Why I'm writing

2026-05-16

I'm David. I've shipped a couple of indie apps linked from the homepage — Movr and Fernweh — alongside a decade in product engineering (BCG Digital Ventures, Blinkist, currently Sensorfact). Long before that I co-founded a fitness startup called Evomo, where we trained models to detect movements like pushups from smartphone sensor data — roughly a decade before "AI in the app" became table stakes.

What I want to use this space for is narrower than "indie dev blog." For a while now I've been building apps with agentic workflows. Claude and a handful of specialised agents do a serious chunk of the work; I'm the one steering, setting the constraints, and reviewing the output. It has changed how I ship software more than any tool I've picked up in the last ten years — Evomo's deep-learning pipeline included.

So this is where I'll write about that. Concrete things — how a feature in Movr or Fernweh actually got built end-to-end with agents in the loop, the tooling choices that paid off and the ones that didn't, the harness I've built around the agents to keep their output trustworthy, and what other engineering leaders should expect from this in 2026.

Plan is to build in the open.

← Back to writing