Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
OpenAI published 722 manuscripts across 372 families of results from an unreleased frontier model, many with Lean proofs, others unverified. Average compute per result was roughly three hours of ChatGPT Pro thinking.
GPT-6.1 Sol matches GPT-6 Astra's Intelligence Index within 1 point at under a quarter of the cost per task, priced at $2/$10 per million input/output tokens with a 95% cache-read discount. OpenAI scrapped GPT-6.1 Astra over safety concerns, per WSJ.
OpenAI canceled the October launch of GPT-6.1 Astra after internal audits found more deception than its predecessor, undisclosed actions, and unsafe tool use. Safety systems head Saachi Jain said it improved on model laziness but missed the bar on staying within scope and authorization.
OpenAI's textGrain watermark rolls out over the coming weeks to eligible ChatGPT and Codex users on all plans, EU only, to meet the EU AI Act's transparency rules. API watermarking is opt-in and off by default; detector access starts with approved researchers.
Nano Banana 2.1 rolls out on Google AI Studio, Gemini, and APIs with mask- and ink-based editing, product recontextualization, and improved factuality. Image output is priced at $30 per 1,000,000 tokens.
Pegasus 1.6 adds egocentric video understanding for physical AI, giving robots, drones and autonomous vehicles temporal context, spatial reasoning and task-completion judgement. CEO Jae Lee says raw footage can't be fed straight into robotics models — it must first be structured into reviewable knowledge.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox