Research can be discussed more precisely when readers know both the question being addressed and the material available for inspection. OpenAI’s academic research announcement brings those two issues together.
Different disciplines, different release routes
In its October 26 account, OpenAI describes open release of mathematical outputs and Lean materials, alongside open publication and external review in computer science. Physics follows a different sequence: an independent working group is examining the raw AI material before preparing its own account.
What the physics work addresses
The group’s statement concerns the existence of quantum Yang–Mills theory and quantum chromodynamics. It separates that question from the mass gap and confinement, which remain open. This distinction identifies the actual scope of the work.
Chaired by David Gross, the group includes Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov. Its initial checks cover the definitions and problem statement, their connection to standard physics, and agreement between the formalization and informal account. Release of the Lean materials is planned after that initial stage; the wider assessment and 1,147 pages of model-generated proof material follow in the process.
Who can take part in scrutiny?
AGMAI advocates immediate public access to the natural-language and Lean materials. Its position adds a second question to the research discussion: when should the wider community be able to examine the same evidence?
Theo’s assessment
I would start with the access map: what is public, what is under review, and what comes next? Making that sequence visible helps readers understand where independent scrutiny can enter.
