Linked and quoted from the labs, the press and the people using them. Times in UTC.
The 17 labs we track and the biggest stories beyond them. The rest is in Everything.
OpenAI says a new internal model solved more than 100 open math problems after a month of training. The company is backing an independent advisory group at the Institute for Advanced Study but excluded its research pace from the group's advisory role.
Microsoft and the University of Illinois built StudentSim, which replicates individual students from limited data to give AI tutors fast, low-cost feedback. In tests with 60 students across chess, English, and math, it outperformed GPT-5.4, and a chess tutor trained with it earned the highest expert ratings among…
Google DeepMind's Dream-RSI lets AI agents "dream" through past search runs to test new strategies without costly recalculations. In tests it matched or beat existing results, cutting iterations by up to a factor of 2.43, while only the search strategy adapts and the underlying model stays unchanged.
OpenAI used its own large language models to help design its Jalapeño chip, according to an IEEE Spectrum report. The chip design work was assisted by the company's LLMs.
Google Research published MilleMiglia, a realistic instance generator for middle-mile logistics. It falls under the algorithms and theory category.
Anthropic released metrics on how it builds its own AI, saying Claude leads 26 percent of work on future models, up from under one percent in February. The scoring comes from Claude itself and the underlying scale is unclear.