AI Makes the Dumb, Dumber & The Smart, Smarter

Everyone sold you the same story: AI democratizes intelligence. Now anyone can write like a professional, code like an engineer, analyze like an expert. The floor rises for everybody, so we all win.

Half of that is true. The floor does rise. That's exactly the problem.

A rising floor with no rising judgment doesn't produce competent people. It produces confident people who are wrong faster, at greater volume, with cleaner formatting. The person who never learned to question a source now has a machine that will produce a fluent, authoritative-sounding answer to anything, and no habit of checking it. They didn't gain intelligence. They gained a louder echo of their own lack of it.

Here's why the split happens, in three mechanisms.

1. The floor rises for everyone. The ceiling only moves for those who can already reach it.

AI hands the incompetent a way to ship mediocre work fast. That feels like progress and it isn't. Mediocre was never the bottleneck. And yes, the famous experiments found AI lifts the weakest performers most. Read what they measured: memos, drafts, brainstorms. Floor work, graded on floor terms. Not one of them measured whether anyone's judgment grew. The equalizer evidence is a study of floors being quoted as if it were a study of ceilings. The ceiling, meaning insight, taste, knowing which of ten plausible answers is actually right, doesn't move unless you already have the judgment to direct the tool toward it. AI is a lever. A lever multiplies force; it does not supply it. Zero force through an infinite lever is still zero. If you bring nothing, it multiplies nothing.

2. Output is only as good as your ability to catch what's wrong with it.

AI answers arrive with errors baked in: subtle, confident, well-dressed errors. There's a tax on using the tool. You have to be able to spot them. The person with real knowledge and context pays that tax easily. They read the output, feel the thing that's off, correct it, and walk away with something better than they could have made alone. The person without context can't pay. They can't tell the good answer from the plausible one, so they ship the plausible one, and every mistake compounds, at scale, with a machine's endurance behind it. Verification is the whole game, and only one group can play.

3. The dumb outsource their thinking and let it rot. The smart outsource the grunt work and reinvest what they save.

Give the same tool to two people. The first person offloads the thinking itself, the deciding, the weighing, the judging, and the muscles that do that work atrophy from disuse. It has a name, cognitive offloading, the same reflex that deleted every phone number you used to know, now pointed at judgment itself. Give it a year and they can't function without the crutch, and the crutch was never load-bearing. The second person offloads the drudgery, the boilerplate, the first draft, the lookup, and takes the hours they clawed back and pours them into the hard part: the strategy, the synthesis, the questions the machine can't ask. Same tool. One mind shrinks. One mind compounds.

This was always the real dividing line

Long before AI, people split into two camps. Not cleanly. Most of us live on a slider between them, rigorous in the few domains we've actually earned and credulous everywhere else. But the poles are real, and AI drags you toward whichever one you feed it most.

One camp lets itself be led. Whatever came through the screen first, the news anchor, the viral post, the loudest voice in the group chat, became the truth. No proof required, no second source, no friction. They outsourced their conclusions their whole lives. AI is simply the most powerful outsourcing engine ever built. Of course it makes them worse. You built a firehose and handed it to someone who never learned to ask where the water came from.

The other camp earned its views. They read, they were wrong and corrected themselves, they lived through things, they built context you can't fake. For them AI isn't a replacement for thinking. It's a sparring partner, a research team, a second brain that takes their already-strong foundation and stress-tests it, extends it, sharpens it. They walk in strong and walk out stronger.

The tool doesn't decide which one you are. It just finds out.

The strongest objection, and why it doesn't save the argument

The honest counter goes like this: AI can build judgment in people who lack it. A patient tutor that never tires, explains anything at any level, and meets you where you are should, in theory, lift the credulous into critical thinkers. The floor doesn't just rise, the argument goes; the tool teaches people how to climb.

I'll grant the theory. I've seen it happen. A motivated beginner with a good AI tutor can learn faster today than at any time in history.

But look at the load-bearing word: motivated. Building judgment through AI requires the exact traits the credulous don't have: the impulse to ask "is that true?", the patience to sit in not-knowing, the willingness to be corrected. The tool can teach critical thinking only to someone already willing to think critically. Everyone else uses the identical tool to get answers faster and question them less. The tutor and the crutch are the same product. Which one you hold depends on what you brought to it, which is exactly the point. The counter doesn't break the thesis. It restates it.

You've read this far assuming you're in the second camp. Everyone does. That assumption, unexamined, is the first camp's whole problem.

So take your shot. Here is what would change my mind: evidence that people with no habit of questioning develop the habit itself from using AI. Not better output. Better judgment, still there after the tool is taken away. Find that at scale and I fold. I do not think you will.

What to actually do about it

Stop treating AI as an oracle and start treating it as an intern with a genius vocabulary and zero accountability. Never accept a first answer you couldn't have argued your way to yourself. Build the context, read the hard things, learn the domain, earn opinions you can defend, before you reach for the tool, because the tool only amplifies what's already there. Use the hours it saves you to think harder, not to think less.

The equalizer story is a comfortable lie. AI doesn't close the gap between the sharp and the credulous. It takes that gap and turns it into a canyon.

Which side you land on was mostly decided before AI ever showed up. But you still have time to change the answer. Barely.