Polylog
← Trends

AI Masters Formal, Verified Reasoning

AI systems increasingly master formal, machine-checkable reasoning — writing verified proofs in assistants like Lean and turning that rigor toward finding real software bugs — with each release pushing automated theorem proving from research benchmark toward practical verification tooling.

weakening · confidence 44 · Emerging (watchlist) · tracking since July 4, 2026 · updated July 7, 2026

Score history

Daily conviction score, 0 to 100. Higher means the thesis is more strongly corroborated.

Jul 6 · 45Jul 7 · 44

Now 44 · -1 since Jul 6 · ranged 44 to 45

Showing the last few days. Unlock full score history.

Why the conviction moved

  • Jul 5
    Strengthened +6

    Mistral updated Leanstral to v1.5, targeting machine-checked proofs in Lean 4 where a compiler, not a benchmark, verifies correctness.

  • Jul 4
    Strengthened

    Mistral open-sourced Apache-licensed Leanstral 1.5, a Lean 4 proof model that solved 587 of 672 PutnamBench problems and reported five previously unknown bugs across real code repositories.

Source trail

  • Supporting · July 5, 2026

    Mistral Updates Leanstral, Its Model for Machine-Checked Proofs in Lean 4

    Mistral updated Leanstral to v1.5, targeting machine-checked proofs in Lean 4 where a compiler, not a benchmark, verifies correctness.

    AI ML Big Data
  • Supporting · July 4, 2026

    Mistral Open-Sources Leanstral 1.5, Saturating a Formal-Math Benchmark

    Mistral open-sourced Apache-licensed Leanstral 1.5, a Lean 4 proof model that solved 587 of 672 PutnamBench problems and reported five previously unknown bugs across real code repositories.

    AI ML Big Data (Telegram)

Unlock full source trail, score history, and daily updates.

Unlock Trends

Affected regions & assets

RegionsGlobal