You have to beat the models at something
A clear argument for identifying the parts of engineering where human judgment still creates durable advantage.
Selected writing I return to or recommend.
-
-
The eval smell
A practical warning about evaluations that look rigorous while measuring the wrong thing.
-
What we can learn from Bun's rapid Rust rewrite with AI
A grounded case study in using AI during a large rewrite without losing sight of engineering constraints.
-
Bun in Rust
Bun's account of a runtime rewrite, and the design and performance trade-offs behind it.
-
A Lean syntax primer
An inviting introduction to the shape of Lean syntax and the ideas that make it feel different from JavaScript.
-
Harness engineering
A useful framework for thinking about the systems around an AI agent, not just the model at its centre.
-
How projects fail
A concise catalogue of the organizational and technical patterns that quietly turn projects into failures.
-
Party tricks
Small demonstrations of technical fluency that also reveal how engineers choose what to explain.
-
Good writing is obvious, not original
A reminder that useful technical writing earns its clarity through structure and attention rather than novelty.
-
Giving and taking credit
A thoughtful guide to recognition, ownership, and the incentives that shape collaborative engineering work.
-
You don't have to be smart if you think clearly
A practical case for disciplined reasoning over flashes of cleverness when making difficult decisions.
-
Thinking clearly
A compact set of habits for separating observations, assumptions, and conclusions.