Axiom formalizes entire Stacks Project in Lean

This is a compelling counterexample to the idea that AI-generated code at scale is inherently slop: formal verification makes correctness mechanically testable, while the real challenge becomes faithful translation and long-term maintainability.

• –The project converts thousands of definitions, lemmas, and theorems into a reusable Lean dependency graph rather than a flat code dump.
• –Lean can catch invalid proofs and broken dependencies, but humans still need to verify that formalized statements accurately reflect the original mathematics.
• –A public, machine-checkable algebraic-geometry library could become powerful infrastructure for theorem search, automated proving, and future AI agents.
• –The important benchmark is not line count; it is whether external mathematicians can build on, audit, and extend the resulting library.