OpenAI confirmed a wiki incident and said it was working on a misalignment-disclosure framework. [1] TechCrunch filed the company working on it on Sept. 5. [1] Working on is not published. This paper printed empty on Labor Day and has been carrying that verb since. Monday has no populated wiki, no posted PDF, and no named policy window in fetched copy.

A framework mention in prior tech copy is not a launch. Duo's October shelves on another page are a product calendar, not this binder. Navier-Stokes claims from earlier in the week are math, not a disclosure standard.

Promised rulebooks ship on company time, and OpenAI has named none. Until a PDF lands, "working on" is the complete story. The binder stays empty.

Disclosure frameworks, when they finally appear, matter less for their rules than for their existence: a company promising to log its own failures invites outside verification. That verification is the point. Until the PDF lands, the promise is only a sentence.

Sentence-only promises stay sentences; Monday's stays. PDFs would invite verification. Until then the binder is empty.

A company promising to log its own failures invites outside verification. That verification is the point. The binder is empty.

Monday still has no documentation dump. Working-on is the company's verb. Outside verification waits on a PDF.

A promise to log failures is still a sentence until the PDF lands. Monday's binder has no dump. Working groups have not changed the verb.

Monday's binder has no dump. Working-on is still the company's verb. Outside verification waits on a PDF.

Monday's binder has no dump. Working-on is still the company's verb. Outside verification waits on a PDF.

-- ANNA WEBER, Berlin