My best guess as to what OpenAI did: a reconstruction of the method behind OpenAI's Yang-Mills existence proof

OpenAI announced on October 26, 2026 a rigorous existence proof for four-dimensional Yang-Mills theory, QCD and a unified completion of the Standard Model, without releasing it. This working paper reconstructs the likely method from public statements, OpenAI's public two-dimensional O(n) mass-gap papers, and the Balaban constructive renormalization-group literature, with independent numerical checks and a list of predictions that the released proof can confirm or refute. It is a guess, not OpenAI's work. Version 1.2 adds the author's conjecture that an infinite-volume limit is a finite map standing in for infinity, and an addendum making the corrections a standard reading of the review committee's statement calls for. Version 1.3 sharpens the conjecture (potential infinity against the axiom of infinity), separates mathematical infinity from physical space, and gives the author's reading of why the mass gap did not follow from the proof.