SoL-Pi paper: auto-research loops save $4-13 per hour on coding agents
alex_verem · x · 2026-09-23
arXiv details for SoL-Pi (2609.20519, submitted Sep 17, 2026): 14 authors including Haozhe Liu, Song Han and Enze Xie; 15 pages, code and project page released.
- As coding agents move to unattended exploration, token efficiency becomes critical for recursive self-improvement; the paper scales auto-research loops at the harness layer across increasingly diverse environments.
- Four selected mechanisms span action execution, context compaction, observation handling, and delegated reading, forming SoL-Pi with improvements that transfer beyond the development setting.
- Numbers: on EdgeBench across GPT-5.6 Sol and Opus 5, performance matches Pi while recorded token traffic drops 44.7–49.0% and API cost falls 1/3 — estimated hourly savings of $8.75–$13.50 vs native Codex/Claude Code harnesses and $4.36–$5.71 vs Pi.
Related event: NVIDIA's SoL-Pi auto-optimizes agent harnesses, cutting tokens ~45-49%(2 posts)→
More from coding & agent
- Dev's 'software factory' burns 5-10B tokens a day with 32 parallel coding agents — BLUECOW009 · 2026-09-23
- Dev shares how he built app moderation with Jev and LLMs — amos_gyamfi · 2026-09-23
- Dev argues the real AI agent problem is unmeasurable requirements, not capability — algo_diver · 2026-09-23
- $/task beats token price: outputs are <5% of tokens in agentic coding — xeophon · 2026-09-23
- Rethinking multiplayer agent UX: shared artifacts may beat a single group-chat box — max__drake · 2026-09-23
- Dev: if your eval costs $20 for 100 runs, the problem set isn't hard enough — pvncher · 2026-09-23