Why I never tell AI what to do
Stop giving AI instructions. Describing the problem and asking questions first produces better results — the AI knows patterns, you know your context.
I’m Manuel Holzrichter — a software developer with over 15 years of experience across industry and government projects. I write about software architecture, automated testing, and what makes developers truly effective, including how to work well with AI. Practical insights from real projects, not textbook theory. More about me.
Stop giving AI instructions. Describing the problem and asking questions first produces better results — the AI knows patterns, you know your context.
External APIs will change without warning. How adapters and hexagonal architecture keep foreign systems at the boundary of your codebase.
Software projects rarely fail on technical work. They fail on long feedback loops, unvalidated trust, and gut-feeling estimates — all three fixable.
Compose micro frontends at runtime with import maps — a lightweight, browser-native alternative to Module Federation. No page reloads, no iframes.
Assume you are wrong by default: why verification built into your workflow beats confidence, and how short feedback loops catch what you cannot see.
A step-by-step guide to migrating a frontend monolith to micro frontends: extract a shared kernel, then split the first module into a standalone SPA.
Why every developer should build a personal website: practice explaining technical work to non-technical people, own your content, grow your career.
Why writing is a developer’s most underrated skill — clear prose sharpens thinking, spreads knowledge, and becomes a superpower in the age of LLMs.
Why small, iterative steps beat big-bang delivery: lessons from growing into an agile developer, one short feedback loop at a time.
How automated tests transformed twelve years of software development — a personal retrospective on quality, confidence, and lessons learned.