AI

Interpretable Context Methodology

How I build AI-assisted workflows with Claude and the Model Workspace Protocol — structured, reviewable, and built for real engineering problems, not demos.

MWP

Model Workspace Protocol

A structured way of organizing project context so Claude has exactly what it needs — no more, no less — for every task. It's the scaffolding underneath everything I build.

ICM

Interpretable Context Methodology

A discipline for keeping AI-assisted work reviewable: every decision traceable, every piece of context legible to a human, not just the model.

Vibe coding

I've done a little vibe coding

I'm not a developer and I don't claim to be one. But I've built enough alongside Claude to hold up my end of the conversation — to read what it wrote, follow why it's put together that way, and tell when an answer is wrong rather than just confident.

WellTrack is the example. A TypeScript app I built by filing issues against my own repo and telling Claude Code to pick one up and fix it — don't ask questions, just fix it. It did, and it did it well. It also stashed API keys in a JSON file, never created a .gitignore, and left security holes I had to catch myself.

That's the part worth keeping. The leverage isn't in generating code, it's in knowing what to reject — which is the same instinct ICM and MWP are built to make repeatable.

WellTrack on GitHub
Toolkit

AI tools I use

My actual toolkit for building structured AI workflows: Claude, Perplexity, n8n, Ollama, and communities where I learn. Transparent affiliate disclosure included.

Explore →

ICM projects — coming soon

This is where the write-ups on my ICM workflow, the CyberArk API change reporter, and other Claude-built projects will live. Check back soon, or follow along on the blog.

In progress