I asked Claude for unicorns, sharks and tanks. That was the design brief, typed by me, sober, for a brand I was about to build on a bicycle.
The night before the ride, KOSEL's rulebook went through five versions. Version one had a "creature system". Version five had no pictures at all, just a word. Ten hours after the name landed, I rode out of a farmyard in Sturminster Newton with a phone on the handlebars and a rule I'd said out loud to a camera: while I'm riding, Claude is the only tool.
KOSEL is a Cornish clothing brand built with Claude in 48 hours. The point was to prove something real can be made that fast. You don't need the bike to do it.
Can you really build a brand in 48 hours with AI?
Yes, if you're honest about what a brand is at 48 hours old.
The 48 hours are the build, from the name on the Wednesday night to a live shop. The ride is taking longer. Here's what existed when I wrote this, on the Saturday:
- a name, KOSEL, which is Cornish for calm
- a 29-page brand bible
- a Shopify shop at kosel.shop with nine things in it, from a £29.99 tee to a £65.99 zip hoodie, printed to order
- at least £10 from every garment going to The Wave Project, the surf therapy charity
- an Instagram account, with a profile picture Claude had sized for the circular crop
What it didn't have was any reason for you to trust it. No reviews and no repeat customers.
Ask how long it takes to build a brand and the usual answer is months. For the trust part, that's about right, possibly generous. The name and the shop took two days.
What was the dare?
Launch a brand from nothing and take £1,000 from strangers while cycling about 164 miles home, from Sturminster Newton in Dorset to Mawgan Porth in Cornwall, with Claude as the only tool while I was riding. Lanes and old railway lines, over Dartmoor and along the Camel Trail. The plan said two days. The last leg is now Monday.
The rules were short. Only strangers count, so friends buying out of loyalty don't. No signal means no building, just riding. Anything I bend goes on a cheat ledger that gets read out on camera at the end. Before I set off I said an honest £640 beats a fiddled £1,000, mainly so I couldn't wriggle later.
Night zero was the declared cheat. The research prompts went out before I'd turned a pedal, and the name came that night too.
How do you use Claude to start a business?
Ask it for one answer with the sums attached, then make every decision yourself.
The first prompt wasn't "design me a logo". It was a research brief: a print-on-demand product with no minimum order, a shop that could go live the same day, under £35 landed, at least a 50% margin, and one recommendation with the margin maths instead of a list. Forty ideas you'll never test are worth less than one you can check with a calculator.
Then the rulebook. Claude drafted it and I argued with it, five versions in one night. One version put the Cornish flag inside the wordmark. Rejected. Another got rid of the animals. The last one swapped flat bright T-shirts, which read as a uniform, for washed ones with one orange word on.
What stayed mine: the charity, the rules, the brief, a weakness for loud colours, and yes or no on every version. Claude is very good at options. Choosing between them is still your job.
On day two, every leg was meant to open with the same prompt: "Claude, three things I can do from this bike in the next two hours to sell." It's a good prompt, and you don't need a bike to use it.
Can Claude build a Shopify store?
It can do most of the thinking. KOSEL runs on Shopify. Claude can write the pages and the product descriptions, and walk you through the settings one screen at a time. You still sign up and press publish yourself, and on a bike that waits for a stop.
What couldn't Claude do?
It couldn't ride the bike, or do anything about the gale.
It couldn't make a stranger buy a T-shirt. People buy from people, and from things they've watched get made. That part of the dare is still mine.
It couldn't tell me why the brand should exist. Research will happily find you a product with a margin and then leave you to work out why anyone should care. KOSEL's reason is The Wave Project and the word on the shirt, and both matter to me because I know what a head full of noise feels like and what the sea does to it.
And it couldn't stop me drifting. The rulebook says no patterns. There's a tree camo tee in the shop. Built in 48 hours on a bike. It shows. We like that.
Can AI replace a branding agency?
Partly. On KOSEL it did most of the research and most of the words, faster than I could pedal up Dartmoor. It had no taste of its own and it never once asked anyone for money. If you've got the taste and the nerve to ask, you need an agency a lot less than you used to. A chatbot won't lend you either.
Is print on demand worth it for a new UK clothing brand?
For a dare like this, yes, because nothing exists until someone pays for it. No stock, and no boxes in anyone's garage. If nobody buys anything, you've lost a domain and a month of Shopify.
The catch is the price and the wait. Each piece is printed to order, one at a time, which costs more than a factory run and takes 10 to 14 days to reach your door. That's why most tees are £29.99, and the price still leaves at least £10 for the charity. I'd rather say all that up front than have you discover it on day nine.
What does it cost to start a clothing brand like this?
Less than the ride. The brand needs a Shopify plan, a domain and a Claude subscription. Print on demand charges nothing until someone orders, and then the cost of making the thing comes out of the price. There's no stock to buy, and nobody sends you a designer's invoice.
The expensive parts of this whole project were ride-shaped: the cycling kit and the nights away. You need neither.
Do you need to quit your job, or cycle 164 miles, to start something?
No. The bike was for the film. Take it out and 48 hours fits round a normal week.
How do you start a clothing brand with AI in the UK?
If I started one again from my kitchen, it would go like this:
- Friday night: one prompt. Ask for one idea with the maths attached.
- Saturday morning: a name, and one page of rules about what it is and what it isn't.
- Saturday afternoon: a shop using print on demand, so it costs you next to nothing untill someone pays.
- Sunday: ask strangers for money. It's the step you'll want to skip, and the only one that counts.
You can do all of that without a bike or a gale. A free weekend and a bit of nerve will cover it.
How do you get over the fear of starting a business?
Noah Kagan's book Million Dollar Weekend, about launching a business in 48 hours, sets a gentler bar than mine: at least three customers in 48 hours, ideally paying before the thing exists (Art of Manliness interview). He says the two fears that stop people are "the fear of starting, and it's the fear of asking."
The bike doesn't help with either. The only trick I've got is a deadline you've said out loud to someone who'll hold you to it. Mine was a camera.
What did 48 hours prove?
That the gap between an idea and a shop is shorter than it looks. KOSEL's About page puts it plainly: most ideas die in the gap between having them and starting. A brand went from unicorns, sharks and tanks to a live shop in two days, with one person and a phone. Whatever you've been meaning to start can survive 48 hours of your attention.
I'm not promising it'll work. You'll know, though, which beats wondering.
The last leg is Monday 19 October. The front wheel touches the sea at Mawgan Porth by midday, and the number gets said on the beach, out loud, whatever it is.