Field notes
Perspectives on AI automation, software delivery, and building things that last.
DevOptix · UtrechtThe $350,000 automation that changed nothing
It worked perfectly, shipped on time, and moved no number that mattered. Most automation projects fail before a line of code is written, at the moment someone picks a process because it looks easy instead of because it holds the business back.
Read articleAnatomy of a team agent harness
A chat loop with tools is not a harness. An agent that works for a team needs four primitives most personal harnesses skip entirely: delegation, a learning loop, self-aware capability gaps, and multi-channel presence, plus three platform rules that keep the whole thing governable.
Read articleA skills repository is not an automation platform
Every organisation is building a central repo of AI skills, prompts, and agent definitions right now. It is the right first step, and it has a ceiling nobody talks about. A skill is advice for an AI assistant; the doing still runs on personal access. A maturity ladder for what comes next.
Read articleStop trusting your LLM to behave. Enforce it.
System prompts are requests, not rules. Human review, LLM judges, and guardrail frameworks all raise the cost of an attack, and none of them is a boundary. Banks did not solve embezzlement by hiring more honest tellers. Give the agent hands, not keys.
Read article