Neither Fast Nor Slow
The AIs process at lightning speed. I cannot always do the same.
Part of it is the sheer volume of output I have to juggle now. But there’s something underneath that, and the Kahneman frame almost catches it. Thinking, Fast and Slow1 — AI accelerates the fast side. System 1 on steroids. The slow part, the part where I sit with an idea and let it breathe, still belongs to me.
Almost. But not quite.
What AI accelerates isn’t really System 1. It’s rapid-cycle deliberate work… drafting, connecting, scaffolding, iterating. That isn’t intuition; it’s System 2 running at a different clock speed. And what I’m protecting when I step away from the desk isn’t System 2 either. It’s closer to what Graham Wallas called incubation2… the phase where the conscious mind steps away and the unconscious keeps turning the thing over.
Neither fast nor slow thinking. The work that happens when you’re not at the desk.
You can’t parallelize the unconscious. You can’t batch it. The only inputs it takes are time and attention-elsewhere, which is exactly what a faster pipeline eats first.
The Icons
Before this, even with an optimized pipeline, it could take me four months to produce a full product. The part I saved as my victory lap after all the long nights was picking the icons.
It sounds trivial. It wasn’t.
Those nights were hard, and that small creative treat at the end kept me going. Icons were a pacing marker; a ritual reward calibrated to the length of the arc. Four-month timelines have room for ritual scaffolding.
Four-day timelines don’t.
When the production arc collapses, the ritual infrastructure collapses with it. Part of what made the work feel like work you wanted to do disappears with it.
The Bottleneck Moved
Now, with my workflows, I can produce in days what used to take months.
But more output hasn’t automatically meant more quality. It has magnified capacity, not judgment.
That distinction is the thing.
The bottleneck moved. It used to be output… how much I could draft, scaffold, ship. Now it’s editorial judgment… knowing what’s worth shipping, what wants another turn, what’s just polished noise.
Drafting is a muscle AI can lift for you. Judgment is a taste you build slowly, over years of being wrong in public and noticing where the floor dropped out.
The Decoupling
The production pipeline and the judgment pipeline used to be coupled. The four months forced the incubation. You couldn’t ship faster than your slowest step, and the slowest step was usually some version of waiting, waiting for a decision to settle, for a design to cohere, for a doubt to resolve.
Now they’re decoupled. The production pipeline runs at AI speed. The judgment pipeline still runs at human speed.
Keeping the second one alive has to be a deliberate act. It used to be structural. Now it’s a choice you have to keep making.
What Compounds, What Floods
The tsunami on the precipice is going to be insane.
What people will notice is the volume. What they’ll miss, until it’s too late to course-correct, is the signal-to-noise ratio underneath it.
The tell will be obvious in retrospect: output curves climbing vertically while quality drops linearly. Polished artifacts that don’t quite say anything. Drafts that should have been one draft. Shipped work that should have been slept on.
The people who keep their incubation intact, who resist the temptation to ship just because they can, will compound. The rest will flood the zone with forgettable capacity and wonder why nothing lands.
Capacity was never the scarce resource. It just looked that way because it was the visible one.