Tag: cognitive bias

Categories
Aircraft History

The Merlin

There is a sound that men who heard it never forgot. Not the roar exactly, though it roared. Something beneath the roar — a note, almost musical, that settled into the chest and stayed there. Four Rolls-Royce Merlins at full throttle on a Lancaster climbing out of Lincolnshire in the dark, and sixty years later old men would close their eyes trying to describe it and find they couldn’t, not quite, which was itself a kind of description.

The engine was a miracle of the wrong era. Liquid-cooled, sixty degrees of vee, twenty-seven liters of displacement producing over a thousand horsepower from something you could fit in a large kitchen. Rolls-Royce had been making engines since 1906, had learned things about metallurgy and tolerance and the behavior of superheated gases under compression that couldn’t be written down, only accumulated, passed hand to hand through decades of making things that had to work when nothing could be allowed to fail. The Merlin was the distillation of all of it.

And then — this is the part that stops you — they couldn’t build enough of them.

Britain in 1940 was a country running on nerve. The factories were working. The workers were willing. But the math was brutal and the math didn’t care about willingness. So someone made a phone call to Detroit. To Packard. A company that had spent thirty years building luxury automobiles for American industrialists, cars with interiors like drawing rooms on wheels, cars that announced their owners had arrived at exactly the place they had always intended to be. Packard looked at the Merlin blueprints, converted the tolerances from imperial to metric and back again, retooled their entire production line, and started building the engine that would power the Spitfire, the Hurricane, the Lancaster, and the P-51 Mustang.

Think about what that required. Not just the engineering, though the engineering was extraordinary. The belief required. That these tolerances mattered. That this particular arrangement of pistons and supercharger vanes and coolant passages was worth the disruption of an entire industrial operation. Packard’s engineers didn’t question the design. The design had already proven itself.

You built the Merlin because the Merlin worked.

The question — the one that takes longer to arrive — is what you do when the thing the Merlin is for doesn’t.

Arthur Harris believed.

That is the first thing to understand about him, and maybe the last. He believed in the bomber the way certain people believe in a technology so new and so powerful that the believing itself feels like vision. Strategic bombing would break Germany. Not assist in breaking Germany. Not contribute to a larger effort that would break Germany. Would, by itself, through the systematic destruction of German cities and the German will to continue, end the war. Harris had held this view before the war began and he held it after the evidence came in and he held it, unmodified, until he died in 1984.

This is not stupidity. The most costly certainties never are. Harris was shrewd, forceful, organizationally gifted, genuinely courageous in the sense that he was willing to send men to die for what he believed and knew he was sending them. He understood logistics, understood morale, understood the brutal arithmetic of attrition. What he could not do — what the structure of his certainty would not permit — was update.

The evidence arrived slowly enough that you could always explain it away. German war production increased through 1943, then through 1944, even as the bombers came night after night. The factories dispersed. The workers adapted. The morale that was supposed to crack showed instead a remarkable tendency to consolidate under pressure, the way populations sometimes do when the threat comes from the sky and cannot be reasoned with. The theorists had a model of human psychology that turned out to be wrong, and the model’s wrongness kept arriving in the data, and Harris kept flying.

Fifty-five thousand men.

Picture Harris alone. The commander in the early morning after the casualty reports come in, before the day’s work begins again. The loneliness of a certainty that has become structural — no longer a belief you hold but a belief that holds you, because the alternative is not just being wrong but having been wrong, which means all those boys went down over the Ruhr for a theory, which is a weight no living person can carry and continue to function. So you don’t revise. You recommit. You ask for more aircraft, more crews, more nights.

You build more Merlins.

This is the mechanism. Not malice. Not indifference. The certainty becomes self-protective, which means it becomes invisible, which means it becomes the water you swim in rather than a position you hold. Harris stopped being a man with a theory about bombing and became a man for whom bombing was the answer to every question, including the question of whether bombing was working.

The Lancaster crews knew something was wrong before Harris did. You could see it in the casualty rates, which they could calculate as well as anyone — better, actually, because they were doing the calculating with their own lives as the variable. Forty-four percent didn’t survive their tours. They knew this. They flew anyway, because courage doesn’t require certainty about the strategic framework, only about the man beside you and the mission tonight.

The Merlin started and you went.

