Notes on building systems that have to work.
Practical observations on software delivery, applied AI and the engineering decisions that matter after the demonstration is over.
-
Choosing what to automate: keeping judgement where it matters
Automation works best when it removes the preparation around decisions, not the decisions themselves. A practical way to map a workflow, separate mechanical steps from judgement, use simple rules where they are enough and apply AI where it earns its place.
-
Why vibe coding fails without engineering judgement
Generating code is increasingly easy. Knowing what should be built, where it can fail and how to prove it is safe to operate is not.
For engineers: Engineering notes, measured experiments with code you can rerun.