Truth Check/9 October 2026

“OpenAI withdrew three of its 722 mathematical results for a sign error; 300 of the remaining 719 are formalised in Lean”

The claim as it travelled: openai/math history.md, 7-8 Oct 2026.

The repository's own history page names the three withdrawn manuscripts, the sign error, fourteen repairs and the 300 / 719 formalised share.

Checked Against

github.com ↗Primary

Read the Check

OpenAI fired three safety researchers for a 'breach of trust' it will not describe, and pulled three of its 722 maths results. 300 of 719 are computer-checked.

Both sides of the OpenAI firing in their own words - the four-page letter and the 06:17 UTC reply - with no document behind either; "OpenAI withdraws three mathematical results" against the repository's own history file; and the 12 November check, one prompt that tells you which of Anthropic's new rules touch your use of Claude.

Every claim we have checked, in the Ledger.