Capture: Treasure In, Treasure Out
The quality of AI output depends entirely on the quality of your inputs. Here's how to mine your own expertise for source material.
"Garbage In, Garbage Out"
— Old programming adage
"On two occasions I have been asked, 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' ... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question."
— Charles Babbage, Passages from the Life of a Philosopher
How can you tell the difference between an article written by a human and something churned out by AI in seconds? If there are typos, it's probably of human origin. When it comes to basic writing mechanics and even certain stylistic elements, AI has gotten good.
How good? Joshua Lisec, one of the most talented and sought-after ghostwriters of our generation, observes that AI is already better than 95% of writers—yes, even the good ones.
For writers to stand apart in this new landscape, Lisec's advice is clear: you have to be "so. much. better. than a good, cheap, fast content generator any schmuck can use over a few days to write, edit, and publish the next great American novel."
The Missing Variable
Yet, Lisec also recognizes that while AI might be capable of producing a high volume of passable content (with impeccable grammar), it is still missing a certain "X factor."
To illustrate his point, Lisec undertook an experiment using various AI tools to write a hypothetical book chapter. When prompted, the AI could only regurgitate the input Lisec gave it—a slightly reworded version of his prompt, padded with fluff. Lisec then rephrased his prompt to inject more inner conflict, and the AI invented a fictional struggle while still failing to add any meaningful details or novel information.
Lisec concludes two things from his experiment: first, that the output is only as good as the input, and second, that you cannot use AI to write a book.
I agree with the first part. Programmers have a phrase for this: "Garbage in, Garbage out." Flawed or nonsensical data inputs lead to nonsensical outputs.
However, I disagree with the second assertion. It all depends on how you use the word "with."
AI "Sourcery"
The missing variable in Lisec's experiment was not skill on his part as a writer—he's one of the best—but rather the lack of source material he gave the AI to work with.
This takes us back to the art of AI alchemy, or what I call "Sourcery": the transmutation of one's own experiences and expertise into polished content. Sourcery begins with mining the depths of your intellect for source material and then refining it with AI's help.
Now, let's be clear: Lisec knows how important quality source material is to the writing process. When a professional ghostwriter sets out to write a book for their client, where do they start? By interviewing them. In this way, you capture the raw material—the ore of original thought.
Treasure In, Treasure Out
The gap between AI-generated and the very best human-authored content remains significant. Though mechanically sound, AI tools are mostly limited to the pool of congealed knowledge in their training data—a stale summation of word vectors. Think Wikipedia but written by a robot.
Great writing still requires inputs provided by a knowledgeable creator who can weave them into a rich tapestry based on subtle context that is not available to AI.
Therefore, as AI-assisted writers, we must adopt a new maxim to emphasize the significance of high-quality inputs:
"Treasure In, Treasure Out."
The richness of the material we feed into AI determines the quality of content we craft. Starting with rich, authentic content is key.
What If I Don't Have a Podcast?
If you don't have a podcast, one simple yet powerful method for generating source material is to take a walk with your phone and record your thoughts and ideas.
This act of speaking freely, perhaps while meandering through a park or walking your neighborhood streets, can capture a fresh stream of consciousness born of sunshine and outdoor air.
The dynamic and spontaneous nature of conversation—whether with another person or with yourself—captures the essence of authentic communication and idea generation. Evolutionary biologist Matt Ridley has a term for this synthesis: "Ideas Having Sex." The interplay of thoughts culminates in the birth of a newborn perspective.
Sources of Source Material
Your accumulated archives represent treasure troves of content waiting to be refined:
Podcast transcripts — The dynamic nature of conversation captures authentic idea generation
Voice memos — Stream-of-consciousness recordings from walks or commutes
Meeting notes — The back-and-forth of professional dialogue
Old blog posts — Ideas you've already articulated but never fully developed
Email threads — Explanations you've written that could serve others
Journal entries — Raw reflections waiting to be shaped
The richness of the material we feed into AI determines the quality of content we craft. Without quality inputs, you're asking AI to create something from nothing—and that's not transformation, it's generation. Generation produces generic. Transformation produces gold.
This post is adapted from "Commanding the Page" (2023).