and improve market efficiency.

• Steps:

  • Develop a centralized rental database for lease registration and rent trends.

  • Enable tenants and landlords to report violations or access dispute resolution services online.

Expand rent controls to Tier-2 and Tier-3 cities

• Objective: Prevent spillover effects from rent-controlled metros to smaller cities.

• Steps:

  • Gradually extend policies to semi-urban areas experiencing rapid rent increases.

  • Customize rent control policies to match local economic conditions and housing markets.

Monitor and revise policies regularly

• Objective: Ensure policies remain effective and adaptable.

• Steps:

  • Conduct periodic reviews (every 3-5 years) to assess policy impact on housing supply and affordability.

  • Adjust rent caps, exemptions, or incentives based on market feedback and economic conditions.

Challenges to anticipate

• Housing Supply Constraints: Balance rent control with policies that incentivize new construction.

• Market Distortions: Address potential disparities between controlled and uncontrolled properties.

• Implementation Costs: Allocate resources for monitoring compliance and resolving disputes.

This builds on the point-of-view from my tweet thread here. By phasing rent control implementation in this manner, India can tackle rising rents systematically, protect vulnerable populations, and encourage equitable urban development while avoiding the pitfalls of rushed or blanket policies. The next part will cover on the need for zoning regulations in India.