GPT Astra improves Anthropic's zeta zeroes certificate from 67.25% to 67.31%
dsanft · reddit · 2026-09-16
A Reddit user used GPT Astra Max (on a 20x Codex plan) to improve on Anthropic's recent zeta zeroes certificate related to the Riemann Hypothesis:
- Anthropic's original result: 67.25007% of simple zeta zeroes on the critical line
- After iteration and several "Tibo resets", Astra pushed it to 67.31056%
- The author reused Anthropic's toolchain but needed manual steering, plus a dedicated robust CI/CD and commit-gate certification setup
- Lean proof and exhaustive numerical certificate are open-sourced at dbsanfte/RiemannGaussian
Further proofs, including a narrow zero-free region not found in the literature, are in progress but unpublished.
More from coding & agent
- Hugo Bowne on the biggest agent failure mode: it's the harness, not the model — hugobowne · 2026-09-16
- Lean Verified Transformers: human-written text, AI-written proofs of core invariants — srush_nlp · 2026-09-16
- One prompt turns any book into a chapter-by-chapter podcast site with parallel subagents — armand_ruiz · 2026-09-16
- Image-to-WebDev Arena: GPT-6 Astra tops at 1733, GLM-5.3-Flash shines on price — arena · 2026-09-16
- Probability-only model tested as LLM judge: 25x faster, 600x cheaper than Fable-class — danshipper · 2026-09-16
- Two years with an AI career coach: my agent now knows my job better than I do — danshipper · 2026-09-16