Two ways to design with AI

The first is to let AI design for you. You prompt, you accept what comes back, then you ship it. In this way, you're not learning how the code works, and you're outsourcing the design thinking. It's fast and easy, and that's why you see so much slop in your feed.

The second is to use AI as a tool for designing in code. You come up with the ideas, then use agents to help you implement and iterate on them in production. This way is much harder because you need to master a new tool (and when to use it), plus get familiar enough with code to work in production. But once you're over the learning curve, it's more direct and raises the quality ceiling of what your team actually ships to users.

I think this is one reason designers are so split on AI. The typical AI-generated interface is obvious (to designers), and because it's everywhere, we fixate on it. The second type of work is less visible because it shows up as designers solving user problems faster and shipping more polished, interactive interfaces.

We should spend less time getting angry at the first way and more time paying attention to the second.