The gold rush and the undo button

There's a gold rush on in my corner of the world, and like most gold rushes it's loud, a little reckless, and genuinely thrilling to stand inside.

The thing that happened is simple to say and large in its consequences. You can now talk to an AI and have it build and change a real website for you, the kind that runs something like four out of every ten sites on the internet. Not generate a mockup. Reach into the live thing and change it. For a lot of people who were locked out of their own websites, who had to file a ticket and wait three days to move a button, this is close to magic.

I'm one of the people building in this rush, so I'm not throwing stones from outside it. I believe in it. But I keep noticing something in how we're all behaving, and gold rushes have a way of setting their habits early and living with them for a long time.

Almost everyone is selling speed. Faster, easier, watch it build a whole site while you drink your coffee. The demos are real, the speed is real. But there's a quiet thing underneath all that speed that hardly anyone talks about, which is what happens when the machine, moving that fast, gets something wrong on a site real people are visiting.

Because it will. Not because the AI is bad, but because it's fast and it doesn't always understand the place it's working in. And the faster you let something move, the more it matters whether you can take a step back.

I've come to think the undo button is the most underrated thing in this whole moment. Everybody wants to talk about how much the AI can do. Almost nobody wants to talk about how cleanly you can come back from what it did. Those aren't the same conversation. One is about power. The other is about trust. You can have a lot of the first and none of the second, and what you get is a tool that's exciting to demo and frightening to use on anything you care about.

What I want, and what I'm building, is a world where you hand the work to the machine and don't hold your breath. Where the answer to “what if it breaks something” isn't “well, you have a backup somewhere,” it's “press the button, it's back, keep going.” That sounds small. I think it's the whole game. The freedom to move fast comes from knowing you can undo, not from pretending you'll never need to.

I wrote the longer, more technical version of this for the people actually building these tools, where the agentic WordPress landscape is going and what I think we owe the sites we touch. It's on the Respira blog: AI can edit your live site now. Can it take it back? This is the human version of the same worry.

There's a temptation in a gold rush, and I've felt it in myself, to describe what you've built in the biggest words available. To say “never” when you mean “usually.” To say “anything” when you mean “the things I tested.” Nobody decides to do this. It just happens, because the big words sell and the precise ones sound timid, and everyone around you is using the big ones too.

I've started to believe the opposite is true. In a room full of people shouting the biggest version of their promise, the rarest and most valuable thing you can do is say the exact version. This is what it does. This is where it stops. Here's how you check. It sounds quieter. It lasts longer, because it's the only claim in the room still standing after someone leans on it.

I learned that the hard way recently. I went to check whether my own undo actually worked on the messy cases, instead of assuming it did because I'd written it to. It didn't, in two specific ways, and one of them was the bad one, the safety feature itself destroying data in a particular sequence. I found it on a throwaway site because I went looking for the failure. If I'd trusted my belief, I'd have found it in someone's email.

So that's the small thing I'm hoping for, for this whole noisy beautiful moment. Not that we slow down, the speed is the gift, take the gift. But that we build the way back as carefully as we build the way forward. That we treat coming home from a mistake as part of the work and not an embarrassment to hide. And that we describe what we've made in true words, even the unflattering ones, because the people trusting these tools with their livelihoods deserve to know exactly what they're holding.

The fast part, we've figured out. The careful part is the work that's left. I think it's the part that decides who's still standing when the rush is over.

- mihai
founder respira.press