This episode features a deep conversation with a mathematician (who goes by David) who left a tenured, permanent research-only position at CNRS to write and think publicly about the foundations of mathematics. The discussion centers on redefining what mathematics actually is, why the traditional philosophical accounts (Platonism and formalism) are broken, and how understanding mathematics as a cognitive practice can transform the way people learn and experience it.
The Definition Problem
There is no agreed-upon definition of mathematics — even Wikipedia acknowledges this. The two dominant historical views are:
Platonism: math is about abstract objects (numbers, shapes) that exist in a timeless, non-physical realm.
Formalism: math is a game of symbol manipulation — axioms, theorems, proofs — with no inherent meaning.
Both views fail when you look at what mathematicians actually do. Mathematicians spend most of their time daydreaming, building intuition, and exploring ideas that feel meaningful — not mechanically proving things or communing with a Platonic realm.
The guest proposes a different view: mathematics is a cognitive technique — a “game of truth” in which you pretend that absolutely true statements and rigorous definitions exist, and this pretense gradually reshapes your brain’s intuition. The feeling that mathematical objects “really exist” is a side effect of this process, not its foundation.
Why Formalism Has a Bug
If mathematics were purely a meaningless formal game, then introducing a false lemma into the literature should cause the whole system to collapse. But this doesn’t happen.
Real example: When Kevin Buzzard’s formalization project found a bug in a foundational lemma of crystalline cohomology (a theory used for 40+ years), he described the formal theory as having “collapsed into nothingness” — yet he knew it was fixable, because the theory had worked in practice for decades.
The concept of “fixing” a proof is meaningless in a purely formal system — there is no notion of “closeness” between two meaningless strings of symbols. The fact that bugs are almost always fixable suggests that mathematicians are tracking meaning, not just syntax.
This is the core argument for conceptualism (the guest’s term): mathematical objects don’t exist in external reality, nor are they arbitrary social conventions — they are produced by human cognition. They exist in our brains as structured mental representations.
Conceptualism and the Brain
The guest draws an analogy to neural networks: just as deep learning systems develop higher-level feature detectors across layers of neurons, the brain fabricates abstractions — concepts — through layered cognitive processes.
“Methodological Platonism” (the working mathematician’s habit of talking as if objects exist) is a useful tool for building intuition, but it becomes harmful when used to tell struggling students that they’re missing a “secret connection” to magical entities.
The real task is to build intuition — to transform something magical and opaque into something transparent and obvious.
The Game of Truth
Mathematics is described as a unique mental game: you pretend you can write absolutely true statements, make perfectly rigorous definitions, and deduce new truths. This “game of truth” has a powerful effect on the brain — it consolidates intuition and creates better internal representations of the world.
There should, in principle, be a machine-learning theorem showing that playing this game with a neural system produces increasingly powerful and useful representations.
Axiomatic Selection and Intuition
How do we choose axioms? Consistency is necessary but not sufficient — there are infinitely many consistent systems. We choose axioms that are meaningful to us, that align with our cognitive sense of what matters.
The example of non-Euclidean geometry: when the parallel postulate was dropped, the result wasn’t “wrong math” — it was a new, interesting geometry. Any consistent system can be explored.
Unintuitive results (like the Banach-Tarski paradox) are not reasons to abandon axioms — they are opportunities to rewire intuition. What seems paradoxical can become obvious with practice.
Lean, Formal Verification, and the Future
The guest is pro-Lean (a proof assistant). He sees formal verification as the future of mathematics — a way to bridge the gap between human-written proofs (which always contain hand-waving) and fully rigorous derivations.
His own paper took 7 years to get accepted, with three referees giving up. A system that could mechanically verify proofs from human arguments would be transformative.
But Lean itself doesn’t eliminate the need for meaning — a mathematician using Lean is still doing mathematics, because the meaning of the theorem is what matters.
