I share what I build as I build it: what worked, what didn't, and how a non-technical operator makes real systems with AI. These are the pieces, one at a time.
The first version was so generic I never launched it. What changed the second time: clarity, and an operating system that already knew my work.
A sticker game and a grant proposal, both confidently wrong. What went wrong, how we rebuilt them, and the exact prompt to give AI before any task.
Substack now scores posts for AI. It measures sentence structure, the one thing that doesn't matter, and the people already gaming it prove why.
I asked Claude to review the transcripts of my own sales calls. The coaching loop I built from what it found, and why the standard you set is what makes it useful.
AI's errors hide inside polished work. The grade-and-fix loop that catches them, and why the quality comes from the judgment you put in, not the tools.
How a small, part-time team runs whole marketing campaigns with a system of AI agents that grade each other's work, and where we keep human judgment in the loop.
Most people I speak with don't know where to start. Onboard AI like a new team member and let it interview you. Prompt included.
I heard one prompt on a podcast and ran it against my AI operating system. Here's how I keep it improving as new features ship every day.
Building Zanzibar's community health system taught me to redesign work around outcomes. Most teams are making the same mistake with AI today.
Everyone hates AI slop. I'm more interested in what it takes to get real value out of AI, and the patterns others can learn from.
Watching Claude build a multi-platform procurement system through MCP connectors, end to end.
The "memory" everyone asks about isn't magic. It's plain text files, read and updated every session. Context engineering, explained.
Not everything works. Here's one I built, measured, and shut down, and what it taught me.
The AI business analyst I'd wanted for five years: every Sunday it reads 14 podcasts, 20 newsletters, and the research, and writes me a brief.
How I built an LLM Wiki in Obsidian over a weekend, and why it became the backbone of how I work.
The build that reached 90,000 people, with no code. What it does, and why so many people wanted one.
These notes are the short version. My newsletter is where I go deep: every prompt, every step, and what I'm learning from the people and organizations I work with.
Subscribe to AI on PurposeEverything I write comes back to one idea: adopt AI deliberately, so it strengthens your thinking, in service of work that matters. That's what I mean by AI on purpose.