Specialization is for insects. I think Heinlein said that, and while he was a bit of a crank, he was dead right on this one. For the last twenty years, the world has been screaming at us to pick a lane, get a niche, and become the ‘go-to person’ for one specific, tiny thing. If you were the world’s leading expert on a specific type of database architecture or a niche tax law in Delaware, you were set for life. You were unkillable.
That world is gone. It died the second someone figured out how to make a large language model pass the Bar exam. If your entire value proposition is based on knowing a massive amount of specific, static information, you aren’t a professional anymore. You’re just a slow, expensive version of a GPT-4 plugin. And honestly? I think a lot of people are terrified to admit it.
The day I realized I was a fraud (and why it was a relief)
In early 2018, I was sitting in a gray, windowless conference room at a fintech startup in San Francisco—I won’t name them, but let’s just say they rhyme with ‘Affirm’—trying to prove I was a ‘Data Scientist.’ I had spent six months taking Python courses and memorizing SQL joins because that’s what the market wanted. The interviewer asked me to write a specific recursive function on a whiteboard. I froze. I couldn’t remember the syntax. I felt like a total failure, a shallow generalist who couldn’t cut it in the ‘real’ world of experts.
I walked out of that building feeling like garbage. But then I got a job at a small logistics firm where I had to do everything: talk to truck drivers, fix the broken Zapier integrations, write the marketing copy, and occasionally look at the data. I realized that my ability to connect the truck driver’s complaint to a bug in the code was worth way more than being able to write a recursive function from memory. I wasn’t an expert in anything. But I was the only person in the room who understood the whole machine. It was the first time I felt actually useful.
Being the ‘glue’ is a much more stable career than being the ‘brick.’
Anyway, back to AI. The thing about these models is that they are the ultimate specialists. They have read every medical journal, every line of documentation, and every legal brief. If you try to compete with them on ‘depth,’ you will lose. Every single time. But where they fall flat—and where they will continue to fall flat for a long time—is the messy, horizontal space between domains.
AI is coming for the people who know exactly one thing
I’ve been tracking my own output lately. I manually logged my Jira tickets and tasks for about 180 days last year. I found that out of 412 distinct ‘problems’ I solved, only 45 of them required what I would call ‘deep domain expertise.’ The other 367 were essentially translation errors between humans or systems. It was stuff like: ‘The marketing team wants X, but the API only gives Y, so we need to tweak the process to make Z happen.’
AI can do X. It can do Y. It might even be able to do Z. But it doesn’t know that the marketing lead, Sarah, is having a bad week and will reject the project if we don’t frame it as a ‘cost-saving’ measure. It doesn’t know that the legacy codebase will catch fire if we push a change on a Friday.
I know people will disagree with this, and they’ll point to ‘AI Agents’ that are supposed to handle cross-functional tasks. But have you actually used them? They’re brittle. Specialists are like high-performance tires on a Formula 1 car: incredible on a dry track, but the second it rains, they’re useless. A generalist is an all-terrain vehicle. We’re slower, we’re louder, and we’re a bit uglier, but we actually get to the destination when the weather turns.
It’s a trap.
The ‘Connective Tissue’ theory of staying employed
The safest place to be right now is in the middle of three or four different departments. If you can speak ‘Developer,’ ‘Accountant,’ and ‘Human,’ you are basically un-fireable. Why? Because the specialist-heavy companies are the ones currently laying people off in droves. When Google or Meta cuts 10,000 people, they aren’t cutting the people who know how everything fits together. They’re cutting the 500 people who all do the same hyper-specific task that an LLM can now do for $20 a month.
I used to think being a ‘Jack of all trades’ was just a cope for people who weren’t smart enough to be neurosurgeons. I was completely wrong. In a world where the cost of ‘expertise’ is trending toward zero, the only thing that retains value is the ability to synthesize.
The value isn’t in the knowledge anymore; it’s in the judgment.
What I mean is—actually, let me put it differently. If I give an AI a prompt to ‘build a marketing plan,’ it will give me a generic, B+ plan. If I give it to a specialist, they will give me an A+ plan for their specific channel (like SEO or PPC). But a generalist will look at the B+ plan and realize the entire product is actually the problem, and we should probably just pivot to a different market entirely. AI can’t tell you to stop doing the wrong thing. It can only help you do the wrong thing faster.
Why I won’t hire an ‘AI Prompt Engineer’
This is my mini-rant for the day, and I know I’m going to sound like a hater, but I don’t care. I refuse to hire anyone who lists ‘AI Prompt Engineer’ as a primary skill on their LinkedIn profile. It’s embarrassing. It’s the 2024 equivalent of putting ‘Google Searcher’ or ‘Microsoft Word Expert’ on your resume in 2005.
- It shows you think the tool is the skill.
- It suggests you’ve specialized in a temporary interface that will be obsolete in eighteen months.
- It screams ‘I don’t have a real background in anything.’
I’ve seen people charging $5,000 for ‘AI Masterclasses’ and I actively tell my friends to avoid them. It’s a scam. Spend ten minutes playing with the tool and you’ve learned 80% of what you need. The other 20% comes from having a generalist background that allows you to know what to ask the AI in the first place. You can’t prompt an AI to solve a supply chain crisis if you don’t know how a warehouse works.
Total waste of time.
A slightly unfair take on LinkedIn
While I’m at it, I’ve developed a visceral hatred for the ‘Top Voice’ badges on LinkedIn. They almost always go to the hyper-specialists who post the same ‘5 tips for [Niche Topic]’ every single morning at 8 AM. These people are building brands on foundations of sand. They are optimizing for an algorithm that wants them to stay in their little box. But the box is shrinking. I’d much rather follow someone who is a bit of a mess, who posts about philosophy one day and Excel macros the next. That person is actually thinking. The ‘Top Voice’ in SEO is just a bot waiting for its turn to be decommissioned.
I might be wrong about this. Maybe the world will become so complex that we need even more specialized experts to manage the AI that manages the other AI. But I doubt it. History usually rewards the people who can pivot. When the industrial revolution hit, the people who survived weren’t the ones who could weave faster than a machine; it was the people who realized they could buy ten machines and start a factory.
The generalist is like a sourdough starter—messy, weird, and capable of making a hundred different things if you just feed it. The specialist is a sourdough loaf. Great for one meal, but once it goes stale, it’s over.
I’m still not sure what I’m going to tell my kids when they ask what they should study in college. Part of me thinks they should just major in ‘Curiosity’ and minor in ‘Not being a jerk,’ but that doesn’t really work on a transcript. I genuinely don’t know if a traditional degree even makes sense for a generalist anymore. Maybe the best move is just to fail at five different jobs in five different industries before you turn 30. That’s basically what I did, and it’s the only reason I’m not worried about my mortgage right now.
Go learn something you have no business learning. Today. That’s the whole trick.