This year I redesigned the website for Hall Green Chiropody, a podiatry clinic run by my university friend Mark. The case study covers the brand, the shopfront and why the site looks the way it does. This post is about how I built it, because the way I work has changed a lot since I first made Mark a website in 2020.
To get there quickly I used the latest AI tools, including Claude, Replit, Codex and ChatGPT. They let me move fast, while the design decisions and the final say stayed with me.
How the new site came together
- A prototype in Replit. I started by building the new look as a single-page app in Replit. It let me quickly try out a new colour scheme and theme, and show Mark a very high-fidelity, visual version of how the site would work with his new logo. That's what he signed off.
- A design system in Claude Design. I fed the prototype into Claude Design, which was brand new at the time, to turn it into a proper design system.
- Back to the original code. I pointed it at the original site I'd built, which was pure HTML and CSS with simple markup, so the new design built on what was already there.
- Building it with Claude Code. In VS Code, using the Claude Code extension, I pulled all of those sources together (the design system, the Replit prototype and the old site) and started writing a skill that set out exactly what I wanted it to achieve.
- One context file for every tool. Alongside that I kept a context file, synced to CLAUDE.md, AGENTS.md and GEMINI.md, so I can use whichever AI tool suits the job on the same codebase.
The result is a system where any change Mark asks for is quick to make. I can jump into the code myself, or simply describe the change, or screenshot the page and circle what needs fixing, and then deploy it straight away with GitHub Actions.
What I love about Claude Code
What I love most about Claude Code is how flexible it is. I can start a change on my laptop, close the lid and carry on in the app on my phone. I can point it at several repos at once, make the same change across more than one site in a single prompt, and deploy to any of the websites I host, all from my phone.
Why I build with AI
I want to be clear about this part. I know my way around HTML, CSS and JavaScript, and I've built sites with Eleventy (11ty), including this one. But I'm deeply curious about what's new: the latest tools, new features to play with and new ways of working.
I use AI heavily in my day job too, as a product manager in ecommerce, for data, prototyping and design. Staying on top of these tools matters to me, because they help me make the most of my time and get the best possible results for Mark and the other clients I work with.
What AI really gives me is the freedom to build what I can imagine, not just what I already know how to code or have time to research. If I'm honest, the results are better than what I would get coding everything by hand, without all the labour and overhead. These new LLM tools have become something I genuinely love working with.
AI doesn't get free rein
AI can do a lot, but the feeling, the design and the thinking behind a site have to come from me, and I steer the AI to build things the way I see them. Churning out pages at huge scale that offer little value is easy now, and search engines are quick to penalise it. What matters to me is a site where patients find what they need, Mark's business keeps growing, and search engines and AI chatbots find the information genuinely useful.
If you want to see the finished result, have a look at the Hall Green Chiropody case study.