Human creativity is not a prompt
Instant generation gives you an output. It does not give you the thing that made the output worth having — and the difference is the whole game.
Watch how a real body of work comes into being and you will notice it almost never arrives at once. It accumulates. Someone keeps returning to the same blue. A motif from an abandoned project resurfaces, transformed, three years later. A half-remembered film bends the way they frame everything afterward. The work is downstream of a long, private process of exposure, contradiction, memory, and revision — most of which never gets articulated, and almost none of which looks like typing an instruction and receiving a result.
This is the part current AI tools are structurally blind to. Not because the models are weak — they are astonishing — but because the interaction is built around a single transaction. You arrive, you ask, you receive, you leave. The system has no idea who you were yesterday. It cannot tell whether the thing it just made for you is consistent with everything you've ever loved or a betrayal of it. It has no stake in your trajectory because it has no record of one.
A prompt is a request. A creative practice is a relationship with your own evolving taste. We have built excellent tools for the first and almost nothing for the second.
What the transaction model optimizes away
The dominant design goal across AI creative products is to compress the distance between intention and artifact. Faster, smoother, fewer steps. As an efficiency play this is unimpeachable. But efficiency assumes the valuable thing is the artifact at the end. In a creative practice, a great deal of the value is in the distance itself — the friction, the detours, the time spent not-yet-knowing. That is where taste is formed and where work becomes specific to a person rather than generic to a prompt.
Collapse the distance to zero and you get output that is competent and anonymous. It resembles the average of everything the model has seen, because nothing in the interaction pulls it toward the particular gravity of one human being. The result is the strange flatness people already sense in a lot of generated work: technically fine, emotionally placeless.
What it would mean to participate instead
Imagine the opposite design. A system that remembers — not your files, but your references, your moods, the concepts you abandoned and why, the motifs that keep returning, the things you called derivative and the things that moved you. A system that, over months, builds a model of where your creative gravity is moving and is willing to say so. One that introduces a tension instead of resolving one, surfaces a reference you didn't ask for, names a contradiction in your work you hadn't seen.
That is a different relationship to the machine entirely. Less vending machine, more interlocutor. Less "what style do you want?" and more "this is the third time you've circled this idea — I think it's the real one." The measure of success stops being how quickly it produced something and becomes whether, over time, you became a more coherent and more fully realized version of the artist you were already becoming.
The goal is not to make artists faster. It is to help them arrive.
Why this is the harder, more important problem
Building this is genuinely difficult — it requires persistence, memory that means something, multimodal continuity, and a model of taste that evolves rather than resets. It resists the clean metrics the industry currently rewards. You cannot benchmark it on output-per-second. Its payoff is measured in months and years, not seconds.
But I think it is where the value is heading. As execution becomes commoditized — as everyone can generate competent output instantly — the scarce thing becomes judgment: taste, direction, originality, emotional coherence, the orchestration of a sensibility over time. A system that helps a human cultivate exactly that is not a faster paintbrush. It is closer to a second mind. That is the category worth building, and it barely exists yet.
Human creativity is not a prompt. It is an evolving internal landscape. The interesting question for the next decade of AI is not how to generate inside that landscape faster — it is how to participate in it. That is the question Strong AI exists to work on.
Essay No. 001 · If this resonates, read the manifesto or write to me.