- Vals AI claims more than 90 Claude Opus 5.5 agents running density-functional-theory simulations over three days identified two Luttinger-compensated magnet candidates for potential spintronic memory applications.
- According to the vals.ai blog post, the predicted candidates—YBaMnFeO₅ and KV[Cr(CN)₆]—carry band gaps of 2.35 eV and 2.1 eV respectively, with magnetic behavior persisting above room temperature in simulation only.
- Vals AI itself acknowledges neither material has been experimentally measured, and the company flags that YBaMnFeO₅ may be impossible to synthesize in the ordered crystal form needed for it to work.
What the Buzz Is All About
Well, hitch up the mule and sharpen your pencil, because Vals AI has gone and told the internet something that's got the tech crowd hollering like a hound dog at a full moon. According to a blog post published October 4, 2026 on vals.ai by Geby Jaff, the company deployed a swarm of more than 90 Claude Opus 5.5 agents—think a whole mess of AI worker bees—to run computational materials simulations over the course of three days. Vals AI claims those agents surfaced two candidate materials, YBaMnFeO₅ and KV[Cr(CN)₆], which the company describes as potential building blocks for next-generation spintronic computer memory.
According to the vals.ai post, the approach the company describes involved two passes of density-functional-theory simulation: a quicker PBE+U run and a slower, higher-fidelity HSE06 run. Vals AI says this multi-agent setup—which the company frames as a kind of parallel computational workforce—let them screen a wide range of material candidates faster than a one-at-a-time approach would allow. Whether that framing survives independent scrutiny is, as we'll get to, a whole different barn dance.
What Vals AI Actually Says It Found
According to the vals.ai blog post, YBaMnFeO₅ is a newly designed oxide compound that the agents predict carries a 2.35 eV band gap and retains its magnetic character up to roughly 420 K—that's above room temperature, which is the whole danged point. KV[Cr(CN)₆], a compound first put together back in 1999, is predicted by the company's simulations to have a 2.1 eV band gap and hold its magnetism up to about 376 K, also above room temperature, according to vals.ai.
The company's post also floats the idea—and this one carries an 'uncertain' label in our own assessment—that spintronic memory built from materials like these could switch roughly a thousand times faster than ferromagnetic memory and kick up less interference with neighboring components. Vals AI makes that claim, but it is entirely unverified pending any actual experimental work, so file it under 'sounds real nice if true' for now.
What Is Actually Known for Certain
Here's what we can say without crossing our fingers: Vals AI published a blog post making these claims, the post was authored by Geby Jaff, and it appeared on October 4, 2026. The Hacker News thread discussing the announcement pulled a score of 222 with around 163 comments, confirming that a good-sized crowd of technically minded folks sat up and took notice—like a cat spotting a laser pointer. Alpha Signal and ai-tldr.dev both covered the announcement, though their accounts flow entirely from the same vals.ai blog post and do not represent independent scientific corroboration.
KV[Cr(CN)₆] is a real compound with a documented synthesis history going back to 1999, so that one at least exists in the physical world. YBaMnFeO₅, by contrast, is a novel design that, as of this writing, appears to exist only in the vals.ai simulation results. No peer-reviewed journal, no independent materials-science laboratory, and no established science-news outlet has verified or replicated any of the computational findings described by vals.ai.
What Remains as Unverified as a Fishing Story
Lord have mercy, the list of unverified items here is longer than a dirt road in July. Neither the band gaps nor the spin-sorting behavior of either material has been measured in a physical laboratory, according to vals.ai's own disclosure. The company acknowledges that experimental synthesis and measurement have not been attempted for YBaMnFeO₅. As for KV[Cr(CN)₆], vals.ai states the next step is to re-synthesize that compound and run the actual tests—meaning even the older material is starting from scratch on the experimental side.
The synthesis challenge for YBaMnFeO₅ is particularly worth keeping an eye on. According to the vals.ai post, the ordered crystal arrangement that would give the material its predicted spin-sorting ability is likely to collapse into a disordered, scrambled form during the high-temperature synthesis conditions—somewhere around 900 to 1300 degrees Celsius—that standard oxide preparation requires. That's a bit like building a fence that falls apart the moment you try to nail it together. Vals AI flags this as a known hurdle, which at least earns the company some credit for not burying the lead.
Community Reaction: Impressed but Not Convinced
The Hacker News crowd, bless their skeptical hearts, showed up in force. While many commenters praised the ambition of the multi-agent computational approach, a notable portion raised pointed questions about reproducibility and whether these results would hold up once someone in a lab coat actually tried to make the stuff. That community engagement is a social signal worth noting, but it is not peer review—not by a country mile.
One digest outlet noted plainly that no human-led synthesis or measurement data accompanied the initial announcement, framing the work as a shortlist for future lab investigation rather than a confirmed discovery. That framing squares with vals.ai's own internal disclosures, which simultaneously present the predicted Curie temperatures as promising and acknowledge that at least one candidate may be practically unsynthesizable in its useful form. That tension, sitting right there in the company's own post, is the kind of thing that makes a careful reader chew on their corn bread a little longer.
Our Analysis: A Promising Shortlist, Not a Discovery
What follows is this publication's analysis, not additional reporting. The most honest way to characterize what Vals AI has shared is this: the company's agents appear to have produced a computationally generated shortlist of candidates worth investigating, not a confirmed materials discovery. Computational screening of this kind has genuine value in materials science—it narrows the haystack before anyone picks up a pitchfork—but the distance between a simulation result and a working material in a spintronic device is vast, and right now the journey hasn't started.
The multi-agent framing is interesting from an AI-capability standpoint, and if the workflow proves reproducible, it could represent a meaningful acceleration of the early-stage screening process. But that reproducibility hasn't been demonstrated publicly either. Until an independent lab synthesizes these candidates and measures their electronic properties, everything Vals AI is describing—the band gaps, the Curie temperatures, the spintronic speed claims—is a hypothesis. A hypothesis worth pursuing, maybe, like a promising fishing hole someone told you about, but you ain't caught a fish yet.
Who is doing the hollering
These links show where the chatter came from. A link is attribution, not our endorsement or independent confirmation.
Last checked Oct 6, 2026, 5:06 AM EDT. Talk Around Town: These are computational predictions only—no lab has synthesized YBaMnFeO₅ or measured either material's electronic properties. The vals.ai team itself flags that YBaMnFeO₅ may be unsynthesizable in its useful ordered form. Independent expert review has not occurred. Treat all capability claims as unconfirmed hypotheses until peer-reviewed experimental results appear.