When the Cost of Intelligence Collapsed

The scarce resource is no longer implementation. It is judgment.

Something changed in how I work, and it happened so gradually I almost missed it.

I run a personal AI pipeline. Every morning, it reads through hundreds of articles, scores them against criteria I’ve defined, and surfaces the ones worth my attention. I didn’t outsource this to a developer. I didn’t hire an agency. I built it myself, over a few months — and most of it was written with Claude Code.

That fact still surprises me when I say it out loud. Not because the tool is impressive. Because I spent years deciding this kind of thing wasn’t worth building.

Here’s the insight at the center of this: the cost of what I’m calling “intelligence” — the kind required to write code, design logic, and build functional software — has dropped significantly. And when a cost drops, it doesn’t just make existing things cheaper. It changes what’s worth attempting.

By intelligence, I don’t mean general cleverness. I mean the kind of applied reasoning that used to be locked behind years of training or an expensive hire. That’s the thing that’s become dramatically less expensive to access.

The clearest examples of this aren’t the headline numbers.

A restaurant chain owner in Hokkaido built a 74-store management dashboard — inventory, sales, staff scheduling — without any engineering background. “Things I used to pay development agencies hundreds of thousands of yen for,” he wrote, “I can now make myself.” On Hacker News, someone posted: “I’m 60, I’ve never written code in my life. I just built my first complete application.” It got hundreds of replies. Not because it was unusual. Because people recognized themselves in it.

These are the important cases. Not the spectacular ones.

A solo developer did sell a company to Wix for $80 million six months after starting. An AI-powered app builder hit $100 million in annual recurring revenue eight months post-launch. These numbers are real, and they tell you something about the speed of the shift. But they’re the wrong signal to focus on. Exceptional outcomes at that scale involve factors beyond tooling.

The stranger part is not that a solo developer can sell a company for $80 million. The stranger part is that a restaurant owner can now build the internal system he used to outsource.

That’s the structural change. The population of people who can build functional software has expanded. Not at the margins — meaningfully.

I have about thirty tools on my personal site. Most were built with Claude Code. Some took an afternoon. One took five minutes.

I want to be precise about what this means.

It doesn’t mean anyone can build anything. The distinction between good software and bad software hasn’t collapsed. Engineering skill remains valuable — if anything, the ability to think clearly about what a system should do matters more now, not less. What’s changed is the threshold for “worth building.”

Think about the pattern that plays out whenever a meaningful cost drops. The desire to travel longer distances wasn’t born in 1950. It existed in 1890. What changed was that cars became affordable. Once that happened, suburbs were invented, commuting distances expanded, and freight logistics reorganized around entirely new assumptions. New behavior didn’t emerge because people wanted different things. It emerged because a cost structure shifted, and suddenly things that were economically irrational became rational.

When broadband became cheap, it didn’t just make existing internet faster. It made YouTube viable. A new cost structure made an old desire satisfiable at scale.

The 60-year-old on Hacker News isn’t a productivity story. It’s a category expansion story. A person outside the population of software builders is now inside it.

Here’s what I notice when I apply this to my own work.

Before, when I had an idea for something I wanted to build, the first question was: “Is it worth the cost?” — meaning the time, the hiring, the coordination. A lot of ideas failed that test. Not because they weren’t useful. Because the cost of realizing them was higher than the value they’d return.

The 74-store dashboard existed as a need long before it got built. The need didn’t change. The cost did.

I look at a long list of things I’ve talked myself out of building. Most of them, I told myself the cost wasn’t worth it. A few of them, I was probably right. But quite a few, I suspect, I was applying a cost model that no longer holds.

There’s one more implication worth naming, and it’s the one I keep coming back to.

When implementation was expensive, the main discipline was restraint. You had to say no to most ideas because building them cost too much. The skill was knowing what not to build.

When implementation becomes cheap, restraint is no longer enough. The harder discipline becomes judgment: knowing which newly possible things are actually worth making.

The resource that’s become scarce isn’t building capacity. It’s the ability to evaluate what should exist in the first place — to look at a list of now-affordable ideas and ask which ones actually matter.

This isn’t just a question for software developers. It applies to anyone whose work involves deciding what to make, what to prioritize, or what to bring into the world. When the cost of realization drops, the value of good judgment about what to realize goes up.

I’m not sure I’ve recalibrated fully. But I’m sure the old model is wrong.