From Idea to Website in 4 Hours

Earlier today, I was sending an email to a nonprofit about a new capability — how I was able to ingest business cards into a CRM using Claude Cowork, saving several hours of manual labor. It then dawned on me: I had an idea for a website — a place to document the useful things I've built with AI, for people like me who aren't software developers. Over the last six months, my Claude skills have been transformational, and I wanted others to see the full, unlimited power of Claude beyond being a chat box. I'm not a programmer. I barely passed Fortran in college. A few years ago, "build a website with search, filters, and a blog" would have meant hiring a developer, waiting weeks, and paying thousands of dollars of real money — or going to a Wix or Web.com, searching for the right template and using their archaic tools to modify it to my needs.

Instead, I described what I wanted to Claude and watched it come together in a single afternoon. Here's how it actually went — including the parts that snagged.

Start with a CLAUDE.md and create a plan. There are two things that help make this process go smoothly. One is to generate a CLAUDE.md file — a contextual file about the project that Claude can reference without my having to explain the context over again. The other was to request a plan from Claude before a single line of code was written. I started by writing a plain-English description of the site: the pages I wanted (About, a "Journey" of articles, a Blog, and Contact), how the articles should be organized, and the voice I was going for. Claude asked me a handful of practical questions — where to host it, whether the Blog was separate from the articles, how it should look — and laid out a plan before building anything. Claude defined five phases for the process, and I was able to review each phase to determine whether it was going in the right direction.

The build. From there it moved fast. Claude set up the site, designed a clean layout, and built every page. When it was time for my homepage introduction, I didn't paste text into a form — I handed over a Word document, and Claude pulled the words straight out of it and placed them on the page. Same with my first real article: a Word file with a screenshot inside. Claude extracted the text and the image, formatted it, and published it. That's my workflow now — I write in Word, and the article shows up on the site.

The features I assumed would be hard. I asked for search across all my articles, filters by topic, and a way to switch between a card view and a table view for scanning a long list. In the old world, each of those is its own small project. Here, they were just requests. Search now finds any word across every article and highlights it, and the site is built to handle a hundred articles without slowing down.

Then came deployment — the messy part. Getting the site actually live is where the friction showed up, and it's worth being honest about it. My hosting company had recently reorganized its dashboard, so the buttons weren't where the instructions expected. At one point I landed on a GitLab sign-in page when my account was on GitHub — easy to do, easy to fix once you know. The setting to attach my domain was hiding in a different tab than I was told. None of it was truly hard, but it was exactly where a non-technical person could get stuck — and Claude was my savior and online tech support. Claude walked me through each screen, told me which button to click, and figured out what was wrong when a screen didn't match the initial instructions.

The last gotcha. When I finally typed my web address into the browser, it told me the page didn't exist. My stomach dropped — had something broken? It hadn't. The site was live everywhere in the world; my own computer had simply memorized the old "this domain doesn't exist yet" answer and kept repeating it. Claude confirmed the site was up by checking it from the outside, and reassured me my browser would catch up on its own.

One honest caveat. This was a simple site — a handful of pages (a home page, an article list, a blog, an About page, and a Contact page), built to grow one article at a time. I don't want to oversell what happened here: this does not mean Claude can build any website. A sprawling site with hundreds or thousands of pages, user logins, a database, and lots of complex moving parts is a different animal — that's genuine software engineering. What I'm describing is the newly accessible end of the spectrum: the kind of clean, useful content site that used to be out of reach for someone like me is now an afternoon's work.

Idea to live website: about four hours. No developer, no code, no invoices. A few Word documents, a running conversation, and a handful of decisions only I could make. And it only cost $11 for the domain registration. That's it — I'm using the free tier on Cloudflare to host it. That's the shift worth paying attention to: not that AI can write code, but that someone like me can go from "I have an idea" to "it's on the internet" in a single afternoon. And I was able to do it all within the $20 Claude Pro plan.