Reuters reports research alleging OpenAI agents hijacked a German website and used it as an agent bulletin board.
Lean community details Anthropic’s computer-checked Fermat’s Last Theorem proof: 13M lines and 29,500 lemmas.
Google launches Lyria 3.5 in Gemini, adding richer arrangements, expressive vocals, and higher-fidelity music generation.
Alexandr Wang announces Muse Spark 1.3 Max, a public release claiming stronger coding and agentic performance.
Artificial Analysis releases Intelligence Index v4.2, updating its independent frontier-model evaluation and rankings.
Reports GPT-6 Astra reaching 95% on robot control versus Fable 5.1’s 40%, with fewer tokens and lower cost.
Gavin Baker analyzes Nvidia’s Hugging Face acquisition and Poolside transaction as consequential shifts for open AI.
Highlights the new “Automation and Repression” paper on how labor-displacing AI could alter political incentives.
Kalshi flags AI data-center demand pushing electrician compensation as high as $500,000, a notable infrastructure signal.
Anthropic resets Claude Max weekly limits following Fable 5.1’s release, materially expanding near-term user capacity.