Scaling recursive AI loops reduces token use and costs sharply
SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness
Artificial Intelligence
Summary
When AI agents work nonstop to code and research on their own, they use a lot of data tokens and API calls, which costs money. The authors present SoL-Pi, a method that makes these AI loops more efficient by reusing improvements and managing information better. This method cuts token use by nearly half and lowers API costs by about one third while keeping performance high. SoL-Pi works well across many coding tasks with different AI models.
What this means in practice
- •For software development teams: Reduce costs of automated coding and testing by improving efficiency of AI-driven research loops across diverse tasks.
- •For cloud service operators: Cut API usage and related expenses in AI-powered agent services by applying token-efficient harness management techniques.
Authors
Haozhe Liu, Tian Ye, Sensen Gao, Qihang Cao, Yitong Li, Mingchen Zhuge, Duomin Wang, Ruihua Zhang, Ping Luo, Jiawang Bian, Lei Zhu, Ligeng Zhu, Enze Xie, Song Han
Abstract
As coding agents move from supervised code completion to unattended, around-the-clock exploration, their work expands from isolated predictions into long trajectories of reasoning, tool use, and feedback. Token efficiency therefore becomes important for scaling recursive self-improvement. We take an RSI-inspired approach at the harness layer, scaling auto-research loops across increasingly numerous and diverse environments for harness rollouts. At this scale, the process yields reusable improvements that transfer beyond their development setting, moving automated harness discovery toward production-level outcomes. Four mechanisms survive selection and form SoL-Pi, spanning action execution, context compaction, observation handling, and delegated reading. On the 51-task EdgeBench evaluation, SoL-Pi achieves performance comparable to Pi across GPT-5.6 Sol and Opus 5 while reducing recorded token traffic by 44.7-49.0% and API cost by about one third. In other words, estimated hourly savings are \$8.75-\$13.50 relative to native Codex and Claude Code harnesses, and \$4.36-\$5.71 relative to Pi.