AI For Noobs, AI For Experts

There are two dominant types of AI tools:

• Those who promise to make an expert vastly more effective

• Those who promise to do things for a novice, no expertise required

Think Claude Code (for the experts) versus Loveable (for the novices).

If you're building an AI tool, or integrating AI capabilities into your existing tool, you'll want to be absolutely clear on which of those two you're aiming for, and be clear about the risk/reward tradeoff: A tool that completely replaces an expert would have higher economic value, but it needs to be able to complete a task one hundred percent, because no matter how little work would be left to do, the novice can't do it. An expert-enhancing tool, however, just needs to be somewhat useful. The only drawback is that only experts would buy it.

If I had to put money into a startup in this space, I'd favour those working on "AI for experts". It's still a very hard task, but it's more tractable, with a more gradual on-ramp to value, than "AI for newbies".