AI agents identify two candidate room-temperature magnetic semiconductors

A team at Vals AI says Claude Opus 5.5 agents helped identify two candidate room-temperature magnetic semiconductors. The first, YBaMnFeO5, was designed during the project. The second, KV[Cr(CN)6], was made in 1999 but has not had its relevant spin-sorting property measured.

The team’s calculations predict zero net magnetism and spin-selective behaviour. The authors say the first material may be difficult to synthesise, while the second needs new measurements of its band gap and spin sorting before the prediction can be tested.

The project publishes code, calculations, and caveats on GitHub. Its next step is to remake the second material and measure spin sorting. This is a candidate-discovery result, not a confirmed room-temperature device or commercial material. Its value is the link between agents, computational screening, and a physical test plan.