I rewrote the same function nine ways in an hour and felt productive the whole time. I had shipped nothing.
The first version worked. I knew it worked because the tests went green. But the assistant offered me a cleaner version, and the cleaner version suggested a more functional version, and the functional version made me wonder what the recursive version would look like, and somewhere around the seventh I was comparing two approaches that differed only in whether I liked the variable names. Each one took ninety seconds to conjure. None of them was wrong. That was the problem.
There's a kind of restaurant that ruins people, and it's not the expensive one. It's the buffet. At a buffet the price of trying a dish drops to zero, so the discipline of ordering disappears. You take a little of everything. You go back for the thing you liked. You end up with a plate that is a survey rather than a meal, and you leave full but unsatisfied, having tasted twelve things and finished none of them. The cost of sampling fell, and somehow that made you worse off.
For most of my working life, choosing was expensive, and the expense was the teacher. If building a prototype took two weeks, you thought hard before building it. You argued. You sketched on whiteboards until the cheap medium had wrung out the bad ideas, because the next medium was costly. Scarcity did your editing for you. The price of the meal forced you to decide what you actually wanted before you sat down.
That price is now approaching zero, and it happened faster than most of us noticed.
The buffet opened
By late 2024, generating an option costs almost nothing. Ask for a landing page and you get three. Ask for a data model and the machine will draft it, critique its own draft, and offer two alternatives in a single sitting, faster than I can read them. The marginal cost of an alternative has collapsed, and we are behaving exactly like people who just walked into a free buffet — loading the plate, sampling everything, finishing little.
I want to be careful here, because the cheapness is genuinely good. I am not nostalgic for the two-week prototype. Cheap iteration is how you find the idea you couldn't have specified in advance. The buffet is not the villain. The villain is the part of us that mistakes a full plate for a finished meal.
I have a particular weakness for this, and I should own it. My mind is wired for the chase — the next option always glitters more than the one in front of me. For decades that pull cost me, because acting on it was expensive; chasing a new idea meant abandoning weeks of work, so reality punished the impulse and I learned, slowly, to sit still. The collapse in cost has quietly removed that punishment. Now I can chase nine rabbits before lunch and call it research. The environment stopped charging me for the thing I most need to be charged for.
What got scarce
Here is the part worth keeping. When the cost of trying an option falls to nothing, the option is no longer the scarce thing. The scarce thing becomes the willingness to stop sampling and commit to a single plate — to pick one path and stay on it long enough to find out whether it was any good.
This is not a small reshuffling. It inverts the skill that mattered. The old discipline was selection under scarcity: choosing well because you only got a few shots. The new discipline is commitment under abundance: choosing to be done choosing, when nothing in the environment forces you to. Saying no to eleven good dishes is harder than affording one, and we have built almost no muscle for it, because we never had to.
You can see the failure mode everywhere once you have the word for it. The team that has explored six architectures and built none. The document with four competing intros, each defensible, none chosen. The product that A/B tests its way into having no opinion. These are not laziness. They are what happens when the friction that used to end deliberation gets removed and nothing replaces it. Deliberation, left to itself, does not terminate. It was never supposed to. The bill used to arrive and end it.
Bringing the bill back
So the work now is to bring the bill back on purpose — to reintroduce, by choice, the scarcity that reality no longer imposes. Decide in advance how many versions you'll generate, and ship the best of those rather than the best imaginable. Time-box the buffet. Treat the ninth alternative not as diligence but as the thing keeping you from eating. The constraint that used to be a cost is now a tool, and you have to install it yourself, because no one will charge you for skipping it.
I deleted eight of my nine functions and kept the first one. It was never about which function was best. Past a certain point they were all fine, and "all fine" is precisely the trap — when every option clears the bar, the bar stops helping you choose, and only commitment does.
The future of this work belongs to people who can do the unglamorous thing the machine makes harder by the month: pick a plate, finish it, and walk away from the rest of the buffet still hungry. The cost of a decision fell to zero. The cost of finishing one didn't.