Case studies on building AI-native internal tooling — written like engineering posts.
Context, what shipped, decisions, and what broke. Each one is also an attempt to extract a methodology that travels — patterns I'd use again on the next problem.
Building finance dashboards when nobody writes the spec.
A one-sentence ask, no shared definitions. The PRD got built backwards: explore the data first with AI, settle the rules together, write the spec against the working prototype.
Building five CS skills for a team that didn't have a standard.
I'm not a CSM. The skills had to bake in pattern-matching I don't have. The way through was to play the meta-skill: substitute hard for missing CSM bandwidth, ship a v0 the team could push back on, and design an intake step that lets their judgment fill in on every run.
Building an AI operations system.
A 50-tool MCP server, automated daily digests, and a ticket workflow — built from frustration, refined through a deliberate reflection loop.
Knowledge architecture, from scratch.
120+ docs, a living decision log, enablement summaries, and a self-updating registry — designed so the team always knows what to do next.
Communicating model changes.
Translating statistical model updates into language that builds client trust instead of triggering escalations. 37 releases shipped without a crisis.