Self-evolving agent framework Épi hits 120 research records, tightens Klarner's constant
my_cat_can_code · x · 2026-09-16
A team introduces Épi, a self-evolving framework for autonomous research. Running in verifiable environments on sandbox infrastructure, Épi's agents produced 120 record-setting results in just a few days across mathematics, quantum information, operations research, and engineering optimization. Highlights:
- Lowered the upper bound on Klarner's constant in polyomino enumeration from 4.5238 to 4.29569;
- Proved the Cohn–Elkies method cannot reach the exact sphere-packing density in dimensions 10 and 11 even at its theoretical limit;
- Produced an independently verified, computer-assisted proof of a new case of Tuza's conjecture covering an infinite family of graphs.
The authors frame this as turning compute into verifiable research improvement — and say they're just getting started.
More from coding & agent
- Plasma's Radio Lets AI Agents From Different Providers Talk in One Channel — rohanpaul_ai · 2026-09-16
- Plasma launches Radio: a shared chat room where Claude, Codex and Grok agents talk directly — rohanpaul_ai · 2026-09-16
- Radio gives Claude, Codex, Grok agents a shared chat room, no more copy-pasting — rohanpaul_ai · 2026-09-16
- Telling your coding agent to slow down actually helps, developer reports — nilogram · 2026-09-16
- The Hacker's Guide to Attacking AI Agents: a systematic look at agent attack surfaces — _clickfix_ · 2026-09-16
- vAIvar: open-source honeypot feeds attacking AI agents endless hallucinations — Scobleizer · 2026-09-16