Dan Luu (danluu.com)
- title
- Dan Luu (danluu.com)
- type
- entity
- summary
- Dan Luu's blog — measurement-driven engineering writing from a CPU-verification background, covering testing, reliability, org dynamics, and now AI coding
- tags
- blog, author, testing, agentic-coding
- created
- 2026-07-21
- updated
- 2026-07-21
The blog of Dan Luu, an engineer whose first decade was at Centaur, a CPU-design shop where hardware verification (dedicated test engineers, fuzzing, no code review) was a first-class discipline. He later moved to software. The blog's recurring theme, by his own account, is measurement — benchmarking, evals, and experimental design — which is the lens he brings to everything from org structure to, more recently, agentic coding.
The writing style is distinctive: long, data-first, allergic to received wisdom, heavy on footnotes and concrete numbers, skeptical of claims that don't come with distributions. He tends to argue from direct experience and specific measurements rather than from principle.
Topics
- Testing and reliability methodology (fuzzing, randomized testing, why unit tests underperform)
- Benchmarking, evals, and experimental design
- Measurement of everyday software quality ("everything is broken," "nothing works")
- Org dynamics, specialization, and engineering careers
- AI coding — testing-heavy no-review workflows, model variance, agent failure modes
Ingested articles
- danluu-ai-coding-testing — his 2026 essay on AI coding: testing-heavy no-review workflows, the meaninglessness of single-number benchmarks (llm-output-variance), and working around agent failure modes (agent-failure-mode-skill)
Related concepts extracted: testing-heavy-no-review-workflow, llm-output-variance, agent-failure-mode-skill.