Obsession Is the Bottleneck
The last 20 years of technology ran on skill scarcity. You needed engineers to write code, designers to build interfaces, analysts to crunch data, writers to produce copy. Entire companies existed because those skills were hard to find, and the people who had them could charge accordingly.
AI is dismantling that premise faster than most people have actually stopped to process. Models can now write serviceable code, generate design systems, summarize research, draft legal documents, build prototypes, and the list keeps getting longer every few months. The cost of turning an idea into something real is in freefall, and it’s not going to stop falling anytime soon.
When execution gets cheap, something else becomes scarce. And it isn’t skill.
Building anything used to hit three walls: knowing how to do something, having the technical ability to do it, and having the time and manpower to see it through. AI compresses all three simultaneously, which still feels strange to say out loud. You no longer need to understand compilers to ship software, or graphic design theory to produce decent visuals, or a team of analysts to work through a large dataset. The tools increasingly handle that work, which shifts the question away from capability entirely. The constraint becomes something much simpler and, it turns out, much harder to manufacture: who actually cares enough to finish?
Most people dramatically underestimate how rare obsession is. Not interest, which is common and almost everybody has it about something. Not motivation, which comes and goes depending on how the week is going. Not ambition, which is mostly just ego with a direction attached to it. Obsession means thinking about the same problem every day for years, noticing small improvements that nobody else would bother to see, refusing to drop something even when progress has become invisible and the reasonable thing, the obviously correct thing, would be to move on. It’s a specific and fairly strange psychological trait, and it doesn’t become more common just because the tools around it get better.
AI lowers the barrier to entry. It doesn’t touch the barrier to persistence.
The next decade is going to produce an enormous flood of half-finished things. Apps, newsletters, research tools, businesses, financial models, media brands. AI makes starting all of them trivially easy, so the internet will fill with projects that made it to 80% and stopped. That stopping point was never really a technical problem though. It was always a psychological one. The tools were never what stood between most people and completion. The will to keep going was, and that hasn’t changed and probably won’t.
As the pool of creators expands, it will split more visibly into two groups. Tourists, who experiment, produce a few things, and move on when the initial excitement fades. And obsessives, who keep iterating on the same idea for years, long after it stops being novel or interesting to anyone watching from the outside. Tourists will multiply dramatically. Obsessives will stay roughly as rare as they’ve always been, which means that rarity becomes the actual moat. Not access to tools, not technical skill, not capital.
There’s a popular framing that AI will democratize creation, and that’s true as far as it goes. But democratization doesn’t flatten outcomes, and history is pretty consistent on this point. It tends to amplify extremes. When everyone can make something, quality distributions get steeper, not flatter. The best work improves faster because obsessives now have better tools. Mediocre work multiplies because tourists have lower barriers. The middle, the competent-but-uninspired work that used to be good enough to get by, largely disappears.
What looked like talent in the past was often just technical ability, the capacity to execute on something that most people couldn’t do. In the AI era, talent will look more like sustained attention allocation. Who keeps returning to the same problem after the initial momentum is gone? Who keeps improving the same product after it’s already good enough? Who keeps sharpening the same thesis after everyone else has moved on to the next thing? AI makes iteration cheap. It doesn’t make anyone care about what they’re iterating on, and that gap is bigger than it looks.
Every major technological revolution removes a constraint. Agriculture removed food scarcity. Industrialization removed manufacturing scarcity. The internet removed information scarcity, though what it replaced it with is its own conversation. AI is removing capability scarcity, and it’s doing it faster than the previous revolutions removed theirs. Once capability becomes abundant, the scarcest resource is something much older and much harder to scale: sustained human attention applied to a single problem over a long period of time. The willingness to stay in the room with something difficult, not because progress is guaranteed, but because you can’t seem to make yourself leave.
AI won’t replace that. It will just make it more obvious how rare it always was.