Coding or Gambling?
Eighteen months ago, I struggled to spend $20 on an app I knew I would use every day. Today, I barely blink at giving Claude Code $25 a few times a month to help me make apps that mostly do not work. That mismatch feels irrational. The value proposition is clearly worse. So what changed?
Part of the answer is obvious. Understanding AI coding tools matters more to me than almost any single-purpose app ever could. These tools are reshaping my profession. I am asked to talk about them constantly, to recommend them, and to help teams use them well. Spending money to stay fluent is defensible. Rational, even.
But that explanation does not fully hold up. I could do all this learning at work, on a work machine, on the company’s dime. No one would object. In fact, they might encourage it. So the motivation is not just professional diligence. Something else is pulling me toward swiping my own card, late at night, tinkering on half-baked ideas.
Here is my hot take. Making software is far more fun than using software. AI coding tools sell that feeling perfectly. They are cheap enough to feel harmless, but they invite endless pulls of the lever. Maybe something great comes out. Maybe it does not. Either way, the act of creating, of being “in the vibes” with a model like Claude, scratches the same itch that got many of us into programming in the first place. The uncomfortable question is whether that itch is now being productized into something closer to a slot machine than a tool.