Guides
Practical, step-by-step writing on spec-driven development, evaluation, and AI product practice
Where the field reports describe what happened, guides show you how to do it: adopting spec-driven development (SDD) on your own team, designing evaluations that catch what matters, and building an AI practice as a product manager. Written from practice, not theory.
Data Products for Robots
AI agents are now first-class consumers of your data. A practical guide to what changes when the reader of your documentation is a machine.
The Spec That Lints Itself
A small, free, no-dependency script that checks a markdown spec for the gaps that force an AI coding agent to guess. Download it and run it today.
The Gate That Fires Every Time
Telling an AI assistant how you like things works for about an hour. A rule that triggers on an observable signal works indefinitely. The pattern, and the file.
Write the Skill, Not the Answer
Solving a problem well and losing the solution is the default. A skill file turns a one-off answer into permanent capability. The structure, and the template.
Right Knowledge, Right Lifetime
An AI assistant forgets everything between sessions unless you build the memory yourself. Three scopes, one rule, and why a lesson has to land in the file that matches its lifetime.
Go Do X and Report Back
Skills are how I do something. Agents are go do it and come back with one report. The line between them, when to dispatch, and the brief that makes a stateless worker useful.
The Record You Can Mine
Every work session my assistant runs is logged to a store I can query. That one habit generates my standups, finds last month's fix, and shows where my own workflow has friction. A script to start with.
A Workspace That Maintains Itself
The difference between a tool you use and an environment you live in is who does the upkeep. Four background tasks that run without me, and the smoke test that catches a broken skill the minute it breaks.
The Contract Comes Before the Pipeline
A data contract is the spec's last mile: the document a consumer can hold a producer to and an agent can build from. What goes in one, why each section exists, and a template.
The Judge Needs a Rubric
A model grading your specs is only as good as the rubric it holds and the calibration set that proves it applies it your way. The rubric, the prompt, the ten-document test, and the five ways it goes wrong.
What Kind of Change Is This?
Specs need version numbers as much as code does, and the number should be decided by rule, not by mood. A small script that reads two versions of a document and says major, minor, or patch, with reasons.
Get New Posts by Email
Three essays a week, Sunday, Wednesday, and Friday evenings. No spam, just writing.
Building With These Ideas
I am building North Bound, a product management app that applies this thinking: AI as a thinking partner through the product requirements process, from messy inputs to a spec an agent can build from.
Join the early access list