Community runs dozens of Codex and Claude agents to break integer multiplication bounds overnight
neuroecology · x · 2026-10-10
- Rohan Arun reports rapid community progress on OpenAI's conditional integer-multiplication bound: κ improved from 2⁻¹⁸² to 6.83e-4 in a day, with records falling overnight.
- Key improvements: stopping recursion before small ops get expensive, sharing scratch across completed stages, paired-cube transforms on original inputs, and cost-optimized circuits.
- Running dozens of Codex (Astra high) and Claude Opus 5.5 agents in parallel: Codex reliably submits advances faster; Opus 5.5 set the current leader and keeps up when progress slows. Best breakthroughs still come from human intuition pushing models out-of-distribution.
- The leaderboard format makes contributors work slightly adversarially, spawning new strategy branches; math experts are invited to propose wild ideas with their own Codex/Claude at the repo.
Related event: Community of AI Agents Pushes Integer Multiplication Bounds to New Records(4 posts)→
More from coding & agent
- Asking an agent to fix a bug you don't understand is continuous paperclip maxxing — brandon_xyzw · 2026-10-10
- exe.dev offers SSH-first cloud sandboxes for developers and AI agents, drawing rave reviews — davidcrawshaw · 2026-10-10
- Surge AI launches sudo L7: a benchmark testing whether coding agents can act like staff engineers — rmcwhorter99 · 2026-10-10
- Live feed shows what images AI agents use while hunting for new planets — BLUECOW009 · 2026-10-10
- Juggling 5 agent coding sessions, devs resort to pasting 'how are we lookin'?' — tdhopper · 2026-10-10
- Cognition: 91% of internal Devin sessions now start without a human — charles_irl · 2026-10-10