The Merlin outlasted the theory. It kept flying for decades after the war, in civilian aircraft, in racing planes, in the occasional restored Lancaster that still tours airshows in Britain, where crowds gather on summer afternoons to watch it pass and hear, carried on the wind, that sound. The note beneath the roar. The thing that settles in the chest.

Beautiful, people say, watching it go.

And it is. It genuinely is.

What they couldn’t know — what none of them could know — was that the engine was the most reliable thing in the entire enterprise.

Categories
AI AI: Transformers Books

The Updating Machine

Tom Chivers puts Bayes’ theorem in plain English and it sounds almost obvious: “the probability of event A, given event B, equals the probability of B given A, times the probability of A on its own, divided by the probability of B on its own.” A formula for revising what you believe when new evidence arrives. You started somewhere. Something changed. Now you believe something slightly different. Repeat.

The obvious part is the mechanics. The hard part is the loop.

Most reasoning errors I catch in myself aren’t failures of logic — they’re failures to update. I hold a view, evidence accumulates against it, and I find reasons the evidence is flawed rather than reasons the view might be.

Psychologists have a name for this: confirmation bias. But I’ve always found that label a bit too clean, like it describes a bug rather than a feature.

The prior isn’t wrong to be sticky. It represents everything you’ve learned up to this point. The problem is when it becomes load-bearing — when the prior stops being a starting position and starts being a conclusion.

“Strong opinions, loosely held” is supposed to solve this. It’s a useful phrase — it captures something true about the right posture toward your own beliefs. But in practice the second half is harder to honor than it sounds. The strong opinion gets stated, new evidence arrives, and changing your mind in public feels like losing. The “loosely held” part quietly becomes decorative.

What Bayes actually demands is something closer to epistemic humility with arithmetic attached. You don’t get to say I don’t know. You have to say I estimate 0.4, and here is what would move me to 0.6. That’s harder. It requires you to specify not just what you believe but how you’d know if you were wrong.

This is why Bayesian thinking keeps surfacing in AI conversations. Modern language models do something structurally adjacent to this — not consciously, but mechanically. Every token generated is a probability distribution revised forward by context. The model doesn’t know the next word; it updates a prior over all possible words, given everything that came before. It’s not reasoning the way humans reason, but it’s updating the way Bayes updates: continuously, contextually, without the luxury of certainty.

Whether that’s comforting or unsettling probably depends on your own prior.

The deeper thing Chivers is pointing at, I think, is that Bayesian reasoning is essentially a description of intellectual honesty as a process rather than a trait. You can’t just decide to be open-minded. You have to build the loop: form a belief, assign it a probability, watch for evidence that should move it, and then actually move it. Most of us do the first three. The fourth step is where it gets expensive.

I’ve been wrong about enough things by now that I’ve started to treat my own confident views with mild suspicion. Not paralysis — you have to act on something — but a background awareness that the prior I’m acting on was formed by a person who had less information than I do now, and less than I’ll have next year.

Strong opinions, loosely held, sounds right. The trick is meaning it.

Categories
AI Business

The Topography of a Face

I found myself staring at the physical geometry of a conversation the other day—not the words, but the topography of the faces delivering them.

Elad Gil recently shared a fascinating experiment during a conversation with Tim Ferriss. He’s been uploading photos of startup founders into AI models and asking the machines to predict if they’d be successful, purely based on their “micro-features.”

“Because if you think about it, we do this all the time when we meet people, right? We quickly try to create an assessment of that person, their personality, and what they’re like. There are all these micro-features—like, do you have crow’s feet by your eyes, which suggests that your smiles are genuine? […] So, I have this whole set of prompts that I’ve been messing around with, just for fun, around: ‘Can you extrapolate a person’s personality based off of a few images?'”

He notes the model breaks down the crow’s feet and the furrowed brows, extrapolating a personality from a static frame. It’s a parlor trick, perhaps. But it works because it holds a mirror to our oldest, most unexamined instinct.

We are all amateur phrenologists of the human face. We sit across a table, measure the crinkle of an eye or the tightness of a jaw, and we build a rapid, invisible architecture of trust or suspicion. Over decades of investing and making career choices, I’ve often leaned heavily on this silent language. I’ve backed founders because their intensity felt genuine, and I’ve passed on others because something in their posture felt misaligned.

