In 1939, an advertising executive named James Webb Young wrote a short book to explain to a class of students how ideas actually worked — not the romantic version, but the mechanical truth. A Technique for Producing Ideas is 48 pages. It makes one argument: every new idea is a new combination of existing elements. Originality is not something that arrives from outside. It emerges from the friction between things you've already collected.

Young meant this structurally, not metaphorically. The mind works by associating things. The more raw material you've accumulated, the more associations are possible. The richer and more varied your inputs, the more surprising the combinations you can produce — and the more likely you are to arrive at something genuinely new rather than something that looks like everything else.

The book stayed in print for decades. Advertising agencies handed it to new copywriters as standard orientation reading. David Ogilvy recommended it. Eugene Schwartz kept a copy on his desk. For eighty years it was considered foundational wisdom for anyone doing creative work professionally.

Then AI changed how creative work happens, and Young's argument became more practically important than it had ever been.

What AI makes visible about creativity

Language models are, at their core, combination machines. They predict likely next tokens based on patterns across enormous bodies of text. They can recombine elements at a scale and speed no human writer could approach.

But they recombine what they have. And what they have is the statistical distribution of their training data — a composite of a very large body of published writing, weighted toward whatever appears most frequently. The output gravitates toward the center of that distribution. The words are appropriate, the structure is correct, the result is professionally adequate and generically forgettable.

Young's argument predicts this exactly. If you want distinctive output, you need distinctive input. Prompts can direct a model, but they can't change what it's drawing from. The raw material is fixed at training time, and the raw material is an average.

Young's five-step process

Young described creativity as a process with identifiable stages. The first two are what most people skip.

The first stage is gathering raw material — both specific research on the problem at hand and what Young called a "general education," meaning a lifetime of curiosity-driven collection. The second stage is turning the material over in your mind, letting different elements come into contact with each other. The third — the famous stage that gets most of the attention — is the moment of combination, which Young argued happens on its own if the first two stages have been done properly.

Most people treat AI as if it starts at stage three. Write a prompt, get an idea. But Young's framework says stage three depends entirely on stages one and two — on the quality and variety of what's been collected, and on how much time has been spent letting different elements associate.

The swipe file is Young's stage one, applied specifically to your domain. It's the practice of collecting, over time, the examples that your judgment identifies as worth keeping.

The kaleidoscope metaphor

Young chose the kaleidoscope to explain what happens when raw material combines. The same pieces of colored glass, rearranged, produce infinite new patterns. The quality and variety of the pieces determines what patterns are possible.

He was writing about human creative work in 1939. He was also, without knowing it, providing the most accurate description of what AI needs to produce work worth reading. The AI is the rearrangement mechanism. Your swipe file is the glass.

When you connect a curated swipe file to an AI assistant through a tool like Kaleidoscope, the model isn't drawing from its training average anymore. It's drawing from the specific examples your attention selected as worth keeping — the headlines that landed, the copy that converted, the campaigns that made you stop. The pattern space changes because the pieces change.

Why Young's book is as relevant today as it was in 1939

Young wrote before computers, before the internet, before anything resembling AI. His subject was human creativity in an advertising context. What he identified, though, was something structural about how new ideas emerge from existing material — a principle that holds regardless of whether the combination mechanism is a human brain or a language model.

The insight that's easy to miss: the mechanism matters less than the material. A sophisticated model with thin inputs produces thin output. The same model with a rich, specific, well-curated set of examples produces something meaningfully different.

Young figured this out in 1939 by watching how the best copywriters worked. They collected obsessively. They kept their collections organized and accessible. They didn't sit down to write a headline from nothing — they sat down with years of accumulated material and let the combinations emerge.

That practice is still the answer. It's just more urgent now.