AI in practiceLinkedIn post

The reward loop of AI coding tools

A personal observation about the variable rewards of AI coding tools and the importance of noticing compulsive iteration.

People say they're “addicted” to AI coding tools.

I don't think that's just a figure of speech.

The loop has the same structure as a slot machine:

  • Rare big rewards (it solves something perfectly oneshot)
  • Frequent small rewards (one step closer, one step closer)
  • Near misses (almost right, just one more prompt)

That's the variable reward schedule. The same mechanic behind gambling addiction, loot boxes, and some of the more predatory corners of e-commerce.

What makes it more compelling? With slots, failure is clearly random. With AI coding, you have more control. It's your prompt, it's your docs, it's your harness. So you iterate. And iterate. And iterate.

I use these tools daily and they genuinely change what's possible. But I think we should be honest about the mechanisms that drive compulsion. So we can avoid the pitfalls.

Is this something you've noticed in yourself or your team?

More notes on engineering, AI and the systems around the work.

Back to all writing