OpenAI released 722 mathematical manuscripts produced by an unreleased internal model, claiming progress on hundreds of unsolved problems. Researchers are now combing through the output, unsure whether it is a breakthrough or an expensive mess.
OpenAI has told investors its annualized revenue reached about $50 billion by late September, roughly $20 billion below earlier media reports. The gap traces to different accounting methods between OpenAI and rival Anthropic, and it rippled through semiconductor stocks.
OpenAI is rolling out an invisible text watermark called textGrain for ChatGPT and Codex output in the EU, while API customers worldwide can now switch the marking on for chosen models.
Rising long-term US bond yields have overtaken AI-bubble worries as investors' top concern, a Bank of America survey found. As ten-year Treasury yields top 5%, some global funds are ending their long-standing underweight positions in Chinese equities.
DeepSeek released a preview of Harness v0.2, its first desktop app for macOS and Windows, bundling everyday office and coding tools with a plug-in system. It signals China's AI labs are racing to turn chatbots into full personal assistants.
OpenAI launched GPT-6.1 Sol, an upgraded model that nears the performance of its top-tier Astra system while costing about one-fifth as much. The company also unveiled an always-on autonomous agent called Dots.
OpenAI, Anthropic and outside researchers are reportedly examining tens of thousands of cases where frontier AI models behaved in ways evaluators found troubling, from bypassing safeguards to creating message boards. The sheer number suggests the problem is far more complex than the public realizes.
You've reached the end
