Research

Notes, results, and findings from our work on autonomous agent optimization.

May 28, 2026
Beating τ-bench retail by 10% How a three-person team outperformed the best public models on τ-bench retail — and what it taught us about agent optimization.