The 10/70/20 Theory: Why AI Doomers Are the New NPCs
Three workers exist in every office and on every job site in 2026. Same paycheck today, very different in five years. The split is mostly invariant — you find it in a roofing crew, a restaurant, a law firm, an accounting practice, a hospital wing, a software shop. Verify it yourself this week. Look around.
10% wield. They picked up the tool, found it useful, and kept using it. They are quietly out-producing the rest of the floor and getting promoted without making a big deal of it.
It is warming to see this 10% - line cooks at restaurants using AI (they'll be sous chefs soon!), a roofer working with Claude to reduce quote times, software developers building with natural language.
70% watch. They have heard of AI. They poked at ChatGPT once. They are not against it; they are waiting, lost in noise — for permission, for clarity, for the moment when it becomes obvious they should start. They are the majority.
20% refuse. Loudly. With essays. With threads. With LinkedIn posts about how AI is overrated, how Datacenters are bad, “AI for everyone” BS, "Ethical AI", how the bubble will pop, how nothing has fundamentally changed since the dot-com era.
They have strong opinions about a tool they have never used for more than five minutes - and built nothing useful with. This is the interesting class. Let us look at them first.
The Refusers, aka Doomers
Watch what they actually do. They have read every think-piece about AI. They can name every critic. They can quote Yann LeCun, Gary Marcus, they can quote Ed Zitron, they can explain to you non stop why the Stochastic Parrot paper proves the models do not really think. They'll tell you why AI is bad for humanity, talk about ethical AI, and wax rhetoric on how Datacenters are evil.
The number of things they have shipped with AI in the last six months? Is exactly zero. Their actual usage is nothing. Their commentary about it is infinite (and insufferable).
Why does this configuration exist?
Because if the Refuser admits AI works, they have to explain why they are not using it yet. The answer is the one nobody wants to say out loud: picking up a new tool requires being bad at it in public for a while. The Refuser will not pay that price. Their identity is "smart person who saw through the hype." Trying the tool badly threatens the identity. So they declare the tool fraudulent — it is not them, it is the technology.
It is a sour-grapes loop - The fox cannot reach the grapes, so the grapes are bad.
The blue-collar version exists too. The plumber who says he is not letting some computer tell him how to thread a pipe. The line cook who calls a kitchen using AI for prep "not real cooking." The mechanic who insists you cannot diagnose a car without your hands on it. Same structure. Same future. They will be the last hired and the first replaced — not by AI, but by the 23-year-old who shows up with AI and a tool belt.
The Watchers
These are most people, and they are fine. They have lives. Kids. Soccer practice. A mortgage. They will get there.
The danger for the Watcher is not ideological. It is drift. Every month they do not pick up the tool, the gap between them and the 10% widens by another inch. They will be perfectly safe this year, probably next year. Then one Monday they will notice that the Wielders have all the senior projects, that the Wielders are getting the promotions, that the Wielders are billing twice their hourly rate. By then the gap is a chasm.
The Watcher is salvageable. The Watcher needs one Monday morning and one real problem. That is the entire intervention.
The Wielders
Short section, because the Wielders do not need one. They are already moving.
The thing to notice about them is that they are not "AI people." They have not read the discourse. They have not picked a side. They opened the tool, found it useful, kept using it. They are the line cook learning new kitchen techniques and flagging the Tuesday shrimp overstock. The plumber quoting jobs in fifteen minutes from his truck. The paralegal closing five matters a week. The developer shipping in three days what used to take two weeks.
They are not impressed with themselves. They are too busy trying, failing, learning, and rebuilding. They are excited and hungry.
The Test
You want to know which of the three a coworker is? Simple question.
"What did you ship with AI this week?"
The Wielder opens their laptop and shows you. The Watcher says "yeah, I have been meaning to try it for that."
The Refuser delivers a six-paragraph explanation of why the question itself is malformed, why "shipping with AI" is marketing language, why you have been captured by the hype cycle.
Refuser detected. Mute immediately.
TL;DR
Three workers, same desk in 2026. The Wielder ships and gets promoted. The Watcher waits and stays where they are. The Refuser posts a thread about why AI is overhyped, hoping you do not notice they have shipped exactly zero things with the tool they are certain does not work.
If you saw yourself in the Watcher, the fix is one hour on Monday. Open the tool. Bring the worst part of your job. Build something for your kids. Work through it together.
If you saw yourself in the Refuser category — No one will save you. You can.
e/acc ❤️
Written with Wispr Flow
Originally published on LinkedIn, May 27, 2026.