Vals AI ran about 90 Claude Opus 5.5 agents and roughly 750 density functional theory jobs in Quantum ESPRESSO on Modal cloud CPUs, with adversarial referee reviews. The predictions put YBaMnFeO5 at a 2.35 eV band gap and magnetic up to about 420 K. KV[Cr(CN)6] comes in at 2.1 eV and 376 K. After calibrating against a known magnet, the YBaMnFeO5 figure rises to about 490 K. The designed compound is likely to scramble during synthesis. Both are predicted to have zero net magnetism yet still sort electrons by spin. One is new, and the other was first made in 1999. Memory built from such materials could switch about 1,000 times faster than ferromagnetic memory, according to the post. Nothing has been tested experimentally. The next step is to re-synthesize KV[Cr(CN)6] and measure its spin sorting. Vals published a public ledger of inputs, outputs and scripts, plus a one-command checker. Hacker News commenters focused on that missing validation. For AI founders, the notable part is the workflow: parallel agents, cheap cloud compute, and a referee layer that checks the agents' work. Vals is an evaluation company, so this reads partly as a demonstration of agent capability. The commercial value depends on whether a lab confirms the predictions.