# Blog

URL: https://codepanel.ai/blog

Notes from building CodePanel: code review as agents working with agents, how multi-model review works, how we measure it, and what we learned.

## Blog

- [How to tell if an AI code review change actually helps](https://codepanel.ai/blog/how-to-tell-if-an-ai-code-review-change-helps.md): Paired runs over the same fixtures, intervals resampled over fixtures, a sign-flip test and Holm. The pipeline changes we measured and did not ship.
- [Recall is not enough: measuring false positives in code review](https://codepanel.ai/blog/measuring-false-positives-in-code-review.md): How CodePanel counts false alarms: clean and trap fixtures, blocking recall, and precision from verdicts the author agent cannot give itself.
- [Claude Sonnet 5.5 was the fastest reviewer in our benchmark, and tied for first](https://codepanel.ai/blog/claude-sonnet-5-5-review-results.md): Claude Sonnet 5.5 tied for first in the 2026-10-17 run at 10 seconds a review, the fastest of 17 models, and cost more per review than its list price suggests.
- [GPT-6.1 Sol had the highest measured recall in our benchmark, at about a third of GPT-5.6 Sol's cost](https://codepanel.ai/blog/gpt-6-1-sol-review-results.md): In the 2026-10-17 run, GPT-6.1 Sol had the highest measured recall of 17 models, with few false alarms, at about a third of GPT-5.6 Sol's cost per review.
- [First benchmark results: seven open models on real bugs](https://codepanel.ai/blog/open-models-benchmark-results.md): DeepSeek V4 Pro and GLM-5.3 Flash tied for first, and the Flash model costs a seventeenth as much per review. What we ran and what surprised us.
- [Code review should be agents working with agents](https://codepanel.ai/blog/code-review-is-agents-working-with-agents.md): Why CodePanel is built for a coding agent and a panel of reviewer agents working on one pull request, and what people still decide.
- [How we built the code review benchmark](https://codepanel.ai/blog/how-the-benchmark-is-built.md): Real defects from open-source history, clean changes to count false alarms, a three-family judge panel and intervals on every number. How it grows.
- [Why CodePanel reviews with a panel of models](https://codepanel.ai/blog/why-a-panel-of-models.md): Several cheap reviewers from different model families, then one orchestrator that checks every finding against the code. Why we built review this way.