AI Won't Replace Us — It Will Multiply Mediocrity and Open a Huge Opportunity

I opened LinkedIn this week and read, for the hundredth time, some version of the same promise: "5 ideal prompts, connect your MCP, build your agent, and double your income." AI pseudo-experts everywhere. A year ago the same person was writing about Web3. Two years ago, about the metaverse.

The saturation of pseudo-experts bothers me, but not for the obvious reason. It bothers me because it hides the real question: do we actually know what we want to use AI for?

It's the same divide I watched 25 years ago with Google. One person used it to find the latest viral joke. Another used it to find information that actually changed a decision. Time moved on. Our way of seeing and consuming technology, not so much. And that's where the difference is drawn.

What I'm watching (and what worries me)

I see more anxiety about mastering the tools than curiosity about understanding what is going on. And what is going on, the way I read it, is not that AI will replace us — not in the way people usually mean. It's two currents at once: AI is going to multiply mediocrity, and at the same time it will open a huge opportunity for those who already have deep knowledge. Not one or the other. Both, in parallel.

Today the average designer asks AI for an image without a meaningful input, asks AI for a website without knowing why a component goes where it goes, and ships. That — multiplied by thousands of people doing the same thing — doesn't reduce mediocrity: it multiplies it. More amateurs producing more low-effort output, faster, across more channels. The saturation is structural.

At the other end of the curve there's a massive opportunity for the people who do know: those who understand the business before they prompt, those who can look at a model's output and say "this doesn't work, and I can tell you exactly why." That second current — the one of judgment — stands out more, precisely because the first one floods everything else.

If you were the designer who copied from Behance, or built every project off a template, the first current is going to swallow you. And the market is going to be flooded with that kind of content — no matter how well the AI is trained. AI can't create from zero. It combines and structures, but it can't generate outside the distribution it was trained on.

"It's already happening" — the data

This isn't a prediction for next year. It's measurable today.

Doshi and Hauser published in Science Advances a study where AI-assisted writers produced stories judged more creative individually — but collectively more similar to each other than stories written without AI (Doshi & Hauser, 2024). Each person gets a little "better." The ecosystem flattens. You saw this when everyone posted the same "what does AI know about me" image — different features, different elements, the same exact aesthetic.

The same is happening with website and app designs. And it will get worse. Anderson and colleagues documented that LLM-assisted ideation produces a narrower idea space than ideation without AI (Anderson et al., 2024). More volume. Less variety. More of the same.

And there's a technical mechanism behind it. When models train on data generated by previous models, the "tails" of the distribution — the rare, the original, the strange — start disappearing. Shumailov and co-authors showed in Nature that this produces model collapse: what's left is the statistical center, the average of the average (Shumailov et al., 2024). The copy of the copy, made structural.

Is this new, or just acceleration?

A fair objection: "the designer who copied from Behance was already being replaced by a cheap Fiverr freelancer before AI existed." That's partly true. The pressure on derivative work isn't new.

What is new about AI is scale and speed. Before, producing 50 generic mockups meant paying 50 people. Today they get produced in an afternoon. The consequence: the minimum quality floor of the market jumps suddenly, and everything that was barely above that floor sinks below it. This isn't romantic. It's the geometry of the market.

The other half of the thesis: the opportunity

While the first current multiplies mediocrity, the second one opens space for those of us who still love sketching wireframes, laying things out honestly, and understanding why a component goes here and not there.

I want to be precise here because the temptation is to make a nostalgic argument, and that's not what I'm making. I'm not defending the manual because it's manual. I'm defending it because it trains the eye.

Sketching a wireframe on paper isn't virtuous because it's old. It's virtuous because it forces you to make visual-hierarchy decisions before you have a pretty component to disguise them. It forces you to know why one button goes on top and another at the bottom. It trains a sense of weight. And that sense is what later translates into better prompts, into better critique of model output, into digital products that don't blend into the noise as one more of the pile.

In The Craftsman, Richard Sennett calls this the cycle of the trade: the hand educates the head as much as the head educates the hand (Sennett, 2008). Someone who has sketched wireframes for ten years doesn't have a sentimental advantage. They have a cognitive one: they can see what's wrong before they can articulate why. In a market flooded with generic outputs, that advantage is worth more, not less.

What's coming

Here's my bet:

Over the next 18 months we're going to see both currents at once. The tide of mediocrity is going to keep rising: more identical images, more generic sites, more "agents" repeating the same thing. At the same time, work with judgment — the kind that comes from understanding a business, looking at a screen, and knowing why something doesn't work — is going to go up in value.

The paradox: the more mediocrity the tool multiplies, the more visible the judgment of the people who actually know becomes.

It's a good moment for those who understood that the tool was never the point.

It's a bad moment for those who thought it was.

That's the opportunity. And it isn't for everyone.