Claude Audits Quantum Gravity news update

Physicist E. O. Kahya submitted a new research auditing paper to arXiv in 2026 detailing a full computer-assisted audit of an established one-loop quantum gravity derivation, using Anthropic's Claude Opus 5.5 model as a checking tool alongside independent SymPy and mpmath code checks. According to github.com's report, the project re-derived two foundational theoretical physics studies published in 2007: Paper I in Physical Review D, which established the graviton correction to a massless, minimally coupled scalar field on de Sitter space, and Paper II, which tracked late-time scalar field equations during cosmological inflation. The audit tracked every algebraic transition across the printed record, uncovering typographical slips, counterterm adjustments, and an overall sign inversion while verifying that the physical conclusion remains sound.

This report covers the mechanics of the verification workflow, the specific mathematical corrections identified across the papers, and the consequences for researchers modeling quantum field theory in curved spacetime. It is intended for theoretical physicists, cosmologists, and computational researchers following how automated reasoning systems audit lengthy multi-page calculations. By walking through eighteen sequential tasks grouped into nine distinct stages, the audit confirmed that inflation-era infrared gravitons do not alter the frozen late-time scalar mode functions, providing a blueprint for verifying complex analytical work.

Claude Audits Quantum Gravity news update: what actually changed

The audit identified thirty-three total discrepancies between the published 2007 papers and the re-derived mathematics, categorized into four substantive findings, six absorbable finite errors, and twenty-three typographical errors. The most prominent substantive finding, designated S1, showed that the bare self-mass-squared printed in Paper I carries an overall factor of negative one relative to the definition set by equation I.168. Both the four-point seagull diagram and the three-point diagram carry this reversed sign, meaning their relative sign remained internally consistent while flipping every entry in Table 5 of Paper II and changing the overall sign of the first-order correction.

A second substantive finding, labeled S2, identified a missing piece in the four-derivative zeta row of Table X, which omitted eight times the natural log of x minus sixteen times x times the natural log of x from Table III. This corrected the row expression from ten-thirds to thirty-four thirds for the leading logarithmic coefficient and shifted the second term from negative twenty-four to negative forty. Finding S3 adjusted the constant term in the sixth-order scale-factor row of Table X from negative fifty-four to positive eighteen, resolving an omitted term equivalent to seventy-two gamma-two. Meanwhile, the divergent pole of negative four divided by D minus four remained unaffected, and findings B1 through B6 adjusted finite counterterm coefficients beta-two and beta-three without physical impact.

Claude Audits Quantum Gravity news update: how it works

To reconstruct the calculation, the audit decomposed the two original papers into eighteen discrete tasks and nine sequential stages. The first two stages recalculated Feynman rules and the four-point seagull diagram contributions spanning equations I.6 through I.44. The subsequent four stages worked through the three-point diagram setup, isolating local and logarithmic terms across equations I.54 through I.67, processing normal terms in Table III, and expanding the operational evaluations across Tables IV and V up to equation I.125. The final three stages evaluated de Sitter symmetry breaking, gauge counterterms, renormalized self-mass in Table X, and the Schwinger-Keldysh in-in field equations.

Anthropic's Claude Opus 5.5 acted as an analytic checking partner during the review, deriving each expression index by index while keeping the spacetime dimension D general wherever the original papers maintained dimensional regularization. Every symbolic expression was paired with dedicated SymPy and mpmath code to run thirty-digit numerical cross-checks across three test coordinate pairs. A global consistency test confirmed that the direct sum of all normal diagram contributions matches the operator expansions of Table X only when both findings S2 and S3 are applied simultaneously.

Claude Audits Quantum Gravity news update: why it matters now

The primary physical conclusion originally published in Paper II survives the audit intact. During cosmological inflation modeled on de Sitter space, the expansion of spacetime pulls long-wavelength, non-conformally invariant gravitons out of the vacuum, creating an infrared bath. Because gravitons couple to massless minimally coupled scalars exclusively through derivative interactions, and because scalar mode functions freeze to constant values after leaving the cosmological horizon, the physical interaction was expected to shut off at late times. The audit proved that the quantum-corrected mode function continues to approach a constant state modified only by a field-strength renormalization term labeled C-tilde.

No secular terms growing as the logarithm of the scale factor, the square of the logarithm, or the scale factor itself appear in the final solution. The substantive correction S2 eliminates sub-leading late-time tails proportional to momentum over Hubble scale factor to the fourth and sixth powers divided by the scale factor, ensuring the remaining late-time terms decay as the logarithm of the scale factor divided by the square of the scale factor. The constant shift from negative fifty-four to positive eighteen in finding S3 is independent of the spatial coordinate y, causing it to drop out of the retarded Schwinger-Keldysh kernel identically.

Claude Audits Quantum Gravity news update: who is affected

Cosmologists and gravitational theorists relying on the Kahya-Woodard one-loop calculations for inflationary perturbations are the primary audience affected by the newly corrected coefficients. While researchers do not need to revise broader cosmological conclusions regarding scalar stability during inflation, anyone writing code based on the operator listings in Table X or the intermediate tables in Paper I must adopt the corrected algebraic forms. Researchers using the earlier literature can now substitute the thirty-four thirds coefficient and the positive eighteen constant into their own symbolic pipelines without tracking intermediate discrepancies by hand.

The project also directly affects practitioners evaluating artificial intelligence in formal scientific research and peer review. Because the audit was led by E. O. Kahya, an original author of both 2007 papers, the findings show how domain experts can use general-purpose reasoning models to audit multi-page algebraic derivations that human referees rarely recompute from scratch. The method provides an operational template for combining large language models with rigorous, independent computer algebra systems to eliminate legacy transcription slips from the scientific literature.

Claude Audits Quantum Gravity news update: what to watch

The immediate focus moves to the arXiv submission of the corrected scalar self-mass paper and potential follow-up evaluations of simplification II.44. The audit noted under finding S4 that the specific algebraic justification given in Paper II for simplification II.44 fails term by term, because two intermediate gradient rows grow with opposite signs before canceling. Without relying on that simplification, the audit verified that non-falling terms cancel through order k to the tenth power, relying on the quadratic logarithmic term present in both the printed and corrected rows.

Physicists will also watch whether similar audit frameworks are applied to the remaining appendices and related loop calculations in quantum cosmology. The author noted that the bookkeeping tables in the appendix of Paper I were audited rather than regenerated from scratch, leaving open potential audits of higher-loop diagrams or alternative gravitational gauges. As reproducible notebooks combining symbolic solvers with machine-assisted proof checking become standard, additional decades-old curved-spacetime calculations are likely to undergo systematic programmatic review.