Leaving Academia
The guest had a permanent, research-only position at CNRS — essentially a dream job with no teaching obligations. He quit after proving a major theorem that resolved his imposter syndrome.
He realized his real interest wasn’t in proving theorems but in understanding how mathematics works — a topic not taken seriously as a professional pursuit. He wanted to tell this story to the general public.
He took a break, saved money, tried writing a book, then founded a MarTech startup (using machine learning on customer data) to fund his writing. The startup lasted 12 years.
Math Ability Is Not Hard-Coded
The belief that math ability is genetic is largely wrong for the vast majority of people. If you can read a complex text, you have the cognitive hardware to do mathematics.
What’s missing is a specific mental practice — invisible, internal operations that must be learned. But they can be learned at any age.
You won’t become Ramanujan or Terence Tao, but you can become 10x, 100x, or even 1,000x better than you are now.
The Pareto Distribution of Mathematical Talent
Mathematical “strength” is not normally distributed — it follows a Pareto distribution, like wealth. There are orders-of-magnitude differences between practitioners.
This arises from a capitalization process: building intuitions on top of intuitions over years of sustained practice. Small early advantages compound dramatically.
The good news: because the gap is created by compounding, even modest improvements in your rate of progress can yield enormous long-term gains.
Understanding = Finding It Obvious
The guest’s definition of truly understanding something in mathematics: finding it obvious. Not just being able to follow the proof — feeling that it could not be otherwise.
His personal example: he spent ~10 years not understanding group cohomology (the bar construction), despite being a world expert in the area. One night, he reinterpreted it through category theory (groupoid coverings, nerves, functors) — and it became trivial. Increasing the abstraction by 100x made it 100x simpler.
This is like Descartes connecting algebra and geometry via coordinates — once the cognitive bridge is built, it feels obvious to everyone. The entire human race has undergone a “spiritual elevation” that makes x-y coordinates feel natural.
The Serre Tactic — Admitting Ignorance
Jean-Pierre Serre once sat through the guest’s 90-minute seminar and then said, “I did not understand a word. You’ll have to repeat everything.” The guest realized this took extraordinary bravery — and that this bravery was part of what made Serre great.
The guest experimented with this tactic: at conferences, he told people “my brain is damaged, I have ADHD” — giving himself permission to ask as many stupid questions as needed.
This changed the social dynamic: the other person stopped serving the “tourist menu” (impressive, formal explanations) and started serving the “local menu” (how they personally make sense of the material). This is where real learning happens.
Practical advice: find peers you can be honest with. Mathematical progress often happens at the blackboard with 2-3 people, each with chalk, making sense of something together.
Ramanujan and Neuroplasticity
Ramanujan’s abilities are often treated as magical or divine. The guest argues they are better explained as the result of extreme cognitive transformation through neuroplasticity — not genetic mutation (which doesn’t produce orders-of-magnitude differences within a species).
Analogy: a violin prodigy looks like a magician, but the skill is learned through practice, not inborn talent. Ramanujan’s brain likely did unusual internal practice that produced extraordinary mathematical intuition.
Concrete evidence: the guest took a course by Xavier Viennot that proved one of Ramanujan’s formulas using purely visual, combinatorial objects (dominoes, polyominoes, trees). After the course, all 30-40 students found it obvious. This shows the formula can be accessed through normal cognitive pathways — it’s not inherently magical.
Rigor and Picture Proofs
The history of mathematics is a back-and-forth between intuition and formalization. Picture proofs (once standard) gave way to epsilon-delta rigor, which may now give way to computer-verified proofs in Lean.
But rigor is a tool, not the goal. The goal is understanding. Sometimes a one-sentence picture proof (as in one of the guest’s own published papers) communicates more than a 50-page formal derivation.
The guest’s PhD advisor’s advice: “If you’re not failing 10% of the time, you’re not trying hard enough.” Don’t self-censor. Submit the ambitious paper. Take the risk.
