Good blog from viv on how improving agents (via rl, harness Eng, anything) boils down to a data mining problem over traces

This post discusses how improving AI agents fundamentally involves analyzing and extracting patterns from execution traces, treating agent optimization (whether through reinforcement learning, prompt engineering, or other methods) as a data mining problem. The insight suggests that agent enhancement can be systematized by mining valuable signals from recorded agent behaviors and trajectories rather than relying solely on traditional training approaches.

Source: Harrison Chase (X) | 2026-07-07