Close the loop: verification and heartbeat for AI coding agents
Agents must verify their own work — but the shape of that loop depends on what you're building. Solo work versus team-scale, runners versus workflows.
Two decades building software and the teams that build software — currently in digital healthcare, working on what AI changes about the craft. This is where I publish what I learn. Field notes on agentic coding, high-performing teams, and what "done" means when the agents write most of the code. Occasionally in Swedish, when the story demands it.
Agents must verify their own work — but the shape of that loop depends on what you're building. Solo work versus team-scale, runners versus workflows.
AI-codingExecutives want AI gains. But AI is a force multiplier — it multiplies whatever you have, including your problems. Why Spotify and Stripe succeeded, what most companies miss,...
ai-codingA practitioner's system for delegating to AI coding agents. The workflow, the rulebook, the mistakes, and why 80% of the work is scoping and verification — not writing...
ai-codingThere are four stages to how developers work with AI coding tools. Most of us are stuck at stage three. Here's the full progression — and what infrastructure...
I've spent the better part of two decades building software and the teams that build software — currently leading engineering at a Swedish healthtech preparing for IPO, where the fun part is shipping AI into healthcare without breaking the things that make it trustworthy.
This site is where I think out loud. Expect field notes on agentic coding, post-mortems on interesting failures, and the occasional Swedish-language piece when the story demands it. Not a content mill. No newsletter pop-up. Just writing, filed when it's ready.
Governed AI execution platform for regulated enterprises. Open source community edition launching H1 2026.
Engineering management platform translating GitHub activity into actionable insights. Born from managing 25+ engineers. Launching 2026.