AI Builders Psychosis: Succeeding with AI is as much about slowing down as keeping up

A product design influencer recently tweeted that the "double diamond" framework of doing research before building is a thing of the past. With AI, we can now build as soon as we want. We just build build build and we never have to do any discovery.

For context, the double diamond framework requires you to do research - it includes tasks like defining your problem before exploring solutions. There are other frameworks that come down to similar principles - this one is most popular with product designers.

Many people agree with her: Anthropic product engineers are no longer writing code, they are just managing multiple agents, prompting one, and while that one is off doing work, they are prompting another one - managing up to 5 agents at a time (maybe more? Who knows, it's not illegal to be insane.) They are constantly prompting something. If you're not prompting, the permanent underclass beckons.

To me, this sounds like a recipe for disaster. But it's a natural and understandable progression when given a near-magical tool like AI. I call it AI Builder's Psychosis. You're so excited about building you forget to think about what you are building.

I love to build, so I understand the excitement about getting our hands dirty, about having ideas and seeing them materialize. This is actually why I started coding in the first place 20 years ago - I wanted to build, to feel empowered, to craft my vision into something other people could interact with, and ideally, be wowed by.

That wow - that reaction I used to get from people when building a prototype - that was my drug. Then as I started working for larger orgs, I found that making lines go up was similarly satisfying, especially if the line was labeled profit or revenue. But anyway - back to the wow.

The wow is gone. Nobody is impressed anymore. Everything looks the same - everything has 3D, animations, rounded corners, hover effects. Things that used to require time and effort and were risky to implement because they might look terrible (they usually do) are now one short misspelled prompt away. So of course, people stop being impressed.

Now the wow comes from things like OpenClaw. Have an agent do stuff while you sleep. We're never sure exactly what the agent is doing while you're sleeping, but hey - it's exciting. It's controversial. Perhaps the most exciting part is that it's dangerous, and security experts hate it.

But this excitement will pass, too.

After everyone is tired of building slop, we'll have to go back to the thing we said is a thing of the past: thinking. That first diamond in the double diamond framework.

The first "diamond" is about research, thinking, and strategy. It's about thinking hard about what you're building and why, and what goals it needs to achieve. Building without a strategy is fun - don't get me wrong - grab some playdoh and create some random stuff if you're stressed. But if you want to build with purpose - to get something out of people - a reaction - maybe a wow - you have to think hard about it. You have to think about them. What's going on with them? What keeps them up at night? What would bring them joy or delight? What world do they find themselves in? What's exciting? How can we make people feel empowered, competent, confident? How can we help them be knowledgeable, collaborative, or effective?

One of my mentors gave me a really good piece of advice about the dichotomy of thinking and building. And lucky you, I will share it with you, because if you've gotten this far down the blog post, you deserve a little tidbit that will help you do this kind of meaningful building.

My mentor saw that I loved to build. He told me, "you know, that's dangerous. People who love to build might find themselves building for no reason at all, just for the love of it. This may not seem dangerous to you, but if you have goals in life - if you want the things you build to get a life of their own - you can't just be building because it feels good. You need to think."

It's also dangerous at a larger scale because it's a trap that humanity's most capable and creative people fall into. The trick he gave me is simple - don't build and think on the same day. The trap is that you think to yourself "I'm thinking - in between prompts" or "I already thought of this." When you love to build, it's easy to cut corners on thinking.

If you dedicate a day to thinking - no building - your results will markedly improve. It's one of those things that helps you 80% of the way, so all other optimizations are pointless to worry about. Think and build on different days. Even if your AI is building for you. Take a step back. Slow down. Your products will be better for it.