Lean, an open‑source theorem prover created by Leo de Moura at Microsoft in 2013, is based on the Calculus of Inductive Constructions (CIC) and supports both programming and mathematical reasoning in a single language. Its central library, mathlib, now contains nearly 300 000 theorems, over 100 000 definitions and about 2.5 million lines of code contributed by more than 700 developers. Autoformalization using large language models became practical in 2026: in September 2025 Math Inc. produced a quasi‑autoformalization of the prime number theorem; in January 2026 J. Urban released 130 k lines of formal topology from Munkres; in March 2026 Math Inc. autoformalized the sphere‑packing problem in 24 dimensions, yielding ~500 k SLOC that were later reduced to ~200 k lines; in May 2026 Meta/Facebook’s ATLAS project autoformalized large parts of 26 mathematics textbooks; on September 4 2026 Anthropic announced an autoformalization of Fermat’s Last Theorem generating 13 million Lean lines in 11 days, and on September 8 2026 OpenAI did the same for the forced Navier‑Stokes blowup theorem. A series of soundness bugs in the Lean kernel were uncovered in the summer of 2026, including one that produced an illicit disproof of the Collatz conjecture; these were identified by AI‑assisted security researchers such as Ramana Kumar and Dan Selsam and promptly patched. Verification efforts include Joachim Breitner’s Con‑Leche kernel, a Lean‑implemented checker with a formal consistency proof (assuming ZF plus inaccessible cardinals) that has been checked by over a dozen proof‑checkers and has validated mathlib.