OpenAI's AI agents cracked the Navier–Stokes equations, one of mathematics' great unsolved problems, in 88 hours. Tristan Buckmaster expressed concerns that his research stored in Codex could have fed into the work of OpenAI.
OpenAI stated that prompts from Buckmaster submitted to Codex during the two months before the announcement could not have influenced the system. Dalmeet Singh Chawla claimed that generative AI has basically disrupted the whole scientific process.
Google DeepMind solved how proteins in living things fold into their precise shapes in 2020. AlphaFold, a model from Google DeepMind, predicted 200 million protein folds in one year.
Dr Pushmeet Kohli said AlphaFold could be seen as a watershed moment. A DeepMind model produced more than 2 million new crystal structures, releasing nearly 800 years worth of knowledge in one go.
Of these new crystal structures, 380,000 were considered promising enough to make in a laboratory. Anthony Cheetham and Ram Seshadri found no strikingly novel compounds in a random sample of the new crystals.
Prof Dmitry Kobak searched more than 1 million biomedical research papers in 2024. Nearly 90% of the searched papers had AI fingerprints by December 2025.
Basecamp Research plans to collect DNA to create an AI that can speak the language of biology.
