Claude agents screen candidate magnetic semiconductors
A Vals AI research team used Claude Opus agents to propose and assess two magnetic-semiconductor candidates, publishing calculations, code and caveats.
vals.ai
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Explore candidate materials with these target properties. Record the data, assumptions, calculation code and uncertainty for each candidate, and separate theoretical predictions from experimentally verified results.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- Geby Jaff (Vals AI)
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The team reported two computational candidates with predicted properties of interest. No synthesis or device result is reported.
- Human involvement
- Human researchers worked with the agent team and published the calculations, code and caveats. The predictions are not presented as experimental confirmation.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- The team identified the spin-related material properties it wanted to investigate for memory applications.
- Claude Opus agents and researchers proposed a new material and examined a previously made material as candidates.
- The team published calculations, code and caveats; the results remain predictions requiring scientific and experimental review.
Geby Jaff describes a research project in which a team of Claude Opus 5.5 agents and human researchers proposed two candidate Luttinger-compensated magnets for possible spintronic memory applications. One candidate was designed by the team; the other was a material first made in 1999 that their calculations suggested might have the desired properties. The authors published their calculations, code and known caveats alongside the explanation.
The result is a pair of computational candidates, not experimentally confirmed materials ready for a device. Jaff’s article explains the physical property they were looking for and why it could matter for memory, then presents the theoretical predictions and limitations. The evidence lets readers inspect the calculations rather than treating the agents’ output as a final scientific conclusion. It does not show that the candidates were synthesized or that a memory device was built.
This is a research-assistant pattern worth copying carefully: use agents to explore candidate materials and generate an auditable calculation trail, publish the code and assumptions, and treat the result as a lead for human scientific review. The source does not reveal a single prompt or exact division of labor between agents and researchers, so the suggested prompt below is ours.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Website · Geby Jaff (Vals AI)
Frequently asked questions
What did the research team identify?
The article presents two candidate Luttinger-compensated magnets: one designed by the team and one previously made in 1999.
Were the candidates experimentally confirmed?
No. The article reports calculations and predictions, not experimental validation or a working memory device.
What evidence did the authors publish?
Jaff says the team shared calculations, code and caveats so readers can inspect the basis for the predictions.
More details
About this account
The authors say a team of Claude Opus 5.5 agents contributed to finding and designing material candidates; the source does not detail exact autonomous steps.
Reviewed the full Vals AI research article, author, publication date, calculations and caveats. It presents theoretical candidates, not experimentally confirmed materials. The source does not provide a verbatim prompt.
Credit: Geby Jaff — Vals AI
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Opus 5.5 agents · Scientific calculations · Code
materials science · agent research · spintronics
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.