“Beauty is truth, truth beauty, — that is all
Ye know on earth, and all ye need to know.”
– Keats

Patrick Kidger wrote a post I read as I was getting into JAX, and even though the post was largely unrelated to what I was looking for, the writing itself is something I sometimes recall. The argument he makes is not complicated. He says the best researchers just “know stuff,” across a wide breadth of domains such as ML, scientific computing, software engineering, mathematics, statistics. Possessing a technical depth of knowledge is how new ideas get generated and bad ones get recognized. Not awareness of adjacent fields but real facility with them, enough that connections form without forcing. The people who do the best work know numerous random facts, not as trivia but as something load-bearing in how they think.

Specialization, in this framing, only works as a last mile on top of something wider. Old ideas resurface constantly with new packaging. Without that base, everything looks like it should be the next SWE-killer, some revolution. People begin to trust things at first sight just because of pedigree and proxies rather than accuracy.

I don’t believe building this base must be constrained simply to the goal of an ML PhD, or even anything in computer science. Across all of our personal aesthetics, each of our experiences drives an inherent intuition. Dirac believed a truly fundamental theory must be mathematically beautiful, and that this wasn’t aesthetic preference but a truly working heuristic. Even when experiments seemed to contradict general relativity, he argued the theory should be trusted: “Anyone who appreciates the fundamental harmony connecting the way nature runs and general mathematical principles must feel that a theory with the beauty and elegance of Einstein’s theory has to be substantially correct.”

Coat Racking

People are mostly consumptive now rather than productive. Shortform media compressed attention into something that rewards recognition over understanding. People end up as a kind of philosopher’s coat rack, collecting frameworks and mental models and recycled takes but never wearing any of them. There is no progressive expansion of ideas brought on by real world perturbations, forcing nuance.

AI, when used without any real intention, exaggerates this particular human failure mode. To some extent, every token corresponds to a decision. When a model writes on someone’s behalf, it collapses thousands of those micro-decisions into a single generation. Rhetoric directly corresponds to how various audiences receive a concept, even if the underlying meaning is the same. Syntactic structure and use of analogies can reinforce or confound a message. Paragraph structure, argument ordering, what to cut (even if the thought itself is meaningful) are each a real writer’s judgment call, and judgment atrophies without use. The people who use these tools best are the ones who have read and written thoughtfully for long enough to steer them, who can describe the rhetoric they want rather than recognize it after the fact.

There’s a tension between robustness and evolvability in the theory of human adaptation that feels relevant. Chimpanzee groups separated by a river are more genetically distinct than humans from different continents — in part because culture itself (theoretically) suppresses genetic diversity, buffering us from the selection pressures that would otherwise demand it. Clothing instead of cold tolerance, medicine instead of immune systems. Robustness consumes biological evolvability. Now the same tradeoff shows up in how people think. Confirmation bias and prestige bias are cognitive robustness strategies — deferring to what’s already worked or already credentialed. In a stable landscape, this makes sense. In a fast-moving field, the interesting bets get made before social proof exists, by people whose model is good enough to evaluate the world directly. These are early adopters. You can’t get there by being robust harder, carrying more coats. You have to make your own.

Genuine intrigue matters much more than extrinsic motivation for building a strong world model. A higher percentage of the musicians I know think this way than founders (alas, being a founder is a stylish thing now). They’ve built a working understanding of the creator economy from the inside, deconstructing how Spotify’s algorithm weights save-to-listen ratios, how YouTube surfaces new channels, which subcultures drive real engagement versus which ones just look active, how to build a visual identity that reads coherently across a 1-inch profile picture and a festival poster. They think about branding and demographics not because someone told them to but because their craft demanded it. Nobody taught them a framework for any of it. They cared enough to figure it out, and the figuring-out required pulling from places that had nothing to do with music. Their understanding feels much richer compared to anything a GSB student can tell you, with all the resources in the world. The difference isn’t information. One group built understanding against a problem they were genuinely trying to solve. The other collected it reflexively, with no pressure to cohere.

What matters is the quality of the internal model. A single way of seeing the world makes thinking brittle in ways that aren’t obvious from the inside. You can execute within a known framework but the moment it stops applying there’s nothing to fall back on.

Pathfinding

The Renaissance ideal was that a single person could understand most of what was knowable. Leon Battista Alberti wrote the first scientific study of perspective in painting, designed church facades, invented the first polyalphabetic cipher, wrote a Latin comedy so convincing it was still being published as genuine Roman text 150 years after his death. “A man can do all things if he will.” That’s not the case anymore. Danny Hillis calls it “The Entanglement” — past the Enlightenment into an era where each expert knows a piece but the whole system has become causally opaque even to the sum of its specialists. In a slower yet complex world, one might say that specializing harder would be the right response. But progress is diffuse now, spread across industries that increasingly borrow from each other. The person who can follow multiple tailwinds and cross-pollinate between them will outpace the person who can only think within one.

