EdgeLogby
Making Methods Philosophy
About
EN / KO
← All Tags

#benchmark

2 posts

  • How Many AI Code Reviews Is Enough?

    I spent 400 LLM calls on one codebase to learn this. One review pass finds 34% of the real defects, and the four most-repeated findings were all false.

    Aug 22, 2026 · 30 min read
  • AI-Written Firmware: Measuring the Gap With 233 Tests

    Sonnet 4.6 scored 68.0% and Haiku 4.5 56.9% on 233 firmware cases. EmbedEval is my open benchmark: 5-layer eval, n=3, Wilson 95% CI, Zephyr, STM32 HAL.

    Apr 17, 2026 · 5 min read
/SUBSCRIBE ยท EMAIL

The notebook, by email

Products made alone, with AI, plus the measurements and failures behind each call. Biweekly, no spam.

Unsubscribe anytime.

Or subscribe via RSS.
HomeMakingMethodsPhilosophyAboutAll essays
© 2026 EdgeLog·made by
GitHub·Email·NeuroTerm·RSS