SoL-Pi: Auto-research loops optimize coding agent harnesses, cutting tokens 44-49%
burny_tech · x · 2026-09-20
A paper from NVIDIA, MIT and collaborators, SoL-Pi applies RSI-inspired auto-research loops at the harness layer, letting coding agents discover their own framework optimizations instead of hand-designing them.
Across diverse environments, the selection process yields four surviving mechanisms—action execution, context compaction, observation handling, and delegated reading—that make up SoL-Pi.
Key results: On the 51-task EdgeBench, SoL-Pi matches Pi's performance with GPT-5.6 Sol and Opus 5 while reducing recorded token traffic by 44.7–49% and API cost by roughly one-third—estimated savings of $8.75–13.50/hour vs. native Codex and Claude Code harnesses, and $4.36–5.71 vs. Pi.
More from coding & agent
- Dev Uses Computer-Use Agent to Install/Uninstall 300+ Mac Apps, Adds 90 Cleanup Rules — vista8 · 2026-09-20
- Spotify ships Xirp, a Mac app born from 36,000 AI coding sessions — shashib · 2026-09-20
- Resetwatch plugin aggregates usage limits and reset times for 14 AI providers in one page — Teknium · 2026-09-20
- An inbox re-ranked live by importance instead of time, built with jev — xkonjin · 2026-09-20
- DAIR.AI launches auto-updating Awesome Jev collection, curated by Jev itself — omarsar0 · 2026-09-20
- Open-source Tampermonkey scripts add paste-to-upload for WeChat, Xiaohongshu and Douyin — vista8 · 2026-09-20