Something I keep seeing: a team has a problem that a simple solution would handle well, and they build something complicated instead. Lately the complicated version often comes from an AI assistant. Ask a model to design a system and it will happily give you queues, agents, caches and three layers of abstraction. It all looks serious. Nobody on the team could explain why each piece is there.
That last part is the problem.
Complex feels like better
We humans are drawn to the idea that complex means better. A diagram with twenty boxes looks like more work went into it than one with four. More features look like more value. In my experience it usually runs the other way. At the higher levels of the craft, you gain more by subtracting than by adding.
Complexity has to earn its place.
A simple solution has one quality a complex one rarely has: you can understand it. You can hold it in your head, predict how it fails, and fix it when it does. That’s worth more than most of the cleverness you could add.
Don’t delegate the understanding
I use AI every day. It writes code, drafts documents and checks my reasoning, and it makes me much faster. I use it to leverage what I know. I don’t use it to know things for me. The moment you hand your understanding of your own product to a model, you lose the ability to judge what it gives you.
That’s how teams end up owning systems nobody chose. Each suggestion looked reasonable on its own. Together they add up to something only the model seems to understand, and the model doesn’t really understand it either.
Be the captain
A captain doesn’t row every oar. A captain decides where the ship is going and knows why. With AI, that means you decide the shape of the solution, you can explain every part of it, and you add complexity only when something simpler has failed and you can say why.
When I start a project, I ask one question before anything else: what’s the simplest thing that could work? Most of the time, it does. When it doesn’t, the failure tells me exactly what the next piece of complexity has to do. That’s how complexity earns its place.
Use AI to go faster. Just make sure you’re still the one steering.