Now

Updated September 2026

Building Flock

Most of my attention is on Flock Synthetics, where I’m co-founder and CEO. We use AI agents to test web products from different user perspectives and find where people get confused or stuck.

Right now, I’m focused on improving what it catches and making the findings easier for teams to act on. We test changes to models, prompts, and workflows against real app experiences with known defects, then bring the findings and browser evidence into the coding tools teams already use.

The question behind a lot of this work: did the change actually make the product better? Sometimes the upgrade isn’t one.

How people use AI

I’m also testing how people engage with the same capabilities in different places: a web app, directly inside an AI platform, or through a general-purpose agent. That includes MCP, the protocol that connects agents to external tools.

When do people want to click through an interface? When would they rather ask for something? And when are they comfortable handing over the whole task?

Family Atlas

I’m building Kelch Family Atlas, a family genealogy project that doubles as an experiment in automated research. I’m exploring how agents can follow leads, compare records, and keep track of what the evidence actually supports.

It’s currently split between my machine and hosted services. I’m gradually moving the work into the cloud, toward fully automated research that keeps progressing when my laptop is closed. The challenge is keeping sources, conflicting evidence, and uncertainty visible as more of the work happens without me.

Advising and writing

I continue to advise founders and leadership teams on what to build, how to organize the work, and where AI can help. I write about the decisions and mistakes that come out of all this.

Recent writing: The Right Tool for the Job, on deciding whether to change models, and How Things Are Made, on the work between a convincing demo and a useful product.