AI is amazing at building the wrong things
When building gets cheap, focus becomes the most expensive skill.
Everyone is building with AI now. People do it for fun on weekends. Companies are rushing to implement it into their workflows. The speed is amazing, the output is impressive, and the barrier to creating things has never been lower.
But there is a catch. And I did not see it coming.
The harder something is to build, the more carefully you think before building it. I did not realize how much of my judgment was secretly outsourced to the cost of development, until that cost nearly disappeared.
The moment it clicked
I was supposed to be shipping one specific thing. Somewhere in the middle, AI suggested a feature — a genuinely useful one. So I built it. Then another and I went that route again.
Two hours later I had something polished and complete. None of it was the thing I sat down to do.
That was the moment I understood. I did not make a bad decision. I made a perfectly reasonable decision, in the wrong direction. And the tool I was using made it very easy to go there.
AI is a slot machine for ideas
Most suggestions are decent. Some are useless. Occasionally one is genuinely brilliant. You never know which is coming, so you keep pulling the lever.
And each response opens the next one. You ask something. The answer is useful but incomplete, so you refine the prompt. That one is better, but now you notice a new problem you had not considered. The partial satisfaction is the engine. Enough to keep going, never enough to stop and ask whether you are going in the right direction.
I have watched friends discover AI coding tools for the first time. They sit down to try one thing. They look up at 3 a.m. They got a lot done — just not the thing they sat down to do.
The side quest trap
AI does not just build what you ask for. It suggests what you did not think of. And the suggestions are good. That is the problem.
You are automating a workflow and AI proposes connecting directly to the data source instead of doing a manual export. Great idea. Days later, you have nothing to show for it because the integration was harder than expected. The manual process was two clicks.
You build something functional and AI suggests it needs version history. You spend a day on it. You never use it once. Then you spend another day on the design — for something only you will ever see.
It is not just code. You sit down to write a document and AI suggests restructuring the entire thing. You came to fix one paragraph. You leave having rewritten three sections you were not planning to touch.
Each side quest feels like an hour. Each side quest becomes a day. And every day spent on a side quest is a day not spent on the thing that actually matters.
What changed when building got cheap
When building was expensive, the question “should I build this?” had teeth. You narrowed the scope. You killed features before they were born because you could not afford to be wrong. Scarcity was painful, but it was also honest. It forced you to think.
The startup world made a religion out of this. The Lean Startup is basically a 300-page version of it: find the single core problem, do only that, prove it works.
AI removed that friction. It did not make us better at deciding what to build. It made us faster at building things we should not have started.
And there is an illusion that makes it worse. AI can prototype anything in an hour. That makes everything feel like it is one hour away. But prototyping is not finishing. The last twenty percent still takes eighty percent of the time. You think you are an hour from done. You are three days from done. By the time you realize it, you have already committed.
Focus used to be enforced by economics. Now you have to enforce it yourself.
This is coming for companies too
Right now, organizations are in the euphoria phase. AI makes teams move faster, do more with fewer people, ship at a pace that was not possible a year ago. The benefits are real.
But moving faster does not always mean moving in the right direction.
Companies are building pipelines that deliver features at incredible speed. Soon, some of them will realize they have been delivering the wrong features very fast. The same pattern that affects individuals — scope creep, side quests, the illusion that everything is one hour away — will play out at the organizational level.
When that happens, the conversation will circle back to the same principles the startup world has been preaching for decades. Think before you build. Validate before you commit. The Lean Startup was written for a world where building was expensive. The principles matter even more now that building is cheap.
What I do now
I wish I could say I have this figured out. I have not. I still slip.
But I have learned one thing: focus is the most important skill I have right now. Not prompting. Not knowing which AI model to use. Focus.
Before I build anything, I write down what I am going to create and why. Not a full specification — just enough to answer one question: what is the most important thing this needs to solve? Sometimes I write a small PRD.
AI makes you feel like you can skip that step, but you should not.
When side quests appear — and they always do — I write them down but I do not act on them. I finish what I set out to do first.
Focus was already hard before AI. The Slack messages, the emails, the meetings — there was always something pulling in another direction. AI added a new kind of pull. It is not an obligation like a late-night email from your boss. It is an invitation. And invitations are harder to turn down.