AI agents just nominated two materials for a job real labs have been chasing for decades: magnetic semiconductor that works at room temperature.
Vals AI, a company known for building benchmarks for AI systems, published a blog post describing how a set of AI agents combed through materials science data and surfaced two semiconductor candidates that may hold onto their magnetism at room temperature rather than only near absolute zero. The post credits the work to AI agents built on Anthropic's Claude models, though it does not specify which exact version did the screening. It went up on October 5, 2026, and picked up 41 points and 25 comments on Hacker News. The post does not name the two candidate materials or publish the underlying screening data, so there is no way yet to independently check the claim.
Room-temperature magnetic semiconductors matter because today's versions need expensive cryogenic cooling to function, which has kept spintronics and some quantum-device designs stuck in research labs. If these candidates hold up, that is a materials science result, not an AI one, regardless of what automated the search. The more durable story here is the method: using AI agents to triage huge numbers of candidate compounds before anyone touches a lab bench, which is quickly becoming standard practice in materials research.
A screening result is not a confirmed material until someone synthesizes it and measures it, and a 41-point Hacker News thread is not peer review.