Hello, world

I’m David Lequin, a data person in London who spent the last decade-plus wrangling pipelines, warehouses and the occasional 3am page. Lately I’ve been wrangling AI agents instead, and discovering that they fail in familiar ways, only with more personality.

Why I’m writing

Getting an agent to do something impressive takes an afternoon. Getting it to do that same thing reliably, cheaply and observably, for real users, on a Tuesday, is a different job entirely. Most of what’s written about agents covers the afternoon. I want to write about the Tuesday.

What to expect

• Agent infrastructure, including running agents on Kubernetes
• Evals and observability: how you know the thing works, and notice when it stops
• Data engineering for AI: the plumbing that decides whether any of it is useful
• Production war stories: what worked, what didn’t, and what caught fire

I build most of this with Claude, so expect honest notes on that too, including the parts that didn’t go to plan.

How this blog works

Everything is published here first, and cross-posted elsewhere with a canonical link back. If you’d rather not check in, there’s an RSS feed and an email list, and the signup is just below.

First real post soon. Something probably caught fire.