SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness

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.

Introduction. Advances in foundation models enable agents to tackle increasingly open-ended tasks over longer horizons with less supervision [1, 2, 3, 4]. This shift supports applications such as autonomous research, software engineering agents, self-evolving personal assistants, and early forms of recursive self-improvement (RSI) [5, 6, 7, 8]. As agents operate over longer horizons, task-level token efficiency becomes a first-order systems concern [9, 10]. Existing efficiency work has primarily focused on lowering the cost per token through faster attention kernels and serving infrastructure [11, 12], model compression techniques such as quantization [13, 14], or the use of cheaper models [15, 16]. In this paper, we explore an orthogonal direction: improving token use through the agent harness that mediates interactions between the model and its environment. Harness-level optimization can improve efficiency without additional model training, complementing infrastructureand model-level approaches [17, 18]. However, optimizing a harness is difficult in practice.

Discussion / Conclusion. We introduce SoL-Pi, an RSI-inspired auto-research system for discovering and combining efficient agent-harness mechanisms. On EdgeBench, the best-performing candidates improve model performance by 5.3–12.8% and token efficiency by 9.8–18.2%, while the complete stack reduces token traffic by 44.7–49.0% and token cost by about one third at comparable performance. The complete stack’s consistent results across GPT-5.6 Sol and Opus 5 demonstrate strong cross-model generalization within the evaluated setting, positioning SoL-Pi as a preliminary step toward scalable RSI systems. Pre-Training the Harness. Generalizing harness artifacts discovered through RSI remains a persistent challenge. Our results provide preliminary evidence that scaling auto-research loops can mitigate this challenge. Analogous to pretraining, the harness is exposed to many tasks and updated from the resulting trajectories. We hypothesize that scaling both executable environments and the diversity of research ideas can yield sustained gains; we call this long-term research direction pretraining the harness.