The Hammer

Something changed when I could test an idea in ten minutes instead of three days.

Not just the speed. The way I think when I'm building.

When iteration is slow you make a decision and you live with it. You spend two days writing something, and by the time it's done you're attached to it. You've put enough in that throwing it away feels like a loss. So you don't throw it away. You ship it and hope it works, because the alternative is starting over and you don't have the time.

Vibe coding removed that. Or at least it can, if you use it right.

I write code. Properly, not just Framer. So when I started using AI tools in my workflow I wasn't looking for something to do the thinking for me. I was looking for somewhere to put the thinking faster. There's a difference. The ideas still come from me. The logic is still mine. I just don't have to spend three hours translating a clear thought in my head into syntax before I can find out if the thought was any good.

That's what iteration speed actually means. It's not about shipping faster. It's about finding out sooner whether what you're building is worth building at all.

The best developers I know have always done this mentally. They hold an idea loosely until it proves itself. They don't fall in love with their own code. They delete things without grieving. That's a discipline that takes years to build, partly because slow iteration makes it expensive to practice. When every test costs you a day, you stop testing things you're not already fairly sure about. You get conservative. You stop exploring.

Fast iteration makes exploration cheap. You can follow a hunch without betting a week on it. You can try the version you don't think will work, just to know for certain. You can stay curious further into a problem than you could before because the cost of being wrong just dropped.

That's what interests me about this moment. Not that non-coders can now ship things, though they can. Not that development teams can move faster, though they can too. It's that the relationship between thinking and building got tighter. The gap between having an idea and knowing if it works narrowed to almost nothing.

What you do with that is still up to you. The tool doesn't make you a better thinker. It just means your thinking gets tested faster. If the thinking is shallow the feedback comes back quickly and you learn that. If it's good you find out sooner and you build on it.

I've started treating AI generated code the way I treat a rough draft. Something to react to, not something to ship. I write a prompt the way I'd sketch an idea on paper - loose, directional, not precious about it. The AI gives me something back. I read it, I understand what it did, I change what needs changing. The first version is almost never right. But it's something to push against, which is more useful than a blank screen.

That's the part the discourse keeps missing. Everyone argues about whether the code is good enough to ship. That's the wrong question for how I actually use these tools. The question is whether having something to react to immediately makes me a better builder than staring at nothing.

For me the answer is yes. Not because I trust the output blindly. Because reacting is faster than originating, and the faster I can react the more ground I can cover in a day.

The hammer is the same hammer. I just spend less time swinging it at the wrong nail.