The second drink was strawberry rum, and something the bartender promised was organic. It was not organic. Nothing at a Florida resort is organic except the mold and the corruption, and both of those predate the building code. The glass was sweating harder than a Series B founder at a bankruptcy hearing, beads rolling down the side and pooling on the arm of a chair that cost more per night than a substitute teacher makes in a year.
I was reading the news on a phone screen gone half-blind with sunlight. Sam Altman, boy pharaoh of the token economy, had gone in front of microphones again to muse, with the practiced bewilderment of a man who has never once been told no by anyone who mattered, that AI adoption was running slower than expected. Sam Altman and others seem surprised that people don’t want to live next to data centers or give over all their data so billionaires get richer, and they get nothing in return except promises of being unemployed.
Somewhere in the same feed, Dario Amodei, the frazzled CS student who wandered out of OpenAI like a man leaving a house fire with one shoe, was saying something earnest about safety and responsible scaling, which is a bit like hearing the second largest dynamite manufacturer talk about the importance of earplugs.
The heat made the glass with the rum sweat. The rum made me sweat. Somewhere in the mid-distance a child was screaming about a pool noodle and I thought: that child understands the economy better than anyone in San Francisco. She wants the noodle. She does not have the noodle. The noodle costs nothing but is being withheld by a sibling with more reach. This is the entire AI market in miniature, except the noodle is your data and the sibling is a billionaire who won’t blink on camera.
The AI technical elite trying to comfort us about how they come in peace while attempting to take over the world is a little too much “Twilight Zone” and too little empathy for most people. The misdeeds might not have arrived yet. They are on their way. You could feel them the way you feel weather coming in off the Gulf, that pressure drop behind the eyes. The darkening skies. Another tech titan questioning our intelligence as he reaches into our pocket.
I had been thinking about Everett Rogers.
Rogers is dead and therefore cannot defend himself, which makes him the ideal authority for this kind of essay. He spent decades studying how human beings actually adopt new technologies, and what he found was so boring and so true that the venture capital world has spent fifty years pretending he didn’t exist. The core finding: people adopt innovations through social proof. Not because the technology is good. Not because some billionaire did a TED talk. Because someone they trust, in a life that looks like theirs, showed them it worked for a problem they already had.
Hybrid corn. That was Rogers’ big case study. Hybrid corn was objectively, measurably, inarguably better than what Iowa farmers were planting. Better yields. Documented results. The farmers could literally count the ears. And adoption still took years, because farmers didn’t adopt until their neighbor adopted, and their neighbor didn’t adopt until HIS neighbor adopted, and the whole chain moved at the speed of trust between people who had to live next to each other after the harvest came in.
The tech world looked at that research and learned absolutely nothing from it. Those of us who have detassled corn might have a different feeling for patents on corn, black SUVs trespassing, and guys with names you see on the side of railway cars trying to get your crop burned. Wait, there is a lot of similarity to what hybrid corn resulted in and where AI tycoons want to go.
AI adoption is not moving at the speed of innovation. It is not moving at the speed of compute. It is not moving at the speed of venture capital, which is just cocaine with a cap table.
It is moving at the speed of paychecks.
A person making fifty-five thousand dollars a year does not look at a hundred-dollar monthly AI subscription and see a productivity multiplier. She sees her streaming budget maybe a good chunk of a dwindling grocery budget. The tool has to earn that money back in the current pay cycle, not in some hypothetical career repositioning play that a McKinsey deck promises will mature in eighteen months. She doesn’t have eighteen months. She has until Thursday.
And here is where the whole diffusion model starts to melt like ice in a Florida poolside strawberry daiquiri.
Rogers identified five factors that determine how fast an innovation spreads. Relative advantage: is it clearly better than what I’m doing now? Compatibility: does it fit my existing life? Complexity: can I figure it out without a graduate degree? Trialability: can I test it without risking real money? Observability: can I watch someone like me use it?
AI fails every single one of these tests for the median American worker. The advantage is unclear because nobody has done the job-specific math. The compatibility is low because the tools don’t fit existing workflows without rework that nobody is paying for. The complexity is absurd. Trialability is gated behind subscriptions. And observability is functionally zero because the people finding real value aren’t visible to their peers in any structured way.
That’s not a slow adoption curve. That’s a technology with no viable path to the early majority, which is the only population that matters.
But it gets worse than Rogers ever imagined, and now I needed the second drink after the first second drink. Tone deaf tech bros in Silicon Valley echo chambers sound a lot like toddlers screaming about pool noodles.
Rogers assumed the innovation could be verified. The neighbor’s corn was taller or it wasn’t. The tractor plowed faster or it didn’t. You could count. You could measure. The innovation might be misunderstood, but it could not actively lie to your face while sounding completely confident. With an undead, unblinking stare as if seeing the singularity looking back at you from a seat in front of Congress.
Like a business executive on ketamine or a law school student on Adderall, AI is different in kind. It generates text that looks authoritative whether the content is accurate or fabricated. The output has no visible seam between truth and hallucination. Which means when your trusted neighbor shows you his great AI result, you have no way to verify the demonstration through normal social proof. The corn is either taller, or it isn’t. The AI output might be eloquent garbage. You can’t count the ears because the ears might be potatoes.