Brian Eno called himself a “non-musician” when he joined Roxy Music — preferring “systems manipulator.” He studied under Roy Ascott’s Groundcourse, where the reading list was more cybernetics than art history. Stafford Beer’s The Brain of the Firm taught him how systems could have integrity and personality using only rules, without specifying anything algorithmically. That idea became generative music — compositions that grow themselves from initial constraints. It also became Oblique Strategies with Peter Schmidt, a deck of cards for disrupting creative process. It became his production style on Bowie’s Berlin trilogy and Talking Heads, treating the studio as an instrument rather than a capture device. And it became the Long Now Foundation, which he co-founded and named, where he designed the chime generator for a clock meant to run ten thousand years. One deep curiosity about emergent structure, operating across music, visual art, systems thinking, and technology for fifty years. The breadth is a byproduct of the curiosity, not the other way around.

One can’t fake nor cram the base. It has to come from having genuinely cared about a lot of things over a long time. Some of the best pretraining is just experiencing things without overthinking them, being shaped by something before deciding what it’s for.

What I’ve been calling “base” is probably better described as longform ideology — the slow accretion of enough real understanding, across enough domains, that it coheres into something you’d defend. A worldview that itself is load-bearing. People like to use the word “taste” to describe what their intuition produces, but taste (both in its usage and what it describes) is a hedge. It feels like UI/UX polishing, not something someone would stake their life’s work on. It gets adopted by people who want to sound discerning without committing to anything, preference dressed up as conviction. Nobody’s taste ever got them arrested. Ideology does.

Ideology, Not Agency

Agency is underspecified. Strong habits can look productive without being agentic, where the actor may simply act under inertia. Meanwhile, a high-information diet paired with an action bias can produce someone extremely agentic who directs all of it toward a hopeless market. Another social media app without a novel perspective, another weekend project without retention dressed up as ARR. Capital without thesis, deployment without differentiation. The ability to gamble everything on Kalshi, quit a job, move across the country — but toward what?

In 1926, Mussolini’s government arrested an Italian political theorist whose writings they considered dangerous enough that the prosecutor reportedly said at trial they had to stop this brain from functioning for twenty years. Instead, Antonio Gramsci spent his last decade in a fascist prison and filled thirty-three notebooks — over three thousand pages of political theory, philosophy, cultural criticism. Sick for most of it, barely able to write. The Prison Notebooks went on to shape how people think about culture, power, and class for decades.

One of his ideas, the philosophy of praxis argued that thought without action is sterile but action without a coherent worldview is just flailing. The people who actually change things aren’t the ones with the most freedom to act but the ones whose understanding runs deep enough to know what to do with whatever freedom they have. The organic intellectual isn’t defined by credentials but by thinking that emerges from genuine engagement with the world, the ability to name what others only feel. A worldview that didn’t need freedom to propagate. Just contact with reality, which he maintained through prison walls.

Compared to the “agentic,” I’d rather bet on someone who feels trapped but desperate to do differentiated and novel things, or someone whose habits run so deep that intuition drives how they act without deliberation. Both are technically less agentic than the person who makes deliberate choices every day, optimizes every hour, ships every weekend. When you don’t have much room, you get precise about what matters quickly. And when intuition runs deep enough, perturbation from externalities is meaningful enough that finding the optimal path to an end state comes easy. Either way there is real potential.

I have a friend from high school, now at an Ivy League, who put down all his internship money to sportsbet on Kalshi. He doesn’t know anything about sports. He is already going to be in extreme debt from student loans, and doesn’t think he will be satisfied working IB or consulting (honestly, understandably). So, thinking from his shoes in this very simplified description of his life, why wouldn’t he bet it all? He and many young Americans across the world don’t believe that they have a meaningful future. If they don’t feel they have a reliable control on their life path, wouldn’t their most sensical path be to take the 1% chance to potentially be satisfied? Now, say he is not restless because his currently bad state, but rather because he finds his life so meaningless where even the standard human goals of making money, having a family, or achieving some high level of status is not enough. He truly wants to change the world, out of delusion, and has many reasons to do so — spite, a grand vision, maybe a machiavellian philanthropy. I say this person exists. I say he should take the leap, and his restlessness will find purchase.

(Before migration, birds undergo a state ornithologists call Zugunruhe — migratory restlessness. Corticosterone floods the body, organs reshape, flight muscles hypertrophy. Lorine Niedecker, paraphrasing the naturalist W.H. Hudson, states: birds feel something akin to pain just before migration, and nothing alleviates it except flight. What else would be so inevitable without their existence?)