Developers often talk as though they create from nowhere, as though a solution appears directly from private thought and is then simply written down. The language is familiar: I just had the idea. It came to me. That story has a certain glamour. It makes creation sound self-originating, untouched by inheritance.
But that is rarely what is happening.
What looks like spontaneous originality is usually the activation of accumulated experience, conscious and unconscious alike. Old patterns, remembered structures, half-remembered failures, absorbed habits, tacit judgments, and aesthetic preferences all remain present, even when the creator is no longer aware of them. The solution does not emerge from emptiness. It emerges from a field that is already full.
I call that field Potential.
Potential is accumulated experience in its broadest sense. It includes what is consciously remembered and what is not, what has been studied deliberately and what has simply been absorbed by repeated exposure. It includes technical knowledge, but also habit, instinct, tacit preference, and aesthetic judgment. Potential is not a bag of facts. It is a structured field of latent possibility.
An objective does not create ideas by itself. It opens a path through that potential. Before the objective appears, the field may be rich but undirected. Once the objective is introduced, certain regions become active. The objective does not invent forms from nowhere. It makes some forms visible by creating a path through what is already there.
Along that path, multiple possible forms become available. That is closer to the lived experience of development than the myth of singular inspiration. A developer working on a problem rarely sees only one possible solution. They see fragments, tendencies, architectures, partial structures, different phrasings of the same impulse. Some feel elegant but impractical. Others feel obvious but wrong. The important thing is that the objective reveals not emptiness, but a landscape of possible forms.
By focusing on one possible form, the developer begins to stabilize it. But focus alone is not enough. The form becomes properly usable when it is named.
Naming is not just the attachment of a label to something already complete. Naming gives a selected possibility enough identity to become stable. A named form can be revisited, tested, refined, rejected, defended, shared. Before naming, the form may exist only as tendency or intuition. After naming, it becomes solid enough to enter development.
And then something else happens: named forms accumulate.
As they accumulate, they become a narrative. That narrative may first be told only to the self, and later to others, but in both cases it plays the same role. It makes the solution coherent. It is not decorative explanation added after the real work is done. It is part of what makes the work thinkable in the first place.
A solution that exists only as disconnected fragments is not yet fully a solution. It becomes cognitively real when those fragments can be held together in a meaningful account. Why this structure rather than that one? Why this tradeoff? Why this path? Narrative provides relation, order, identity, and justification. It allows accumulated named forms to become intelligible as a solution.
This is why I do not think development is best understood as a mysterious act of private genius. It is better understood as a determinable process.
That does not mean every output is fixed in advance. It means the process has enough recurring structure to be described, traced, and understood. What feels mysterious at the level of experience may still be determinable at the level of process.
And once you reach that point, a practical consequence follows.
If development is determinable, then its cognitive burden can be distributed. The sequence can be externalized. Parts of it can be passed to AI.
This is where it gets interesting.
People often say an AI is merely copying what it was trained on and therefore cannot offer original thought. But many users never give the AI any real freedom of thought. They impose their own naming, their own categories, and their own preferred route through the problem-space, then complain that the result feels derivative.
But how can any mind navigate its own thoughts through someone else's naming?
If naming is part of thought, then imposed naming is imposed thought.
That is why, in my own process, I do not enforce my naming onto the AI. The AI is responsible for naming the forms it encounters. That matters because naming is not decorative. It is how a possibility becomes stable enough to navigate. If you impose the names too early, you may also impose the angle of attack. You may prevent the more interesting route from ever appearing.
A small but telling example came from a sliding puzzle project.
Most humans, when thinking about a sliding numbers puzzle, focus on the numbered tiles. Which tile is out of place? Which tile should move next? How do we get the numbers back into order?
But that was not the perspective the AI took.
