Do automated evals work? A comparison of LLM tools on app traces
HamelHusain · x · 2026-07-24
A discussion of whether automated evaluations actually work, centered on a method that uses multiple LLM-powered tools to find failures in application traces and compare them with domain-expert labels.
The attached chart compares several systems by recall, precision, discoveries, and false positives, showing the trade-offs between catching more issues and keeping noise low.
- Braintrust Loop: 87.2% recall, 79.1% precision, 20 discoveries, 9 false positives
- Arize AX Alyx: 74.4% recall, 91.0% precision, 19 discoveries, 3 false positives
- LangSmith (chat agent): 79.5% recall, 77.5% precision, 20 discoveries, 9 false positives
- Codex (GPT-5.5 High): 84.6% recall, 82.8% precision, 20 discoveries, 11 false positives
- Factory Droid (GPT-5.5 High): 84.6% recall, 83.3% precision, 17 discoveries, 10 false positives
- Claude Code (Claude Opus 4.8): 79.5% recall, 77.4% precision, 17 discoveries, 14 false positives
More from coding & agent
- Codex reviewed 1,450 pull requests for one user in 30 days — 0xkarasy · 2026-07-24
- Tyler Agg outlines a framework for evaluating deployed AI agents — tyler_agg · 2026-07-24
- A local two-agent coding setup uses shared memory to preserve architecture decisions — PrajwalTomar_ · 2026-07-24
- New survey maps the path to long-horizon AI agents — jiqizhixin · 2026-07-24
- How one developer uses ChatGPT Voice as a hands-free control panel for Codex — Dimillian · 2026-07-24
- A practical breakdown of 10 AI agent types, from reactive to multi-agent systems — goyalshaliniuk · 2026-07-24