Das Sarma paper written with Claude Opus 5.5, refereed by Codex
Event confirmedAwaiting review
The takeaway
A leading condensed-matter theorist discloses that the analytic derivations, numerics, figures and text of most of a long theory paper were produced by Claude under his direction, with a second AI as internal referee.
Status
- Claim
Event confirmedAwaiting review
- Our reporting
- High confidence
- Verification
- Preprint only
- Importance
- 3 of 5
- Last verified
- 17 October 2026
Your AI and this story
- GPT-6 Astra169 days after its cutoff
- Claude Opus 5.5108 days after its cutoff
- Gemini 4 Argon199 days after its cutoff
- Grok 4.7138 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 108 days before it.
Key facts
- Paper: ‘Exciton superfluidity in two-dimensional bilayers at zero and high magnetic fields’, Sankar Das Sarma & Yang-Zhi Chou, Condensed Matter Theory Center and Joint Quantum Institute, University of Maryland; arXiv 2610.22037 [cond-mat.mes-hall]; 13 appendices (A–M)
- Physics claims: in the ladder approximation the apparent plasmon enhancement of pairing is ‘largely cancelled by the second-order crossed-rung correction’; at large separation the pairing scale falls exponentially with squared separation while the instability persists; at νT = 1 a Hartree–Fock collective-mode instability favours ‘an excitonic supersolid’; the authors conjecture the quantum Hall bilayer evolves continuously into weak pairing of composite fermions, which could mimic a finite-temperature critical separation without a quantum phase transition
- AI statement (quoted): ‘This work made substantial use of artificial intelligence tools … The author posed the problems, specified the approach, and directed the work at each stage. For the zero-field theory (Secs. III–X and Appendices A–F) the analytic derivations, the numerical codes and calculations, the figures, and the text were produced with the AI system Claude (Anthropic; model identifier claude-opus-5-5) under that direction. Successive drafts of that part were criticized by a second AI system, Codex (OpenAI, GPT-6-Astra), acting as a referee, and were revised in response.’
- Checks named in the paper: the plasmon-induced coupling by two independent quadratures, the gap-equation solver against the BCS result, and limiting forms of the net second-order coupling against the full quadrature (Appendices D and F)
- The same AI was used to revise the νT = 1 theory text (Sec. XI, Appendices G–M) and to write the abstract, introduction, conclusion and Secs. II and XII; data ‘available from the authors upon reasonable request’
- No Hacker News thread or press coverage found by Oct 17
What happened
The preprint appeared on arXiv’s condensed-matter listing of 16 Oct 2026. Its long acknowledgments section splits the work by section: the zero-field theory (Secs. III–X, Appendices A–F) was produced by Claude Opus 5.5 under the author’s direction, and a second model, Codex running GPT-6 Astra, critiqued successive drafts “acting as a referee”. The quantum Hall part was human-derived, with AI help on the text.
Why it matters
Most AI disclosures in physics papers so far cover code or editing. This one comes from a prominent condensed-matter theorist (University of Maryland Condensed Matter Theory Center) and puts the analytic derivations themselves, plus code, figures and text, on the model, with an AI-vs-AI review loop before human submission. It is a concrete template for the APS disclosure policy the paper cites.
What is disputed or not yet verified
| Verification | Unrefereed preprint; internal numerical cross-checks listed in the paper |
|---|
Sources
1 source from 1 site. Numbers match the chips in the text.
1 source: 1 primary
Primary
Changes
- Filed (arXiv /list/cond-mat/new page for 16 Oct 2026; PDF read with pdftotext)