Don’t Be AI’s Editor
Of course you should be using AI, but not as your first step. Do this instead.
I go back a long way with AI, producing some of the first automated narratives for Automated Insights back in 2010.
But I’ve only recently begun defacing my deliverables with the statement: “No AI was used to produce this document or the content within it.”
That doesn’t mean I don’t use AI. It means I refuse to be AI’s editor. Because I’ve seen it get a lot of shit wrong and I know that my clients, bosses, and editors are counting on me to be truthful and accurate in the places where they’re not experts.
For example, several years ago, when I was saying, “LLMs aren’t thinking. They just chunk word parts together in a way that instantly predicts the next most accurate word part,” my clients and bosses and editors were pretty skeptical and kept asking me if I was sure.
Yeah, I said. Replace “word part” with “action” and you’ve encompassed the entirety of AI.
To their credit, they trusted me. These days they have a sharper perspective, because the truth is no longer obscured behind a wizard’s curtain.
So I ask you. How much do you trust Claude to speak to those things you admittedly don’t know much about? How many times has he failed? I mean, I know it’s a lot, but did those failures leave scars painful enough for him to learn real lessons instead of just learning to mask the behavior?
I’ve got scars, in some cases real physical ones. Those scars are why we pay more for human experience. This is what we’re losing with the rush to AI.
We’re all just editing whatever AI spits out now. And that’s backwards. Here’s why.
AI Is a Gambling Machine
Let me explain. All decisions that are based on quantified, data-driven, predictive analysis will always require at some point down the chain relying on the wisdom of the crowd.
If you want to rely on the crowd to gamble with your job, your career, and reputation, first let me explain how gambling spreads work.
You might think that in 2025, with all the data we can and do collect around athletes and the sports they play, that gambling spreads — those data points that predict whether one team or the other will win the game and by how much — would crunch all that data and produce the most likely outcome.
They don’t. Because that’s not the purpose of gambling spreads.
Gambling spreads are initially determined by some old guy who has a knack for predicting the wisdom of the crowd. Then the sports books — the places that let you gamble — follow the money as it comes in and adjust those spreads by what the crowd is telling them it thinks. It’s how the books maximize the money they make whether one half of the crowd wins or the other half wins.
One of them once put it to me this way: “Well, what do the idiots think?”
It’s all wisdom of the crowd. It’s polling data at scale. This is why those spreads sometimes seem remarkably accurate, but it’s also not super rare for major upsets to happen across the sporting world. And those not-super-rare instances — when the sure thing loses — are when addicted gamblers get their legs broken.
An AI user is not relying on intelligence. They are relying on consensus.
But for whatever the reason, we’re all fine with that.
I care about you. Here’s how to keep your legs intact.
Don’t Use AI To Produce
Like I said, I go back a long way with this science.
I’m not just talking about using AI to write your school paper or your TPS report — or having it mess with your spreadsheets or summarize your support calls. I’m also talking about machine learning and predictive analytics all the way up to self-driving cars, bank fraud, and military robots.
But the school paper and TPS report scenario is the easiest to understand, so let’s call it a metaphor.
Don’t use AI to produce your school paper or TPS report.
Yeah, I know. It’s not cheating if everyone is doing it.
I’m not talking about that. I’m talking about a major upset.
In this case, the “major upset” is AI getting a foundational fact completely wrong and building a hell of an eloquent, leak-proof case around it — with charts and references and quotes and everything.
And just like with major upsets in sports, the crowd never sees these coming and there’s nothing they can do to prevent it. And I know the likelihood is low, but it’s also, by its very nature, a likelihood completely proportional to the lack of knowledge one has in the subject — which is why one might turn to AI in the first place.
And because AI is nothing more than the wisdom of the crowd — polled really fast and even using a lot of context — this is not a problem that’s going to be fixed, no matter how many PhDs spend billions of dollars training a model on millions of GPUs.
It’s baked into the tech.
So how should you be using AI instead?
What Do the Idiots Think?
Simple. Produce your work first and see how well it holds up against the wisdom of the crowd.
AI is great for catching concepts and counterarguments and outliers you might have missed. It’s awesome for addressing doubters, naysayers, and haters. It’s perfect for getting a birds-eye, panoramic view of how someone else might have done your assignment, whether that’s writing about the War of 1812 or predicting next fiscal year’s revenue growth or figuring out a hacker group’s next target.
When I gamble on sports, it’s almost always on the sports I follow closely. I keep it fun, but I do my own research, based on decades of experience playing, following, and wagering on the sport.
Yeah — wagering — there’s nothing wrong with learning from the wisdom of the crowd.
Then I look at the spread for the game, to see what the crowd thinks, and sometimes I’m a little off, and then I do a little more research — “Oh, star player guy is hurt. Didn’t know that. That changes things.”
In a school or work context, it’s working on the assignment, producing your result, and checking it against the wisdom of AI. Hell, you can even data-ify your own work and throw that into Claude and see if it finds any holes.
It’s not having Claude write your code, skimming it, and rolling it to production.
Learn from the idiots. Don’t be AI’s editor. Make AI be yours instead.
If you learned something from this, or even if you didn’t, please join my email list and get a heads up when I’m published.
canonical=https://www.inc.com/joe-procopio/dont-be-ais-editor/91206885