Post

Fine-tuning is the new stored procedure. Logic pushed into the wrong layer. It's fast, of course. No remote lookups. It's as close to the model as you can possibly get. But it's a serious pain to version, to test, or to understand six months later.