Blog
Writing about AI, engineering leadership, and what actually works in software delivery.
Choosing Engineering Metrics, Six Years On
I reread an article on engineering metrics that I wrote six years ago. Most of it still holds. I now start with four questions, give each metric one use, and prefer metrics that the author of a change did not write.
AI-Ready Data, From Talk to Book
This year, one idea about agents and data became three talks and an article. A book chapter comes next. Here is what changed at each step, so that you can pick the version that fits your time.
Engineer What Surrounds the Model
A coding agent has a model at its center and two rings around it. The vendor builds the inner ring. You build the outer ring, and a better model will not build it for you.
The Architect's Playbook for AI-Assisted Development
AI coding tools change what an architect does, not whether they're needed. Here is how to create the conditions where AI-generated code stays correct, safe, and architecturally sound.
What Does Spec-Driven Development Need to Actually Work?
Most discussions of Spec-Driven Development flatten it into one step. Write a spec, get the output. The version that survives real systems is a five-layer system, and the spec is only one of them.
AI-Augmented SDLC: Part 3 - Making It Work
AI tools accelerate delivery when BOTH conditions hold: pre-coding is settled and post-coding is representable in the coding process. Here's the formula — and what to fix if you want to actually ship faster.
AI-Augmented SDLC: Part 2 - The Bottlenecks
AI tools compressed coding time. But requirements, architecture, cross-team dependencies, and post-coding friction didn't compress with it. Faster coding meant faster arrival at the next blocker.
AI-Augmented SDLC: Part 1 - The Ecosystem
AI coding tools represent a genuine step change. Skills encode domain expertise, subagents handle scale, MCP integrations keep you in flow. Here's what the ecosystem can actually do — across three real projects.