Chris Wharton has been building things on the internet since before most people knew what the internet was. Framesets. Custom CMS systems. Clients like Coutts Bank, RBS, IKEA, Grant Thornton and RAC. Fifteen years running his own agency, 2nd Floor. Seven years building Ready Steady Websites - a website subscription model for small businesses that is now evolving into a full website, automation and CRM offering.
When Claude Design launched, I thought of Chris immediately. Not because I expected him to be worried - but because I wanted to know what someone with that much real-world design experience actually thought of it. The answer was more interesting than a simple verdict either way.
"If you look at Instagram and TikTok, all the posts say designers are dead, SEO specialists are dead, everyone's panicking," Chris told me. "But when I first got hold of ChatGPT and tried to get it to do design work, it was literally stick man stuff. I thought, my job's safe."
Then he found Claude.
What changed his mind wasn't Claude Design specifically - it was Claude Cowork. As a designer who has always done some front end development, Chris found the Cowork environment a natural fit. "You've got a workspace where you can see things happening in the chat and then it's showing you stuff in the info pane on the right. You can develop and iterate as you go. It's a really good middle ground for a designer."
On Claude Design itself, Chris is direct. He tried to set up his Ready Steady Websites design kit through it and burned through his token allowance almost immediately. His more instructive test was a client PowerPoint rebrand - he gave it a basic slide deck, his written brand guidelines and an example one-sheet.
"It did a beautiful PowerPoint presentation. Took me five to ten minutes. Would have taken me three hours. But then I downloaded it all excited and it was just static slides - I couldn't even edit the text."
He went back, asked for editable slides and got them - but the text positioning was off throughout. His conclusion: "Claude Design as a standalone app is just in its infancy. I'm quite excited to see where it goes. But right now I actually think you get better results in Claude Code or Claude Cowork."
The deeper observation though is more interesting than any single test result.
"Claude out of the box does pretty good designs. But they're always samey. You can get a modern SaaS brand or a prestige brand but they all feel the same - same fonts, same spacing, same colour patterns. AI can do pattern recognition. It can repeat what's been done in the past. But if you're a frontier designer coming up with genuinely new concepts, AI isn't there yet. It can't do new thinking. It can only regurgitate old thinking."
That, he says, is actually reassuring for designers working at the higher end. The threat is real for commoditised, run-of-the-mill work. For original creative thinking, the gap remains.
The one thing AI still can't do? New thinking. It can regurgitate what exists. It cannot conceive what doesn't.
What's changed more dramatically for Chris is what he can now build.
"I've always had ideas I could never really realise. Those limits are now gone. I can vibe code things, check through them, make sure they look good and ship apps whenever I want. I feel like I've plugged into the Matrix."
He pulled out a handwritten list of everything he'd built with Claude since starting on 5th March - a page and a half. Web apps, static sites, SEO tools, data scrapers, a Chrome extension, PowerPoint presentations, scheduled tasks and branded onboarding documents. "I've done probably six months worth of work in about six or seven weeks through Claude. Some of which I never ever would have been able to do on my own."
The first thing he built was a cache warmer app - something he'd had an idea for years. A tool that pings client websites to keep them responsive on first load. "I got Claude to build it completely. It walked me through the whole process, told me what services to use and what to sign up for. And it just worked first time. No issues. It's still working now."
The Chrome extension came later - and that's the one that still surprises him most. Built entirely with Claude, guided step by step through the architecture, the submission process and the Google Chrome extension store approval. "Claude even told me what to write and how to get it submitted. It's now approved and live. I just say to customers, search for it in the Chrome store and install it."
That extension is now available in the Google Chrome Extension Store.
Chris's process has become methodical. Before building anything he gets Claude to ask him questions until it fully understands the brief, then generates a markdown spec document. That spec file travels with every new chat window - solving the context compression problem that trips up a lot of vibe coders.
"The more you put into a single chat window the harder it gets to process. So I take that summary file, start a new chat and it knows exactly what's going on. I get Claude to update that file every time it finishes work."
On model switching: he uses Opus for planning and Sonnet for implementation. "I did some planning yesterday and I blitzed my session allowance in 30 minutes on Opus. So I get it to plan, ask if Sonnet can handle the implementation and hand it over. That stretches your token budget significantly."
Chris's background gives him a different lens on security than most vibe coders. He's worked with major corporates including RBS and Coutts Bank where penetration testing was standard practice on every system built. He takes it seriously in a way that most people building with AI currently don't.
"It's going to be the Wild West out there pretty soon with all these apps coming out. If something's been built by someone with absolutely no technical knowledge, it will be pretty much hackable."
His practical advice: before deploying anything, ask Claude directly what security checks it has done, where API keys and environment variables are stored and whether there are brute force protections in place. "It will go okay, I'll do this, I'll do that - and it will make it more secure. But you have to ask."
His broader point echoes something I've been saying since my own piece on vibe coding security. The YouTube tutorials showing how to build apps in an afternoon rarely mention what happens before you deploy it. "If you're not talking about it, did you even think about it? And if you didn't, what does that mean?"
We're planning a follow-up session with Chris specifically on security for vibe coders - what the basics look like and what non-technical founders should be checking before they put anything live.
When I asked Chris what he'd say to non-technical founders who are building with AI or thinking about starting, his answer was simple.
"Don't aim too big too quickly. If you've got a wild, world-changing idea, that's fine. But don't try to build it from day one in one prompt. Take a portion of it, build that out, see how it works and iterate. There will always be learning outcomes from each attempt. Start small."
And on AI more broadly - whether to engage or stay on the sidelines?
"If you're not doing it, your competitors are. Whatever business you're in. Even if it's just an accountant - you still have systems, patterns and repeatable processes. Those are exactly the things to give AI. It can save you hours, make you more efficient and help you take on more work. But find what works for you, build it into your workflow and start with that. Don't chase the shiny stuff."
You can find Chris at readysteadywebsites.com and readysteadywebsites.com/handbook. Follow him on Instagram and TikTok at @cdwharton and connect on LinkedIn.
Building with AI and want to connect with other founders doing the same? Come and join us at Vibe Coding Lab - it's free.
No comments yet. Be the first to share your thoughts.
More from Stories
He co-founded OpenAI, fell out with it spectacularly, sued it, tried to buy it and then built a rival. Here's the story of Elon Musk, xAI and Grok - and why it matters for anyone following the AI space.
Read Story→
It started as a non-profit with a $1 billion pledge and a mission to save humanity from AI. A decade later it's worth $850 billion, its founders have scattered and its CEO was fired and reinstated within five days. Here's the full story.
Read Story→
Most of us are deep in the Claude rabbit hole. But how many of us actually know the story of the two siblings who built it - and what they stand to gain if Anthropic goes public later this year?
Read Story→