But if I am brutally honest, that intuition has sometimes been a mask for my own blind spots. I’ve held on to failing investments for far too long because I trusted a reassuring smile. We like to think our gut instinct is a sophisticated instrument. Often, it is just a pattern-matching engine running on deeply flawed historical data.

Now, we are handing that very human habit over to a machine. We prompt the AI to become a “cold reader,” and it obliges, predicting who will be the quiet observer and who will deliver the dry wit.

The unsettling part isn’t that the machine might get it wrong. The unsettling part is that it might get it exactly right—by mimicking the very same rapid, superficial judgments we make every day, just at a terrifying scale.

We are teaching silicon to read the human code. The future will belong to those who realize the code was always written in our own biases.

Categories
Probabilities

The Fiction of Certainty

There is a profound discomfort in the space between zero and one.

In her book Spies, Lies, and Algorithms, Amy B. Zegart notes a fundamental flaw in our cognitive architecture:

“Humans are atrocious at understanding probabilities.”

It is a sharp, unsparing observation, but it is not an insult. It is an evolutionary receipt. We are atrocious at probabilities because we were designed for causality, not calculus. On the savanna, if you heard a rustle in the tall grass, you didn’t perform a Bayesian analysis to determine the statistical likelihood of a lion versus the wind. You ran. The cost of a false positive was a wasted sprint; the cost of a false negative was death.

We are the descendants of the paranoid pattern-seekers. We survived because we treated possibilities as certainties.

The Binary Trap

Today, this ancient wiring misfires. We live in a world governed by complex systems, subtle variables, and sliding scales of risk. Yet, our brains still crave the binary. We want “Safe” or “Dangerous.” We want “Guilty” or “Innocent.” We want “It will rain” or “It will be sunny.”

When a meteorologist says there is a 30% chance of rain, and it rains, we scream that they were wrong. We feel betrayed. We forget that 30% is a very real number; it means that in three out of ten parallel universes, you got wet. We just happened to occupy one of the three.

Zegart operates in the world of intelligence—a misty domain of “moderate confidence” and “low likelihood assessments.” In that world, failing to grasp probability leads to catastrophic policy failures. But in our personal lives, it leads to a different kind of failure: the inability to find peace in uncertainty.

Stories > Statistics

We tell ourselves stories to bridge the gap. We prefer a terrifying narrative with a clear cause to a benign reality based on random chance. Stories have arcs; statistics have variance. Stories have heroes and villains; probabilities only have outcomes.

To accept that we are bad at probability is an act of intellectual humility. It forces us to pause when we feel that rush of certainty. It asks us to look at the rustling grass and admit, “I don’t know what that is,” and be okay with sitting in that discomfort.

We may never be good at understanding probabilities—our biology fights against it—but we can get better at forgiving the universe for being random.

Categories
Authors Business Living

The Terror of the Empty Chair

It is comforting to believe that when the world breaks—when housing markets collapse, when “unicorn” startups vaporize, or when seasoned scouts overlook generational talent—it is because of a miscalculation. We want to believe the math was wrong, the data was bad, or the algorithm was flawed. We want to believe it was a glitch in the intellect.

I heard a commentator recently mention that Michael Lewis, the chronicler of our most expensive delusions in his best selling books, has suggested something far more unsettling. In looking at the connective tissue between The Big Short, Moneyball, and Going Infinite, he identifies a different culprit. He notes that the “glue” holding these irrational systems together isn’t incompetence. It is FOMO: The Fear Of Missing Out.

“They are more afraid of being left behind than they are of being wrong.”

This observation completely reframes the narrative of catastrophic failure. It explains why high-IQ individuals—people paid millions to be rational—consistently make decisions that look insane in retrospect. The banker, the VC, and the scout aren’t necessarily blinded by greed, though greed is certainly a passenger in the car. They are blinded by the terror of the empty chair.

Lewis points out that for the social animal, the pain of being left behind is acute and immediate, whereas the pain of being wrong is often abstract and distant. If you sit out a bubble and the bubble keeps inflating, you look like a fool today. You are isolated. You are the cynic at the party who refuses to dance. If you join the bubble and it bursts, well, you have company. As the old financial adage goes, “It is better to fail conventionally than to succeed unconventionally.”

