Topic
Use of AI and machine learning in scientific research, from analyzing data and literature to designing and interpreting experiments.
OpenAI announced an agent-generated Navier-Stokes solution a day after Tristan Buckmaster and Levent Alpöge shared a year of work, Nature reported. No independent check shows OpenAI or Anthropic misused data, so researchers' practical exposure depends on account settings and on who publishes first.
Reality
Chalmers University of Technology researchers report an AI system that designs and interprets its own brewer's yeast experiments in a closed loop. The account does not say how often its hypotheses held up, the figure a lab head needs before planning around one.
Reality
Anthropic says about 950 Claude agents found an enzyme system reminiscent of Crispr in 21.5 hours, backed so far by one unreviewed lab experiment. Scientists credit the speed of the search, so a lab can budget for faster triage today while the bench works out what the enzyme does.
Reality
Oriol Vinyals has left Google DeepMind to build Discovery Loop with Jeff Dean, Sanjay Ghemawat and Quoc Le, on the argument that AI already writes the code and runs the experiments while idea generation and judging results lag.
Reality
A Nature review recasts experimental design as a search over hardware configurations, which moves the hard part onto simulator fidelity and how the goal gets written down, and leaves buildability as its own separate question.
Reality
Physical Superintelligence launches with $58 million, though the number that matters is the $15 million probe now riding on unreviewed physics its software produced in three days of mostly autonomous work.
Reality
A model trained at NTNU did work its authors estimate would take a person 40,000 hours. The output: flowering times have moved 2.5 days a decade, and the tropics moved most.
Reality