Lines of Code Got a Better Publicist
A case against measuring AI-assisted engineering by code volume and other vanity metrics instead of outcomes.
Read more →Posts
Is coordinating a fleet of AI agents the same as managing a team of people? Yes, but mostly no.
Arguing that coordinating AI agents is closer to staff engineering than people management — and why that framing matters.
Read more →
Calibrating on Capability, Not Activity
A case for measuring AI-assisted engineering by observable capability, not tool telemetry, prompt counts, or PR volume.
Read more →
My AI Context System: The Sequel
A month-on update to my AI context system — what I added, what I threw away, and why selective automation beat full automation every time.
Read more →
Teach the AI to Fish
Why your documentation needs to serve AI agents as first-class readers — and why writing a CLAUDE.md should be as deliberate as choosing a framework.
Read more →
Your Theory is the Gold, not your Code.
Naur's 1985 idea that programs are theories in developers' minds reframes what AI-assisted development puts at risk, and why I'm not that worried.
Read more →
Here's My AI Context System
A walkthrough of my open-source, markdown-based AI context and memory system, why plain files beat vector databases, how the models have caught up, and how to get it running yourself.
Read more →
My Posts Started Looking Like Everyone Else's
How leaning on AI for writing structure turned my blog posts into generic slop, and what I learned about detecting and avoiding it.
Read more →
Throwing away my fancy AI memory system
Why a simple markdown-file-based AI memory system outperformed a vector and graph database setup for personal knowledge management.
Read more →