What was already known
Earlier work gave regularity at density-one points and smooth rectifiability almost everywhere, but did not show that the entire nearby support forms one minimal submanifold. Brena, Decio and De Lellis posed the measure-zero singular-set statement as a conjecture.
AI involvement
OpenAI is the credited author and says its internal model produced the manuscript family. It does not give a session-level account of human input or identify a human checker for this proof.
The dated manuscript is authored by OpenAI and the release says its internal model produced the mathematical manuscripts; no individual workflow record establishes the stricter autonomy level.
Importance
The theorem settles that almost-everywhere regularity conjecture for all stationary integral varifolds in Euclidean space.
Limitations
The claim assumes no stability, minimizing property, orientability, density cap or codimension-one restriction. No completed check of this proof is documented. A later companion claims a stronger Hausdorff-dimension bound and uses estimates from this manuscript; it is a separate result.
Technical verification
The September manuscript presents a lengthy proof, but neither it nor the cited release documents a completed human check, matching Lean theorem, or exact certificate for Theorem 1.1. The October companion imports its estimates and is not an independent check of the original argument.
Human understanding
No named human explanation of the signed-excess, fitting, and inward-frequency mechanism was located. Few is the missing-evidence default, not an assertion that the proof is unintelligible.
Correction and publication history
If correct, the theorem settles that almost-everywhere regularity conjecture for all stationary integral varifolds in Euclidean space.
Read this published revision