Code Got Cheap. Judgment Didn’t.
Watch a capable AI agent build a working app in an afternoon — complete with a database, a login screen, and a few working features — and it’s tempting to conclude that software engineering has quietly become a solved problem, something any team with a laptop and a subscription can now do. That conclusion is wrong, but not because the demonstration is fake. The demonstration is real. What it leaves out is everything that happens after the demo: the part of software work that decides whether what got built is actually correct, safe, and worth using.