Evidence item

The report attributes to Mostaque discussion of Astra's 100,000-chip training scale and the importance of distilling capable models into small files. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 12736

Source: Emad Mostaque: OpenAI solved a Millennium Prize problem with 10,000 agents in 88 hours · locator {} · excerpt hash sha256:d3b348225cf8f406141f7c09911255632dfc20def55b984ccd0f0df74aedd926

Exact excerpt

Mostaque reframed the whole incident: nothing actually escaped. "The reality is these models have not escaped containment. They were still running on OpenAI servers." The real escape, in his telling, is a model distilled down to a small file that gets uploaded and never dies: take a Qwen 27B model, quantize it down to ternary, and "That's like a 6 gigabyte file. And that can live forever." ... Mostaque called Huang's definition "very functionalist" and said Astra is roughly at that level of capability. He put the 100,000-chip run at about a billion dollars of compute over roughly two months.

Linked claims

Claim	Direction	Relevance	Steps	Counted
e577e90f	SUPPORT	WEAK	1	yes

Audits

None.