A year ago, I couldn’t build software. I could hack together websites, copy-paste scripts I half-understood, tweak some CSS. But actual tools? That required programming knowledge I didn’t have.

Now I build things constantly. Small tools, automations, scripts for specific workflows. The barrier between “I want this” and “I have this” collapsed almost overnight.

Here’s what I’ve noticed: the hard part isn’t building anymore. The hard part is knowing what’s worth building.

Execution Got Solved⌗

AI coding tools solved the execution problem. Not perfectly, but well enough that someone like me - technical-adjacent, not a developer - can build real things that work.

This is genuinely new. The skill that used to gate software creation - knowing how to code - is no longer the bottleneck. Describe what you want clearly enough, iterate through the rough patches, and you’ll get something functional.

The bottleneck moved upstream. It’s not “can I build this?” anymore. It’s “should I build this?”

And that question didn’t get any easier to answer.

The New Differentiator⌗

Watch how different people use these tools and a pattern emerges.

Some people build fewer things, but the things they build become permanent parts of their workflow. They’re solving real problems - tasks they do daily, friction points that actually cost them time. Every tool earns its place.

Other people build constantly. Dozens of small projects, most of which get used once or twice and forgotten. They’re solving problems they had once, or problems they might have someday, or problems that sound interesting but don’t actually matter. Activity that looks like productivity.

The difference isn’t technical skill. Both groups have access to the same AI tools. The difference is taste - knowing what’s worth building before you build it.

What Taste Actually Means⌗

Taste in this context is judgment about value. It’s the instinct that distinguishes “this saves me 30 minutes a week” from “this saves me 30 minutes once.” It’s understanding the difference between a real problem and a problem that feels urgent in the moment but isn’t.

You can’t prompt your way to better taste. The AI will build whatever you ask for. It has no opinion about whether the thing should exist.

That judgment is entirely on you. And it’s the skill that separates people who become meaningfully more productive with AI tools from people who just produce more stuff.

The Trap⌗

Because building is cheap, it’s tempting to build instead of think.

Vague idea for a tool? Why sit with it when you can prototype in an hour? The feedback loop is so fast that building feels like thinking.

It’s not.

Building and thinking are different activities. When you substitute one for the other, you end up with a lot of built things that shouldn’t exist. Functional solutions to non-problems.

The people who use AI tools well don’t fall into this trap. They’re ruthless about what they won’t build. They’ve internalized a filter - probably through years of the old friction teaching them which ideas are actually good - and they apply it before they start, not after.

The Filter Questions⌗

Before starting anything, a few questions are worth asking:

  • Is this solving a problem I have daily, or a problem I had once?
  • If this works perfectly, what actually changes about my week?
  • Am I building this because it’s useful, or because it would be cool to have?

“Cool to have” is a trap. It feels like a reason, but it’s not. Cool things that don’t get used are just clutter with better marketing.

The hardest question: if this works perfectly, what changes? Not “what could this do” - what actually changes? Be specific.

If you can’t answer that clearly, you’re probably about to build something that shouldn’t exist.

What This Means⌗

The tools have never been better. Anyone who can describe what they want can now build it.

But the constraint was never really execution. It was knowing what was worth executing. The old friction just masked that by making execution hard enough that only the best ideas survived the filter.

Now execution is easy. The filter is gone. And a lot of people are discovering they don’t actually know what’s worth building.

That’s the skill that matters now. Not prompting, not “AI literacy,” not any of the buzzwords. Just taste. Judgment about what deserves to exist.

The AI can’t help with that part.