When I started using AI, I made a lot of things that seemed useful at the time. Most of them I never used again.
Successfully executing an idea used to be the advantage over your peers: understanding how to get things done, faster than anyone else. Then execution got cheap. Now if you need a dashboard, you don’t go find a business intelligence analyst to build it for you—you build it yourself.
But expensive execution had a positive side effect. Three weeks of engineering time; a design team with no capacity; budget you don’t have… Those constraints were annoying, but they usually made people sit down and ask some important questions: Should we do this? Is it worth it?
AI has removed constraints, which is generally a good thing. But it also removed the moment that used to force us to ask whether the next thing was worth doing. Until recently, we couldn’t do everything, so we had to choose what got done and what didn’t—we had to make a decision, at some point, to stop.
Now it’s “Claude, add this.” “Claude, build that.” “Claude, what if we also build this?” And so you keep going.
What comes out of that isn’t AI slop in the ugly generated-content sense. It’s slop by accumulation. Too many features, too many workflows, too many automations, too many things that seemed individually reasonable to build, but en masse drown a product in needless nice-to-haves.
The risk of cheap execution therefore isn’t bad execution. It’s endless execution.
As AI gets better at executing our ideas, we’ll need to become ruthless at assessing our own work. The question isn’t complicated—is this a must-have or a nice-to-have?—but nothing is going to force you to ask it anymore. We’ll need to learn to question whether the next thing is actually worth doing, and stopping when the answer is no.
