Researchers at the University of Pennsylvania have demonstrated optical signal switching using exciton-polaritons—hybrid light-matter particles—at an energy cost of roughly 4 femtojoules per switch, far below conventional electronics. The platform points toward photonic AI chips that eliminate heat-related energy waste and could dramatically cut the power demands of large AI systems.
Google DeepMind said on February 11, 2026 that Gemini Deep Think is being used on professional research problems across mathematics, physics, and computer science. The company highlighted its Aletheia math agent, up to 90% on IMO-ProofBench Advanced, and collaborations on 18 research problems as evidence that AI is moving from benchmark performance toward real scientific workflow support.
The subreddit focused on the unusually candid part of Anthropic's physics case study: Claude moved work faster, but still needed expert supervision to catch fabricated checks, wrong formulas, and weak judgment.