Notes
Essays on product, photography and the in-between. Long-form when it's worth it; nothing scheduled.
Too small to tell. Say so.
I built an agentic system that runs marketing experiments end to end. Its most valuable output was a refusal to answer.
Your agent works. Prove it.
A single run isn't a measurement. Agents are nondeterministic, most teams score them as if they aren't, and the number that ends up on the slide is often quietly dishonest.
Show, don't tell.
No, AI isn't making product management obsolete. It just let me stop telling people things and start showing them.
Nobody gave it the address.
Your agent logic is probably fine. The information underneath it isn't. That's the job now.
I asked Claude to build my MCP server. Here's what I actually got.
Standing up an MCP server took a day. Understanding what I built took longer. The interesting part wasn't the engineering.
Don't build the agent. Enable the access.
Why the next twenty years of product work is about exposing capabilities to agents, not designing destinations for users.
Published from a single manifest — a new essay is one file plus one entry, and the home list updates itself.