The Programming Language That Referees Mathematics — Leo de Moura @MachineLearningStreetTalk
The Programming Language That Referees Mathematics — Leo de Moura  @MachineLearningStreetTalk
Uploaded September 2026 | Updated November 2026, 1 day ago
Leonardo de Moura created Lean and co-created Z3.

---
This episode is sponsored by Parallel.
Parallel, where agents find answers: web search, extraction and deep research APIs built for AI agents.
Start free with the Parallel MCP server and $5 of credits every month: parallel.ai/mlst?utm_source=creator&utm_medium=podcast&utm_content=MLST
---

Tim Scarfe talks with Leo about how Lean escaped its original audience, why dependent types and Mathlib made it useful to working mathematicians, and what happens when formal verification leaves the lab. De Moura explains the small trusted kernel and independent checkers, and gives his account of the recent Collatz incident, in which a purported proof was accepted by both Lean's official kernel and Nanoda, apparently by exploiting a different bug in each.

The conversation then turns to AI-produced proofs and software. Claude agents rebuild zlib in Lean and the result is verified, yet the example exposes the loophole at the centre of the show: a proof only certifies the specification humans chose to write. De Moura describes proof search as a game, separates formal competence from mathematical understanding, and asks what authorship means when models can generate proofs at scale.

The closing argument puts human responsibility upstream: choosing definitions, judging abstractions, curating Mathlib, inspecting certificates, and deciding which problems deserve effort. He also reflects on Lean's future beyond its founder and gives a practical starting point for new users.

---
TIMESTAMPS:
00:00:00 Cold open: the green checkmark can lie
00:00:59 Cathedral or bazaar: who controls Lean's core?
00:04:55 Why Lean's core stays small and protected
00:08:07 The Slack purge, Brandolini's law and the Lean FRO
00:11:12 The Collatz exploit: two kernels, two bugs
00:16:44 More kernels, reward hacking and safety by transparency
00:21:04 Sponsor: Parallel
00:21:59 Kim Morrison, Claude and the zlib proof
00:25:25 Can we specify complex systems?
00:28:04 Specs change: proofs are cheaper to redo with AI
00:31:27 From Lean 1 to Lean 4
00:34:57 Dependent types in plain terms
00:37:25 Lean 4's extensibility and Mathlib's growth
00:41:44 Mathlib as infrastructure: Formal Frontiers
00:44:15 Creativity, abstraction and nut-sniping
00:48:46 Breadcrumbs, not learning: what AI agents lack
00:53:03 Competence without comprehension, and verified guardrails
00:56:21 Is the human still the author?
01:01:44 AlphaProof, LLMs and why certificates still matter
01:06:38 What's next for Lean, and its legacy
01:11:42 How to start learning Lean

---
REFERENCES:
tool:
[00:00:48] Lean
lean-lang.org
[00:01:24] Mathlib
github.com/leanprover-community/mathlib4
[00:12:09] nanoda_lib
github.com/ammkrn/nanoda_lib
[00:12:19] CollatzLean
github.com/xrchz/CollatzLean/blob/a79357462a33d2a6babd4cf6c8d8bcd25425d653/README.md
[00:13:06] Lean issue 14576
github.com/leanprover/lean4/issues/14576
[00:13:21] Lean pull request 14577
github.com/leanprover/lean4/pull/14577
[00:13:45] nanoda_lib pull request 22
github.com/ammkrn/nanoda_lib/pull/22
[00:15:33] Lean comparator
github.com/leanprover/comparator
[00:18:02] Lean4Lean
github.com/digama0/lean4lean
[00:19:50] ARC-AGI-3
arcprize.org/arc-agi/3
[00:21:59] lean-zip
github.com/kim-em/lean-zip
[00:29:45] CompCert
compcert.org
[00:29:45] seL4
sel4.org
[00:30:23] Z3
github.com/Z3Prover/z3
[00:38:10] Veil
github.com/verse-lab/veil
[00:38:10] Velvet
github.com/verse-lab/velvet
organization:
[00:00:52] Lean FRO
lean-lang.org/fro
[00:42:39] Mathlib Initiative
mathlib-initiative.org/about
[01:11:42] Computerphile
youtube.com/@Computerphile
person:
[00:05:11] Ilya Sergey
ilyasergey.net
[00:10:31] Joachim Breitner
joachim-breitner.de
[00:32:58] Adam Chlipala
adam.chlipala.net
[00:38:42] Kevin Buzzard
ma.imperial.ac.uk/~buzzard
[01:10:16] Terence Tao
terrytao.wordpress.com
book:
[00:08:19] The Proof in the Code
us.macmillan.com/books/9780374620059/theproofinthecode
other:
[00:10:05] Brandolini's law
en.wikipedia.org/wiki/Brandolini%27s_law
[00:45:33] A new result on unit distances
openai.com/index/model-disproves-discrete-geometry-conjecture
[01:01:08] Fermat's Last Theorem formalisation
imperialcollegelondon.github.io/FLT
paper:
[00:34:36] The Lean 4 theorem prover and programming language
doi.org/10.1007/978-3-030-79876-5_37
[01:02:06] AlphaProof
doi.org/10.1038/s41586-025-09833-y
[01:04:24] AlphaZero
arxiv.org/abs/1712.01815

---
RESCRIPT:
app.rescript.info/share/7d3d4a0059443236a01f6c9acbf4db58
app.rescript.info/api/public/sessions/b007264c0ce89047/pdf
The Programming Language That Referees Mathematics — Leo de MouraWhen It Doesnt Matter That Its Slop (Paul Lessard)A CERN like effort for AI (Jakob Foerster)Strange Geometric Shapes Found Inside AIs — Tom McGrathAI Agents can write 10,000 lines of hacking code in seconds [Dr. Ilia Shumailov]Is the Mind More Than Just the Brain? - Tom FroeseThe Weird ChatGPT Hack That Leaked Training Data [Dr. Yannic Kilcher / Prof. Florian Tramer]AI training data will never be fully synthetic [SPONSORED]Why Voice AI Still Needs a Screen | Pavan MuddireddyA Physicist Found the Hidden Phase Transitions in Society — Cristopher MooreThe Dangerous Illusion of AI Coding? - Jeremy HowardWe Built Calculators Because Were STUPID! [Prof. David Krakauer]
Machine Learning Street Talk |

The Programming Language That Referees Mathematics — Leo de Moura

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER