I shipped an AI assistant adopted by 20+ customers in its first month — and I use AI daily as a PM. Here's the working notebook: what I'm learning, the prompts I reach for, and things I've written down.
How agents that plan and act (not just answer) change product design — building on what we learned shipping Quinn's "suggest the next action" agents in QuickHub.
You can't A/B test your way out of a hallucination. Learning to write evals that catch quality regressions before customers do.
Quinn won 20+ customers in month one — the real work was making week 4 look like week 1. Novelty gets the first prompt; usefulness gets the hundredth.
Non-technical users freeze at an empty chat box. Suggestions, templates, and proactive agents beat "ask me anything" every single time.
Prompts I actually use as a PM — copy any of them.
"Teach me [topic] from first principles. Structure it as a 30-page PDF: start with fundamentals, build up to advanced concepts, include examples and diagrams where useful, and end with a summary and further reading. Write it like a textbook chapter, not a listicle."
OPEN — COPY IT + SEE A RESULT → MARKET RESEARCH"Research the market for [product/industry]: market size and growth, key trends, major players, regulatory factors, and where demand is underserved. Cite sources where possible and flag anything that's an estimate vs. confirmed data."
OPEN — COPY IT + SEE A RESULT → USER RESEARCH"Design a user research plan for [problem/feature]: research questions, ideal participant profile, interview script (open-ended, no leading questions), and how to synthesize findings into themes ranked by frequency and impact."
OPEN — COPY IT + SEE A RESULT → COMPETITOR RESEARCH"Compare [competitor 1], [competitor 2], [competitor 3] on pricing model, ICP, onboarding flow, and core differentiator. Then tell me: what segment do they all underserve, and what would a product built only for that segment leave out?"
OPEN — COPY IT + SEE A RESULT →A practical framework for PMs to get sharper, more useful output from AI — structuring context, constraints, and desired format instead of one-line asks.
READ THE FULL POST → POST 02 · BUILDING WITH AIHow I actually go from idea to shipped feature using AI — from scoping and prototyping to validation — and where AI speeds things up vs. where it doesn't.
READ THE FULL POST → POST 03 · PRODUCT SENSEAI can generate options, but it can't make the call. What good product judgment actually looks like, and why it's the one thing AI hasn't replaced.
READ THE FULL POST →