Notes on how good software actually gets built.
Field notes from our delivery work: the problems every engineering team runs into, and what we built into ADIS to handle them.
What walks out the door when your senior engineer quits
Every team has one: the person who knows why the payment service retries three times, which module nobody touches on a Friday, what that undocumented flag really does. None of it is in a repo. It's in their head, and it leaves when they do. What it costs to keep it there, and what to do instead.
Nobody knows when the documentation was last true
Ask a team for their requirements docs and you get a link. Ask when it was last accurate and the room goes quiet. Why the docs drift, what the gap costs once people start making decisions on it, and what changes when you derive the description from the code instead.
Everyone has their own AI now. That's the problem.
Every role got its own AI, everyone got faster, and projects ship on the same dates as before. The reason is structural: each assistant runs in its own context window and can't hand anything to the next one. What we changed so the work carries forward.
Nobody had time to read four thousand alerts
Security usually gets a frantic once-over the week before release, plus a pile of scanner output nobody opens. Why that keeps happening, what it costs to find out late, and how we made the scanning continuous and the output short enough to read.
Technical debt never sends an invoice
It arrives as a release that slips by a week, a change everyone called simple taking three days, and a roadmap that keeps getting shorter. Why debt compounds, where the real cost hides, and how we keep it on a list with owners and dates on it.