There is a profound, empathetic tragedy in this. It suggests that our systems don’t fail because we aren’t smart enough; they fail because we are too human. We are wired for the herd. The biological imperative to stay with the group—originally a survival mechanism against predators—has been warped into a financial suicide pact.

When we look at the irrational exuberance of a market, we aren’t seeing a mathematical error. We are seeing a materialized anxiety. We are seeing a collective hallucination held together not by logic, but by the sticky, desperate glue of not wanting to be the only one who didn’t buy the ticket.

The antidote, then, isn’t just better data or faster computers. It is the emotional discipline to be lonely. It is the willingness to stand apart from the warmth of the herd and accept the short-term social cost of being “out” for the long-term reward of being right.

Categories
Living Mathematics

The Curve That Blinds Us

There is a fundamental mismatch between the hardware in our heads and the software of the modern world. We are linear creatures living in an exponential age. We can be stunned by exponential growth.

Our ancestors evolved in a world where inputs matched outputs. If you walked for a day, you covered a specific distance. If you walked for two days, you covered twice that distance. If you gathered firewood for an hour, you had a pile; for two hours, you had a bigger pile. Survival depended on the ability to predict the path of a spear or the changing of seasons—linear, predictable progressions.

But nature and technology often behave differently. They follow a curve that our intuition simply cannot map.

If a lily pad doubles in size every day and covers the entire pond on the 30th day, on which day does it cover half the pond? Our linear intuition wants to say the 15th day. But the answer, of course, is the 29th day.

For twenty-nine days, the pond looks mostly empty. The growth is happening, but it feels deceptively slow. We look at the water on day 20, or even day 25, and think, “Nothing is happening here. This is manageable.” We mistake the early flatness of an exponential curve for a lack of progress.

This is the “deception phase” of exponential growth. It is where dreams die because the results haven’t shown up yet. It is where we ignore a virus because the case numbers seem low. It is where we dismiss a new technology because the early versions are clumsy and comical.

Ernest Hemingway captured this feeling perfectly in The Sun Also Rises when a character is asked how he went bankrupt. His answer:

“Two ways. Gradually, then suddenly.”

That is the essence of the exponential. The “gradually” is the long, flat lead-up where we feel safe. The “suddenly” is the vertical wall that appears overnight.

The tragedy is not that we fail to do the math—we can all multiply by two. The tragedy is that we fail to feel the math. We judge the future by looking in the rearview mirror, projecting a straight line from yesterday into tomorrow. But when the road curves upward toward the sky, looking backward is the fastest way to crash.

To navigate this world, we must learn to distrust our gut when it says “nothing is changing.” We have to look for the compounding mechanisms beneath the surface. We have to respect the 29th day.

Categories
AI AI: Large Language Models Investing

The Digital Devil’s Advocate

There is a seduction in the handwritten note. When I scribble down a company name in a notebook, it is purely additive. It represents potential upside, a future win, a brilliant insight caught in ink. The notebook is a safe harbor for optimism because it lacks a “Reply” button. It doesn’t argue back.

But optimism is an expensive luxury in investing.

After my initial experiment—using Gemini 3 Pro to transcribe my messy list into tickers—I felt a surge of productivity. But productivity is not the same as discernment or understanding. I had a list of stocks, but I didn’t have a thesis. I just had digitized hope.

So, I took the next step. I didn’t ask the AI for validation; I asked for a fight. I fed the tickers back into the model with a specific directive: “Act as a contrarian hedge fund analyst. Find the red flags. Kill my enthusiasm.”

“I didn’t ask the AI for validation; I asked for a fight.”

The results were immediate and sobering. The “promising tech play” I had noted? The AI highlighted a massive deceleration in user growth hidden in the footnotes of their latest 10-Q. The “stable dividend payer”? It flagged a payout ratio that was mathematically unsustainable.

In seconds, the warm glow of my handwritten discovery was doused with the cold water of 10-K realities. And it was fantastic.

We often view AI as a tool for creation—generating text, images, and code. But its highest leverage application might actually be destruction. By using it to stress-test our assumptions, we outsource the emotional labor of being the “bad cop.” It allows us to kill bad ideas quickly, cheapy, and privately, before we pay the market tuition for them.

My notebook is still where the dreams live. But the digital realm is now where they go to survive the interrogation.