The 10x knowledge worker is dead. Meet the 100x Curator

My daughter is going to grow up interacting with AI in basically every part of her life. That single fact shapes how I think about knowledge work right now.

She won't be valued for what she knows, or for how much she can produce. Machines will have both of those covered. She'll be valued for how she directs AI, how sharply she asks questions, and how well she synthesizes across domains. Which means a lot of the skills we still optimize for are quietly going out of date.

The 10x knowledge worker, the one who just produced more than everyone else, is finished. When output gets cheap, producing more of it stops being a differentiator. The next decade belongs to a different kind of worker. Call them the 100x Curator: the person who directs the machines and stays accountable for the judgment.

From producing to curating

For most of the knowledge economy, knowing things was the bottleneck. Memorization mattered. Recall was a real skill. We built schools and careers around it.

That bottleneck is gone. The new bottleneck is judgment. Knowing which sources to trust. Doing the due diligence on what the AI tells you, instead of pasting it straight into a board deck. The expensive mistakes of the next decade won't come from people who couldn't produce. They'll come from people who produced confidently and never checked.

We're still training for the old job. Schools optimize for recall. They should be teaching verification, synthesis, and taste. The freed-up capacity from AI is only worth something if people use it to think deeper, not to automate mediocrity faster.

The five skills that actually compound

"Learn AI" is useless advice. Learn which part? Here's a more concrete answer: the five skills that separate people who get real leverage from AI from people who just generate more text.

Verification. Checking what the model claims before you act on it. The single highest-value habit, and the one most people skip.
Synthesis. Pulling a clear answer out of many sources, instead of accepting the first plausible one.
Framing. Asking the question well. A sharp prompt beats a powerful model fed a vague one.
Judgment. Knowing when to trust the output, when to dig, and when to walk away.
Orchestration. Chaining tools and steps into something that actually finishes a job, not just a clever demo.

Notice what's missing: "knows the most prompts." Prompt tricks age in months. These five compound for years. And you don't build them with a tool tutorial. You build them by doing real work with AI in the loop and reviewing the misses.

AI-Native versus AI-Tourist

Here's what keeps me up at night about the next five years. Companies are splitting into two classes, and the gap between them will be hard to close.

AI-Tourists visit AI. They bolt a chatbot onto existing processes, add a few automations, call it transformation, and wonder why a competitor is moving so much faster. AI-Natives live there. They rebuild the workflow around what AI makes possible. Different processes, different hiring, different expectations of what a day's work even looks like.

The tourist optimizes the old way by 10%. The native asks whether the old way should exist at all. This isn't about company size or budget. I've seen 30-person firms that are deeply AI-native and big enterprises that are pure tourism with a large bill. The productivity gap between the two compounds, and that's exactly what makes it dangerous.

The fastest wins are boring

When people imagine AI transformation, they picture something flashy. In practice, the highest-value wins are dull, and they almost all come back to one thing: finding what you already know.

The scale of the waste is well documented. The McKinsey Global Institute found that knowledge workers spend an average of 1.8 hours every day, around 9.3 hours a week, searching and gathering information. That's nearly a fifth of the working week spent looking for things that already exist somewhere in the organization.

On top of that sits a tax nobody puts on a P&L: corporate amnesia. The same problem solved three times by three teams who didn't know the others had already done it. The proposal rewritten from scratch because nobody could find last year's winning one. The new hire who takes months to learn what already lives in someone's old emails. The expert who leaves and takes a decade of context out the door.

Most companies think they have a knowledge base. What they actually have is a parking lot for documents, because storage is not retrieval. The maturity curve runs in four stages. Stage 1 is storage: files exist somewhere, and finding them depends on who you ask. Stage 2 is keyword search, which works only if you already know the exact term. Stage 3 is retrieval: you ask a question in plain language and get the relevant passages back, with sources. Stage 4 is synthesis: the system pulls across systems and gives one sourced answer, respecting who's allowed to see what.

The jump that changes everything is from search to retrieval. That's the moment your archive stops being a cost and starts being an asset people actually use. Most organizations are stuck between stage 1 and 2 and don't realize stages 3 and 4 are now within reach.

Give the hours back

When AI does free up time, most companies waste the dividend. They measure "more done per hour," then immediately fill the saved hour with more tasks. Net effect on a human life: zero. You've just automated the treadmill to run faster.

The evidence that AI frees real time is solid. In a study of 5,179 customer support agents, researchers Brynjolfsson, Li, and Raymond found a 14% average increase in issues resolved per hour, rising to about 34% for the least experienced workers. The gains are real. The question is what you do with them.

I set my own company a goal that sounds odd for a tech founder: give hours back to the team, without cutting pay or people. Because if our AI actually works, time is what it should produce. The way to protect that is deliberate. Find the time AI genuinely frees up and measure it honestly. Decide on purpose where it goes, into deeper work, better thinking, or actual rest. Then defend it, because saved time evaporates by default. And reinvest some of it into verification and judgment, so quality rises with speed instead of falling behind it.

AI won't replace your job. It'll replace your company

Everyone's debating whether AI takes jobs. I think that's the wrong unit of analysis.

Look at Kodak. They didn't die because digital photography was unstoppable. They invented it. An engineer named Steven Sasson built the first digital camera prototype inside Kodak in 1975. Management buried it, afraid it would cannibalize the film business. The result, decades later: a workforce that went from 145,000 employees at its 1988 peak to roughly 3,500 today. Whole communities lost pensions, healthcare, and a way of life, because the industry around them shifted and their employer refused to shift with it.

Kodak had the technology. What they lacked was the courage to disrupt themselves. Here's the uncomfortable version for today: your job might be perfectly safe while your employer quietly becomes obsolete. The real disruption isn't roles being automated. It's business models that stop making sense, and the people attached to them. The question to ask isn't "will AI take my job?" It's "will AI take my company, and is leadership brave enough to change before it does?"

Human work becomes a luxury good

One more prediction for the decade ahead. "Made by humans" becomes a premium label, the way "handcrafted" is today.

The question stops being whether AI can do your job. It becomes whether anyone will pay extra for the human version. In a lot of cases, they will. There will be AI-generated contracts and human-drafted ones, AI marketing copy and human copy, AI-designed buildings and human-designed ones. Some clients pick AI for speed and price. Others pay the premium for human judgment, taste, and accountability, for a name attached to the decision.

If output is becoming a commodity, the premium moves to judgment, trust, and taste. That's a good thing to build toward, and it's the same thing the 100x Curator is built on.

Bet on judgment

The skills that made you a great 10x producer aren't the ones that'll matter most next. Verification, synthesis, framing, judgment, orchestration: that's the new core. Pair it with retrieval that actually works, time that's deliberately protected, and a business model brave enough to change, and you're on the right side of the divide.

I expanded the five skills into a free guide, with how to build each one, alongside the storage-to-retrieval maturity model. You can get it through our newsletter here.

If output is becoming cheap, what's the skill you're betting on becoming valuable? I'd genuinely like to know. My DMs are open.