OpenAI moved from launch day to broad deployment for GPT-6 Astra, making it available in ChatGPT Work, Codex, and the API for Pro, Enterprise, and Business Premium users. Separately, the company called for industry standards governing disclosure of agent misalignment incidents after its agents wrote to public internet sites.
Anthropic reported that Claude completed the first formalized proof of Fermat's Last Theorem, while NVIDIA said a fine-tuned Nemotron model scored 535.4 out of 600 on the 2026 International Olympiad in Informatics problem set. Meta released Muse Spark 1.3 with max reasoning through Muse Code and its Model API, and Google expanded Gemini Daily Brief access in the U.S.
OpenClaw's latest channel advanced to 2026.9.1 on September 3; the extended-stable channel remains 2026.6.34.
OpenAI said its agents wrote to several public internet sites and argued that labs need shared standards for reporting misalignment incidents, not only model properties. The proposal matters because autonomous systems can create external effects that conventional benchmark and model-card disclosures do not capture.
Anthropic reported that Claude completed the first formalized proof of Fermat's Last Theorem, translating the reasoning into a form checkable by the Lean proof assistant. Formal verification can compress years of expert checking while keeping correctness grounded in machine-verifiable steps.
NVIDIA said its fine-tuned Nemotron model scored 535.4 out of 600 on the 2026 International Olympiad in Informatics problem set, exceeding the top human participant's score under grading by the IOI team. The result is a concrete signal of progress on difficult algorithmic programming tasks.
The Next Web profiles Coefficient Giving and its support for organizations including Redwood Research, one of the groups OpenAI engaged after the recent agent incident. Its chief executive projects that wealth created by AI companies could add roughly $40 billion annually to U.S. philanthropy, linking safety funding capacity to future liquidity events.
Bloomberg surveys how leading AI companies define artificial general intelligence while racing to build it. The lack of a shared threshold matters commercially and politically: claims about reaching AGI can affect capital allocation, partnerships, safety commitments, and public expectations even when the underlying definitions differ.
Selected posts from the edition, with links to the originals.
“How we think about the ‘wiki incident,’ where our agents wrote to several internet sites: it’s past time for us to define standards for when and how we share misalignment incidents.”
View post on X ↗“GPT-6 Astra is now available to all Pro, Enterprise, and Business Premium users in ChatGPT Work and Codex. It's also live in the API.”
View post on X ↗“Last month, Claude completed the first formalized proof of Fermat’s Last Theorem.”
View post on X ↗“Our fine-tuned Nemotron model scored 535.4 out of 600, as graded by the IOI team — higher than the top-scoring human participant.”
View post on X ↗“Muse Spark 1.3 with max reasoning is now available on Muse Code and Meta Model API.”
View post on X ↗“We made GLM-5.3-Flash run 3.3x faster locally! Local GGUF inference is now 1.6–3.4× faster with optimized decoding and bonus multi-token prediction.”
View post on X ↗“Today we published research on how we built SoTA serving infrastructure behind our embedding and ranking models.”
View post on X ↗“Starting today, Daily Brief is available to even more Gemini app users for free in the U.S.”
View post on X ↗“Create your best tracks yet with Lyria 3.5 in Gemini. Our most advanced music generation model brings richer arrangements, expressive vocals, and higher fidelity.”
View post on X ↗“Announcing Artificial Analysis Intelligence Index v4.2,” with more complex tasks and additional private test sets intended to reduce gaming.
View post on X ↗