DeepRSI benchmark gives AI agents 48 hours to improve real ML research tasks
Hillclimb's 15 tasks are scored by private held-out verifiers; Claude Fable 5.1 leads at 1548 Elo ahead of GPT-6 Astra at 1232.
- Nasdaq slides 2.9% as Intel has its worst day since the 1970s and chip stocks fall
- S&P revises Meta's outlook to negative over AI capital spending and negative free cash flow
- Intel shares fall 29%, their worst day since the 1970s, cutting about $12 billion from the US government's stake
- Mistral releases open weights for Mistral Large 4, its 1-trillion-parameter model
- Google's Suncatcher satellite returns the first AI response from a data-centre TPU in orbit
- FDA grants accelerated approval to Inovio's Voceryq, the first US plasmid DNA immunotherapy, for adult RRP
- Anthropic renames Claude Max 20x to Ultra and drops the 5-hour limit for verified Ultra accounts
- OpenAI reorganizes its Model Spec around five core values