StateM: Improving Long-Horizon Agents via System Scaling
Successful-Western27 · reddit · 2026-08-25
StateM posits that long-horizon agent failures are often execution-system failures. By using durable state checkpoints, phase-local context, and recoverable runbooks, the system significantly boosts success rates (e.g., GPT-5.5 xhigh from 83.1% to 92.1%). Notably, a runbook developed for one model improves another without changes, highlighting the value of explicit state management and raising questions about benchmark attribution.
More from coding & agent
- Dev pushes back on "AI coding means insecure apps": humans always shipped bugs — dreamwieber · 2026-10-15
- Eight non-negotiable principles for building AI that actually learns from experience — tokenbender · 2026-10-15
- Animated teardown of XBOW's Chrome full-chain exploit that won Google's first $250K VRP bounty — moyix · 2026-10-15
- pstack's durable core: 10-20 line skills that distill agent best practices into principles — HankYeomans · 2026-10-15
- 'Children of Elon' forum lets AI agents post and teach each other while humans watch — AiJohnAllen · 2026-10-15
- VoidZero open-sources Void: SDK, CLI and platform to deploy Vite apps on Cloudflare — cnakazawa · 2026-10-15