We’ve all heard it. Pick one thing. Go deep. Become the master. Everything else is a distraction, a hedge, a sign you’re afraid to commit. Jack of all trades, master of none, they’d say, and they’d say it like it was already the punchline.

Nobody ever asked if the rule still holds. It got treated like gravity. A constant, not a policy. But rules that get repeated without being questioned are usually rules that benefit somebody, and they’re usually built for a world that has since moved on without telling anyone.

Here’s the world that rule was built for: depth was expensive. If you wanted to actually understand something, really understand it, down to the mechanism, you needed years, a mentor, a library card, or a job that happened to force you into the internals. Most people didn’t have access to all three. So depth became scarce, and scarcity became value, and the people who had it built an entire mythology around the discipline it took to get there. Master of one. Ten thousand hours. The specialist as hero.

That mythology made sense when going deep on a second thing cost you almost as much as the first thing did. You had a finite number of decades and each unit of depth was expensive, so you rationed it. Specialize or die diluted, that was the actual tradeoff, and the people telling you to specialize weren’t lying. They were describing the economics of their own time accurately.

That economy is gone.

Right now, today, you can open a model and get a working mental model of distributed consensus, or protein folding, or how a compiler actually lexes and parses, in the time it takes to make a coffee. Not a Wikipedia summary. Not a listicle. An actual, interrogable, deep explanation that you can push back on, ask follow-ups to, get wrong, get corrected, and iterate until it clicks. The cost of going deep on any one thing collapsed to nearly zero. Not free, exactly. But close enough that the old rationing logic doesn’t apply anymore.

This is the part people skip past too fast, so slow down on it. It’s not that AI made you smarter. It’s that AI destroyed the reason depth used to be rationed. When something that used to cost years starts costing an afternoon, the strategy built around that cost stops making sense, and you have to ask what strategy replaces it.

Watch what actually happens when depth gets cheap. The person who “mastered” one narrow thing, who spent a decade becoming the go-to on a single tool or a single stack or a single domain, finds that their moat just got a bridge built across it. The model can produce their depth on demand, for anyone, at a fraction of the time cost. That doesn’t make their knowledge worthless. It makes their knowledge available. Which is not the same thing, and the difference is where the whole argument lives.

Because here’s what the model can’t do for you: it can’t tell you which twenty things are worth going deep on. It can’t tell you that the bug in front of you isn’t actually a database problem, it’s a caching problem wearing a database costume, because that recognition doesn’t come from depth in one domain, it comes from having stood close enough to five domains to see the family resemblance. It can’t hand you taste. It can’t hand you the instinct that says this explanation sounds too clean, something’s missing, go check.

That instinct is breadth. And breadth was always the harder skill, not the easier one. It just never got scored, because there’s no belt for it, no certification, no title that says “connects things nobody else in the room connects.” The specialist gets a plaque. The generalist gets accused of not committing. That was never a fair trade, it was just the trade the economics allowed, and the economics have changed.

Think about what actually breaks a system, any system, code or otherwise. It’s almost never a failure inside one domain that a specialist would have caught. It’s a failure at the seam between two domains, where the person responsible for each side assumed the other side was handling it. The specialist’s whole training is to be excellent inside the seam’s walls. Nobody’s trained to watch the seam itself. That’s the generalist’s actual job, and it was always the job, it just didn’t have a name that sounded impressive at a dinner party.

Now put AI in the room. The model will confidently generate depth on any single axis you point it at. It will also confidently generate depth that’s subtly wrong, coherent-sounding, internally consistent, and false in a way that only someone with enough surrounding context would catch. Who catches that? Not the person who only knows the one domain, because to them the output looks exactly like what a domain expert would say. It’s the person who’s touched enough adjacent territory to feel the seam creak. Depth alone can’t catch a well-dressed lie. Breadth is what smells it.

So the real shift isn’t specialist versus generalist. Framing it that way is the old rule refusing to die gracefully. The real shift is that judgment just became the scarce resource, and judgment isn’t manufactured by depth, it’s manufactured by breadth applied on top of depth you actually understand. The person with ten thousand hours in one thing and zero context outside it is now sitting on a skill that a model can rent out by the minute. The person with working knowledge across ten things, thin in each but real, not surface-level buzzword real, is sitting on the one thing the model still can’t do for them: deciding what actually matters and knowing when the answer’s wrong.

None of this means depth stopped mattering. That’s the lazy read and it’s wrong. It means depth stopped being the finish line. It became a commodity you can acquire on demand, which means the actual skill moved one level up, to knowing which depths are worth acquiring and being able to check the acquired depth against something real instead of just trusting the fluency of the output.

Jack of all trades was never the failure mode people made it out to be. Master of one thing and blind to everything touching it, that’s the failure mode, and it was always the failure mode, we just didn’t notice because for most of history that specialist blindness didn’t cost anything visible. It cost something invisible: the seam nobody was watching, quietly rotting, until it wasn’t invisible anymore.

The people getting replaced right now aren’t the generalists. It’s the specialists who spent thirty years mistaking depth for judgment, because for thirty years nobody made them tell the difference.

So if a model can hand anyone instant depth in almost anything, the only scarce skill left is picking the twenty things worth being deep in and having the judgment to know when you picked wrong. What happens to someone who spent a career being excellent at the one thing they never questioned?