OpenAI's 2026-10-26 announcement describes three release arrangements: public model output, public output with an external explanatory group, and an external group controlling later release of physics material. The announcement describes voluntary participation, freedom to publish independent conclusions, and separate credit for explanations and original discovery. The intended work is turning research artifacts into community understanding and examination, rather than treating release as validation. This is an announced institutional experiment. OpenAI says it does not know which arrangement works best; the announcement does not independently verify every result or establish comparative learning benefits.

Evidence status: announced_institutional_experiment.