The Curious Mind: When Consensus Is Camouflage
Surviving Inflection Points, When Price Is A Liar, Explaining Einstein
“Information is just bits of data. Knowledge is putting them together. Wisdom is transcending them.”
- Ram Dass
"Never tell people how to do things. Tell them what to do and they will surprise you with their ingenuity."
- George S. Patton, War As I Knew It.
"Stability is destabilising."
- Hyman Minsky.
"The reasonable man adapts himself to the world; the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man."
- George Bernard Shaw
"A society grows great when old men plant trees in whose shade they shall never sit."
- Greek proverb
Hi. I am in Kauai for a few days with family. One of the beauties of summer holidays apart from time with the family is being able to step away from things you take for granted.
That partly is the core idea of today’s email: the most important things are invisible from inside the system that needs to see them, and the standard instruments (the map, the price, the consensus) aren’t just failing to show you, they’re actively reassuring you.
Surviving Inflection Points
Angus Fletcher spoke to Rick Rubin. We had discussed Fletcher’s awesome book Primal Intelligence in March in the essay, You Brain Is Not A Computer.
The big idea from the conversation is that systems built on the past (logic, data, rules, playbooks) fail at inflection points, and inflection points are where all the value is created and destroyed. Sounds obvious but so much of our world is built on the past.
5 BIG IDEAS:
Probability is priced in. Possibility is the only edge left: Fletcher’s deepest claim isn’t “humans are creative” — it’s structural: computers compute probability (what happened before), humans compute possibility (what hasn’t happened yet). Probability requires data frozen in time, which means every probability-based system — quant models, economic forecasts, LLMs — is definitionally backward-looking and degrades exactly when regimes change. AI commoditises everything derivable from historical data, returns migrate to whatever isn’t in the training set.
The anomaly is the signal. The average is the trap: Statistics vacuum up data and return the mean, which tells you what’s already consensus. Darwin didn’t survey; he stared at weird finches. The operational discipline is his conversation technique: never ask why (why-questions retrieve someone’s canned explanation, an old answer smuggled into a new question), ask who/what/when/where/how and follow surprise.
Buffett doesn’t underwrite the plan. He underwrites the improviser: A pre-existing plan for a genuinely novel event is proof they’re bluffing or pattern-matching. The pass is a pause followed by live improvisation. You’re not buying the answer; you’re buying the answer-generating machine.
Tell people where. Never tell them how: The recruit who ran around the obstacle course and rang the bell in record time wasn’t thrown out — he’d nailed the objective, and that’s the entire special-forces psychology: fixed target, infinitely fluid path (their model, unexpectedly, is Beethoven — same destination every time, never the same route). Most KPI/OKR cultures fail on both axes at once: abstract targets and prescribed methods.
Anxiety is a volatility gauge and almost everyone trades it backwards:
Anxiety isn’t dysfunction, it’s your brain’s real-time measurement of how fast the environment is changing. Common sense is matching the newness of your response to that reading: familiar situation, run the SOP; novel situation, throw the playbook out. That’s a precise description of market behavior — max leverage at the calm top, liquidation into the panic.
Curious Thinking: The past shouldn’t be a guide, it doesn’t matter if it hasn’t happened before or no one has done it before. Focus on possibilities not probabilities. This might be even more true going forward. Don’t spend time with “experts” but people actually people doing shit. Or as Taleb would say, “don’t tell me what you think, tell me what you are doing”.
When Price Is A Liar
Matthew Smith spoke to Invest Like The Best. Matt is the founder and CIO of Chronometer Partners. They cover how LNG exports and the demand for compute are creating a structural gas shortage, the infrastructure bottlenecks that cap how fast new supply can reach the market, and why building more large-scale nuclear reactors is the only durable solution.
It’s a story about a system whose warning lights all failed at once. Nothing looks wrong.
The 5 BIG IDEAS:
Nobody has actually voted on 2028: Here’s the thing that got me. Everyone says the market disagrees with Smith. Gas is $3.50 out to 2030, the curve is flat, case closed. But almost nobody trades 2028 gas. The curve is flat because it’s empty. And yet that empty price is being used everywhere. Hyperscalers plug it into their models to justify the biggest capex boom in history. Utilities look at it and don’t hedge. Producers look at it and don’t drill. Everyone is making real decisions off a number that no one with real information ever set. The data centers being justified by cheap gas are the thing that makes gas expensive.
Every crisis has a shock absorber, and the shock absorber is why nobody sees it coming: Gas prices have spiked before. Polar vortex, Russia-Ukraine — $8, $10 — and every time they came back down. So people learned the lesson: spikes are temporary. But they came back down because of storage. Four trillion cubic feet sitting underground. Storage is why gas prices behave. Smith’s math says storage runs out. And when the buffer is gone, you’re not in a market anymore, you’re in an auction — a Japanese utility with a contract, a data center that can’t go dark, and a family heating their house, all bidding on the same molecule. The lesson is that don’t watch the price, watch the buffer.
Reminded me of John Burbank’s statement: Price is a liar.
The future is already built. Everyone’s forecasting something that’s already decided: This is the part people miss about physical stuff. A processing plant takes three years. A pipeline takes a decade of permits — we’ve built one in twelve years. LNG terminals are financed and contracted years ahead. Which means 2028 isn’t a forecast. It’s arithmetic. If it’s not under construction today, it doesn’t exist in 2028. The people who own the molecules are already betting on the shortage. But their bet is invisible — it shows up as a lower rig count, which everyone reads as “see, no shortage coming.” The smartest money in the room is positioned, and its positioning looks exactly like complacency.
