Last week an agent wrote a feature for Phora in eleven minutes. Same morning, I spent forty minutes naming a single component.
Both decisions were on purpose. The agent saved me half a day. The naming saved me from a future I didn't want — one where every part of my product sounded like a model wrote it, because one did.
The interesting question moved
Two years ago the question was can I automate this? The answer was usually no, and that made it feel important when it became yes.
Now the answer is yes for almost everything. Which means the question changed. The interesting one is should I? — and more specifically, what do I lose if I do?
For the boring parts the loss is zero. For the parts of the work that carry taste, voice, or judgment, the loss is the work itself. You don't notice it the day you automate it. You notice it three months later, when nothing you make feels like yours anymore.
The list keeps getting shorter and more specific
I let agents do the obvious things. Refactors. Boilerplate. Tests for code I already wrote. The first pass at a function I know the shape of.
I don't let them do the parts where I'm still figuring out what I think. Naming. The first version of a piece of copy. The decision about what not to build. A reply to a frustrated client. The opening paragraph of an article like this one — if a model writes it, the whole article goes nowhere, because I never had to find the angle myself.
This isn't about effort. Most of these are quick. It's about who's doing the thinking. If I outsource the thinking, I get faster output and a slower brain. Bad trade.
Speed is becoming a commodity
The narrative right now is: automate more, ship more, win. Mostly true, and mostly boring. Everyone has the same agents. Speed isn't the moat it pretends to be.
What isn't a commodity is the thing only you would have made — the small choices a model wouldn't make because they're not statistically optimal. Those choices live in the parts you decided to keep in your hands.
Pick the shortlist. Defend it. Automate the rest.
There's always a next level.
If you like what you see — whether you're building a product or a team — I'd love to hear about it.