That collapses the entire peer-demonstration mechanism Rogers proved was the engine of adoption. Not a crack in the theory. A sinkhole big enough for the profits of Nvidia. The foundation is gone.
And then. AND THEN. You look at who is selling this technology and the whole picture goes from broken to psychotic. Rogers showed that adoption accelerates when opinion leaders are perceived as “like us, but slightly ahead.” The county’s best farmer. The teacher down the hall who tried the new curriculum first. Someone close enough to your life that their experience reads as relevant evidence about yours.
The opinion leaders for AI are Elon Musk, Sam Altman, Mark Zuckerberg, and a rotating cast of men who name their children like they’re generating passwords. These are people who fire twelve thousand workers by email, build panic bunkers in New Zealand, buy public park land and fence it off from the communities that used it, keep children by various non-wives across multiple time zones, and maintain the unblinking thousand-yard stare of a man who has been told he is a genius so many times he forgot to keep being human.
We’re talking about the Epstein class. Not because they all flew to the island, though the flight logs make for interesting reading. But because they share the same visible contempt for ordinary social contracts. The rules are for you. The consequences are for you. The data is yours until it’s theirs, which is the moment you type it.
A normal person looks at these men and does not think, “I should trust his product.” A normal person thinks, “That is the guy who will automate my job and then post about it on a platform he bought to make himself feel powerful.” We’re talking about the same corporate inhumanity that outsourced apprenticeships to colleges, forced workers to pay for their own training, and then outsourced their work to a third-world nation. We’re so far past trust but verify that we’ve landed in the world of you gotta be kidding me, where is my Sawzall territory of corporate brinksmanship?
This is not an adoption barrier. It is an anti-adoption engine. The faces of the technology are actively repelling the population that would need to adopt it. Rogers never modeled that because it had never happened before. No prior innovation was championed exclusively by people the target market found viscerally repulsive. Of course, I half expect the titans to find influencers to help them, but from what I can tell, even Gweneth decided Sam was radioactive.
And we haven’t even gotten to the part where they rob you.
A hundred dollars a month. That is what the premium tier costs. A hundred dollars a month to upload your original thinking into a training pipeline you cannot audit, to have your unique analytical frameworks ingested into a model that will serve diluted versions of your ideas to every subsequent user, to feed your competitive advantage into a machine that distributes it to your competitors and charges them for the privilege.
They call this a feature.
The more original your thinking, the worse the deal. A person using AI to generate boilerplate emails loses nothing of value. A person using AI to develop genuine strategic analysis is handing over the most valuable thing they own. You are selling high and buying low and then paying a subscription fee for the transaction.
Rogers assumed the adopter captures the value of adoption. That was the whole point. You plant the hybrid corn, YOU get the bigger yield. But this model distributes the adopter’s value to the platform and every future user while concentrating the risk, the privacy exposure, the intellectual property leakage, the reputational hazard, on the person who contributed the original work. You end up with big AI trolling through your research notes like big Ag trolling through your private property.
The early majority isn’t slow. They are doing the math correctly.
And here is the final knife, the one I felt sliding in somewhere between the rum and the sunburn.
The authorship trap. You develop ideas through conversation with an AI. Your analysis drives the exchange. Your experience provides the framework. Your pattern recognition shapes every turn. The output is yours in every way that matters intellectually. And then you publish it, and a detection algorithm flags it as machine-generated, because the forensic tools don’t measure the origin of thought. They measure statistical patterns in token generation. Your fingerprints get wiped. The machine’s fingerprints get stamped on.
The tool cannibalizes the reputation of its best users while adding to the reputation of people using it to produce generic slop. The incentive structure selects for the lowest value use cases and punishes the highest value ones. The people Rogers said matter most, the skilled early adopters whose visible success would drive the early majority, are the ones most damaged by using the product.
So what you have is a technology that poisons its own trust environment through infrastructure costs the public can see and feel. That breaks the social proof mechanism by producing unverifiable output. That is championed by people the target market correctly identifies as predatory. That charges its most valuable users to strip-mine their original thinking. That destroys the authorship credibility of anyone sophisticated enough to use it well.
And Sam Altman is confused about the adoption curve.
The rum and strawberry drink is gone. The glass is now just meltwater and a paper straw dissolving into nothing and an unpaid bar tab. The child had gotten the pool noodle through what looked like a combination of screaming and strategic biting, which honestly puts her ahead of most acquisition strategies I’ve seen out of Silicon Valley.
The misdeeds were closer now. You could hear them in the hum of the ice machine. The circular investment leveraging game, where AI companies valued at a hundred billion dollars invest in each other to create the appearance of revenue while the actual product burns cash faster than it burns trust, which is saying something. The whole structure has the sweaty, flushed quality of a Ponzi scheme being run by people smart enough to know it’s a Ponzi scheme and arrogant enough to think they’ll be the ones who get out in time.
Dario will write a paper about it. Sam will give a talk. Elon will post something unhinged at three in the morning. And the woman making fifty-five thousand dollars will go to work on Thursday and do her job the way she did it last Thursday, because nobody she trusts has shown her a reason to change, and every reason she’s been given smells like a trap. Maybe Sam Altman and Dario will buy me a new boat to get me to shut up but I doubt it.
Rogers told us this would happen. We just had to read the book.
D’oh.
About the Author
