Six months in, somebody finally asks what happens when a customer cancels an order that's already half shipped. It's a fair question at the worst possible time. Before anyone wrote code, one sentence would have closed that gap. Now it's a migration and three teams who thought they were done. That's the cost-of-change curve.

Nobody draws the other curve: what it costs to go looking for gaps in the first place. It's another review, another round of what-ifs, another expensive person asking what did we miss. Every team stops asking and starts building well before the questions run out. You don't run out of gaps. You run out of hours.

AI can change what looking costs. It won't do that on its own, and getting it to is most of the job.

So we'll walk the loop: how to run it, which gaps get expensive later, and when to stop. Edge cases reveal themselves eventually, usually after the code is written, or worse, after it's in production. Or you look for them first. The best time to catch an edge case is while it's still a sentence.