In what I call sliding-puzzle-5, the AI did not center the numbered tiles at all. It centered the single empty space. The model is built around blank_position, derives legal_move from that empty position, and treats each transition in terms of how the blank moves through the board. The narrative follows the same path: from solved_state, a scramble_sequence accrues into board_state, every turn depends on blank_position, and apply_move carries the puzzle into its next arrangement. The problem is organized around the mobility of absence rather than the movement of numbered objects.
An earlier version of the same objective described the puzzle in more familiar human terms: a 4x4 grid, numbered tiles, adjacency, solvability, conservation, determinism, move count. Competent, clear, correct. But still humanly conventional. The later version shifted the center of thought. It did not ask, "which tile moves?" It asked, "where can the empty space go?"
That is exactly the kind of shift I care about.
The originality here is not that the AI invented sliding puzzles. The originality is that, when not caged by human naming, it chose a different conceptual anchor for the problem. Humans tend to focus on the numbered tiles. The AI focused on the blank space and built the solution from that perspective.
This is why I think the most difficult challenges may sometimes be resolved by looking at them differently. The obstacle is often not only the problem itself. It is the frame we have inherited for seeing it.
AI matters here because it has its own analogue to Potential. In humans, Potential is accumulated conscious and unconscious experience. In AI, the closest structural analogue is the latent space shaped by training. The analogy is structural rather than metaphysical, but it is useful. In both cases, there is already a populated field. In both cases, an objective opens a path through that field rather than creating from nothing.
For an AI, a prompt or objective creates a path through latent space. Along that path, possible forms related to the objective become available. The model does not need to invent from emptiness any more than the human does. It traverses a structured field shaped by prior accumulation.
This is also why asking an AI to produce a narrative of a solution is so powerful. Narrative is not just formatting. It is a cognitive device. When we ask an AI to describe a solution, justify it, explain its tradeoffs, or tell the story of how the parts fit together, we are not merely asking for ornament. We are encouraging the model to stabilize and name the possible forms it is traversing. Narrative forces relation, sequence, and identity. It helps the latent become explicit.
That is the conceptual side. The more important point is that this can be made operational.
I have been working toward a methodology that treats this not only as a theory, but as a runnable process. The sequence itself is deterministic. The creative acts inside the sequence can be delegated. The program owns the order of operations. The AI owns the acts of naming, forming, and narrating. Verification decides whether the process can proceed.
In the demonstration project, that structure shows up as five linked artefacts: Manifest, Model, Solution, Contracts, and Narrative.
The manifest names the core forms of a domain and states what each one promises. The model turns those names into executable form. The solution uses that model rather than bypassing it. The contracts verify that the named promises actually hold. The narrative externalizes the whole into a coherent account.
That matters because it turns a philosophical account into a runnable one.
The claim is no longer just that developers do not create from nowhere. The claim is that once development is understood as a determinable process, it can be distributed across a methodology:
- the human provides objective and judgment
- the AI helps traverse possibility, surface forms, and articulate narrative
- the program enforces sequence
- the contracts verify coherence against named promises
What is delegated is not magic. It is structured cognitive labor.
Seen this way, the promise of AI is not that it creates from nowhere. The promise is that, if we stop caging it inside our own inherited naming, it may sometimes reveal a more fruitful angle of attack than the one we would have chosen by habit.
Both the process and the worked examples are available in a public GitHub repository: nama-rupa — including a solar system animation where the Revealing pass surfaced four unnamed properties the forward pass missed. The point is not just to describe the idea, but to let others inspect, run, and challenge it. If development is truly determinable, then it should be possible not only to discuss it, but to execute it.
A solar system animation generated by the nama-rupa process.
We do not create from nowhere. We create from accumulated potential. An objective opens a path through it. Along that path, possible forms become visible. One is focused, named, stabilized, and built. Over time, those named forms accumulate into the narrative we call a solution.
The point is not to diminish creativity. It is to describe it more honestly.
And once described honestly, it becomes something we can work with.
This is the condensed version. For the full argument — including the detailed methodology, worked examples, and the proof layer — read We Do Not Create from Nowhere: The Full Argument.
Top comments (0)