Dump your website and start again
I press a button, I talk. It types, designs and deploys.
I was running a workshop this week. I spoke to my laptop for a few moments. Out loud, in front of the room, like a man ordering a taxi. I asked Claude to build me a new page for my website. I briefed it, kinda. Three or four minutes later there was a new page on my website. Content, images, the UX properly done, all in my brand.
Done. Live.
The interesting part of this, is what’s missing.
There’s no CMS. No admin screen, no plugins, no Squarespace workspace where the big creative decision is which template comes closest to the thing in your head. I killed all of that.
The website is code. Claude runs on my laptop and can read and write that code. A folder of files that lives on GitHub, which keeps a copy of every version that has ever existed. It is hosted on Vercel but Claude talks to it, not me.
I press a button, I talk, it types, designs and deploys. That’s the whole machine. None of it is exotic or expensive. Speed without quality is just a faster way to embarrass yourself, so the quality lives somewhere specific: skills. A skill is a brief written down once, properly.
My brand skill holds the colours, the typefaces, the spacing rules, what a hero section is allowed to do, how the chatbot speaks, what good UX looks like on a phone. When I ask for a new page, Claude Code reads that skill before it types a word. So I’m never asking an AI to guess what good looks like. The guessing happened months ago, with me in the room. On the day, the sentence is just an instruction.
And once the machine is in place, the relationship with your own website changes. It stops being the thing you’re afraid to touch and starts becoming an active part of your marketing.
If you want a landing page, say you want a landing page. If a new client needs a private section behind a password, ask for that. If the home page hero feels static, make it a rotating hero. Gate a piece of content. Ungate it the following week when you change your mind.
If you want ten chatbots, one for each section of the site, you can have ten chatbots. I’m not saying that’s a good idea. I’m saying it’s an afternoon. The small stuff barely registers. More space between two sections. A new line in the chatbot. Show your new product. Or your first book. Bestseller by the way…
A case study added the day it happened. Each one is a few sentences. None of them is a project.
There’s a safety net under all of this too. Because the website is code on GitHub, every change is a saved version. If I try something and it’s wrong, I roll it back to how it was an hour ago. Experiments become cheap. And cheap experiments are how websites get better, because the usual reason a company site stays mediocre is that touching it feels dangerous.
There is a cost. Not money. I can build a website in a few days.
The costs is giving up old website. A sunk cost. The instinct is to keep the CMS and bolt the AI onto the side, given you just redesigned the thing. But that won’t work. Bite the bullet and kill the old technology first. Before your competitors do.
Dump the website. Start again.