When every solution needs the same ingredient, the ingredient is the story. Look at the fixes everyone’s excited about. Fuel cells? Run on gas. Distributed gensets? Gas. “Bring your own generation”? Gas. More Permian supply? Only works if oil prices go up enough to hurt the same consumers you’re trying to protect. There’s a useful habit here: when you want to know what’s actually scarce, don’t listen to what people say. Look at what all their solutions consume. The remedies are the confession. In a bottleneck, you don’t have to own the bottleneck. Solar’s fuel is free forever, and gas sets the power price — so every dollar gas rises is pure margin for assets that spent nothing to get it. The best seat isn’t always at the choke point. Sometimes it’s the seat that gets paid by the choke point and owes it nothing.
The real shortage is people who remember how to build things: The money for nuclear exists — $260 billion sitting at a government loan office. The plans exist. Nothing happens, because nobody wants to build reactor number one. And honestly, that’s rational. Whoever goes first eats the disaster — Vogtle 3 took fifteen years and 3x the budget. Whoever goes second gets all the lessons for free — Vogtle 4 was dramatically better. So everyone waits to be second, and no one is first, forever. But underneath the game theory is something sadder. For thirty years America didn’t build reactors, and the knowledge didn’t sit in a warehouse waiting. It retired. It died. It moved on. China built 39 reactors while we forgot how — a third of them based on our design. Skill is a depreciating asset that appears on no balance sheet. You find out it’s gone only when you attempt the thing. Smith’s czar-for-a-day answer: government just builds the first four, whatever it costs — isn’t really about electricity. It’s paying tuition to re-learn something we already knew.
Curious Thinking: Loved the idea around the best seat isn’t always at the choke point. Sometimes it’s the seat that gets paid by the choke point and owes it nothing. Also what is a depreciating asset on your own balance sheet?
Explaining Einstein
Have always been fascinated by Relativity and have unsuccessfully tried to read Stephen Hawking’s classic: A Brief History of Time many times. So I was excited when I saw Dwarkesh interviewed Adam Brown. Adam is a founder and lead of BlueShift which is cracking maths and reasoning at Google DeepMind and a theoretical physicist at Stanford.
You can think of the episode as A Brief History of Time's greatest hits, updated with 35 years of science.
The 4 BIG IDEAS:
Einstein’s breakthrough started with a “coincidence” everyone else ignored for 200 years: Mass does two unrelated jobs in physics. It measures how much an object resists being pushed (why a truck is harder to shove than a bike), and it measures how hard gravity pulls on it. There’s no logical reason these should be the same number — in electricity, the equivalent properties (mass and charge) have nothing to do with each other. Yet for gravity, they’re identical to 15 decimal places. Einstein’s insight: there’s exactly one type of force where this equality is guaranteed rather than accidental — fake forces, like the centrifugal push you feel in a turning car. Those aren’t real forces; they’re just your own inertia experienced from an accelerating viewpoint. So Einstein asked: what if gravity is a fake force too? Ten years of math later, that hunch became general relativity. The lesson: when a pattern looks too perfect to be a coincidence, ask what kind of process would guarantee it. Most breakthroughs aren’t hidden in exotic places — they’re sitting inside boring facts everyone stopped questioning.
Gravity isn’t a force pulling you down. It’s the floor pushing you off course. If gravity is a fake force, then the people who feel no gravity — astronauts, anything in free fall — are the ones moving naturally, in straight lines. And you, sitting in a chair, are the one being accelerated: the floor is shoving you off your natural path every second. What you feel as “weight” isn’t gravity pulling down. It’s the ground pushing up.
But if a thrown ball’s arc and the Moon’s orbit are “straight lines,” then straightness itself must be redefined — which only works if space and time are curved. Think of a flight map: the route from San Francisco to London looks like a pointless detour over Greenland, but it’s actually the shortest path. The map lies because it flattens a curved surface. Same deal: mass curves spacetime, and “gravity” is the error you get when you assume it’s flat.
Two people can watch the same event, see incompatible things, and both be right. Watch a friend fall into a black hole and you’ll see them slow down, turn red, and freeze at the edge — you never see them cross. Their experience: nothing. Clock ticking normally, no jolt, no alarm. They sail across the point of no return without any way to detect the moment it happened. Both stories are exactly correct. There is no “what really happened” — only what happened from where you stood. This isn’t philosophy — it’s engineering. Time genuinely runs slower deeper in a gravity well, and GPS satellites must correct for it daily or your maps would drift by kilometers. The darker takeaway is that the event horizon — the most consequential threshold in the universe is completely undetectable as you cross it. In a large enough black hole, you could live comfortably for decades on the wrong side of it. Doomed, healthy, and unaware. The most important points of no return don’t feel like anything from the inside, and “nothing feels wrong” is not evidence you haven’t already crossed one.
Einstein’s method — pure thinking, almost no data — is now the central question of the AI era: General relativity was built from basically two facts (light has a fixed speed; the two masses are equal). No labs, no expensive instruments. The modern question, which Brown works on at Google DeepMind: can millions of AI “Einsteins” repeat the trick? His answer: only in domains where logical consistency is brutally constraining — where few possible theories survive. In messier fields, no amount of thinking replaces the experiment that reveals which possibility nature actually picked. Know which type of problem you’re facing before betting on thinking over measuring. But AI has one real edge, and it isn’t intelligence — it’s the absence of embarrassment. A famous math conjecture went undisproven for decades, likely because everyone assumed it was true and nobody would waste career capital attacking it. An AI, not knowing what it was “supposed” to believe, checked — and disproved it. Einstein’s superpower was refusing to shrug at one coincidence. AI’s superpower is that it never learned to shrug at all.
Curious Thinking: This conversation reminded me of the Arthur Schopenhauer quote: "Talent hits a target no one else can hit; Genius hits a target no one else can see."