Disclosure index

How much of what each organisation has announced can be inspected. A result is identified when the announcement names the problem it concerns, written up when a paper or proof is public, and checkable when an artifact anyone can rebuild is published. The counts come from the disclosure ledger, as of 30 Sep 2026, and each announcement has a first-hour kit with the detail.

By organisation

How disclosure is graded

Anthropic

2 announcements·2 results claimed·2 identified·2 written up·2 checkable

100% checkable

OpenAI

3 announcements·112 results claimed·12 identified·12 written up·12 checkable

100 of the 112 results claimed have not been identified.

11% checkable

Tristan Buckmaster and Levent Alpöge

1 announcement·4 results claimed·4 identified·3 written up·0 checkable

0% checkable

Anthropic

2

Formalizing Fermat's Last Theorem

4 Sep 2026·a complete formalisation of Fermat's Last Theorem

D31 of 1 identified1 checkable

Learning more about Claude's mathematical capabilities

10 Aug 2026·a higher proven proportion of zeta zeros on the critical line

D31 of 1 identified1 checkable

OpenAI

3

Advisory Group on Mathematics and Artificial Intelligence

21 Sep 2026·more than 100 long-standing open problems across most areas of mathematics

D00 of 100 identified0 checkable

On the Navier–Stokes Millennium Prize Problem

8 Sep 2026·Navier–Stokes blowup with smooth forcing, and unforced Euler blowup

D32 of 2 identified2 checkable

Ten advances on open problems

1 Aug 2026·ten advances on open problems in mathematics and theoretical computer science

D310 of 10 identified10 checkable

Tristan Buckmaster and Levent Alpöge

1

Finite-time blowup with smooth forcing for porous media, Boussinesq and Euler

7 Sep 2026·three blowup results with smooth forcing, and blowup for hypodissipative Navier–Stokes held back until its Lean verification finishes

D14 of 4 identified0 checkable