Pat Walls runs Starter Story, a business doing over $1M a year, with 500,000 YouTube subscribers behind it.
And in April 2025 he sat down on camera, dog sunbathing behind him, and made a public bet: he would make $100,000 using only AI build tools. No team. No plan. Filmed from day one.
He called the series "Vibe coding until I make $100K". Everyone else calls it the vibe coding challenge, and Day 1 is the most copyable 26 minutes on the whole channel.
Day 1 ends with a real app, live on a real server, with a name and a domain. One afternoon.
And the uncomfortable part nobody mentions: the domain he bought on camera that day is a dead parked page today. I checked.
So here's the Day 1 playbook first, then what actually happened to it.
The three rules that make it a real challenge
Most "I'll build in public" announcements die quietly because there are no stakes. Pat set three rules before writing a single line:
→ Rule 1: nothing happens off camera. If he works on it, he films it. No secret progress.
→ Rule 2: AI build tools only. Pat was a software engineer before Starter Story. He knows how to code. The whole point is proving he doesn't need to.
→ Rule 3: he doesn't stop until he hits $100,000 or flops completely trying a bunch of different things.
That third rule matters most: a dead project isn't the end of the challenge, just another episode. Keep it in mind, because the series leans on that rule hard later.
The idea came from a $200 invoice
Pat didn't brainstorm for a week. The idea arrived while he was literally sleeping: an AI YouTube thumbnail generator. You draw a rough sketch, upload some reference shots, write a prompt, and get five YouTube-ready thumbnails back.
He picked it because he was the customer, with receipts.
He was paying a designer $50 per thumbnail, four thumbnails per video. That's $200 per video, every video. Plus about 30 minutes per video shooting reference photos to get the right one.
Then he opened a tweet that had been sitting in his bookmarks. A guy named Lane, 8,400 followers, posted a thumbnail-from-drawing demo that pulled 1.9 million views. Pat's take: if that tweet had a product link in the replies, it would have printed $10K MRR from one post.
(His exact words: "this is the one time in history where this may be possible." A 5,000-follower account getting 2 million views on a build demo. That window is why he started now.)
The validation document? One Google Doc, five minutes, two headings:
→ Problem: great thumbnails make or break a video, and they cost $200 a video to get.
→ My edge: a channel with 500K subscribers, an audience full of YouTube creators, and he needs the tool himself.
That edge line is the part most people skip. Pat's logic: anyone can clone a thumbnail generator in a weekend. But when the bugs pile up and support tickets arrive, only the founder who actually feels the problem keeps going. Copycats quit at the boring part.
A $16 domain, a $5 price, and a Stripe link before features
Three money decisions on Day 1, all made in under ten minutes:
The domain. He searched what people actually type into Google ("ai thumbnail maker", "free ai thumbnail maker") and bought the exact-match aithumbnailsmaker.com for $16. Thumbnailcreator.com was $2,800. He laughed and moved on.
He also admitted on camera he was doing "the thing we're not supposed to do," buying a domain before building anything. He did it anyway so nobody would steal the name after the video published. Honest, at least.
The price. $5 a month to start. Low on purpose: against a $200-per-video designer bill, a five dollar subscription is a rounding error. Raise it later once someone proves they'll pay at all.
The paywall. No free tier. Every generated thumbnail hits the OpenAI API, and the API costs real money per call. So the plan was one Stripe payment link in front of the product, before polish, before more features. If nobody pays five bucks, he learns that for free instead of funding a thousand tourists' experiments.
Payment first as validation. One line of code instead of a pricing-page redesign.
Live on the internet in one afternoon
Pat gave himself a hard deadline: launch the skeleton in 4 hours.
The deadline came from a founder named Nico he'd interviewed, who built 17 apps and sold two of them for $265,000. Nico's biggest lesson wasn't a growth tactic. It was forcing himself to ship by the end of the day or the end of the week, or else.
The Day 1 stack, for anyone following along:
→ Cursor as the editor and agent
→ ChatGPT as the step-by-step guide for anything confusing
→ Ruby on Rails for the app, Heroku for hosting
The one move I'd steal from this entire episode: he deployed to a real server before building a single feature. His reasoning is pure software-engineer scar tissue. Deployment is the headache that kills projects at the end, so he moved it to the start, while the app was still a blank page.
And when the deploy came back with a 404, he didn't debug it. He pasted the error into Cursor's agent mode. Cursor found the missing controller, drew the routes, wrote the commit, and pushed the fix to Heroku. Pat clicked okay, okay, okay.
His reaction on camera: it just did the things "I had to spend years learning as a software engineer."
By the end of the session: welcome page live at a real URL, copy centered, product named. Idea to deployed app, one afternoon, roughly $16 spent.
What Day 1 doesn't show you
Now the part that makes this series actually useful instead of just fun.
Look at the next episode titles in the challenge playlist. Day 3: "I Screwed Up… Big Time." Day 4: "I think I need to start over…" Day 5: "I Deleted Everything… and It Worked."
That arc is what vibe coding looks like past the demo. The hello-world afternoon is genuinely easy now. The two weeks after it, when a change breaks something the AI wrote and you can't read, is where most of these projects die. Pat warned about exactly this on Day 1 (his words: one bug can send you "down a rabbit hole of two weeks") and then lived it by Day 3.
The challenge itself followed Rule 3 and kept morphing. By Day 9 it was Ross in the chair, and his first dollar took over 30 days and one brutal launch. I broke that episode down in his failed product launch that got called AI slop. Later the challenge ditched subscriptions for a lifetime deal, which is its own lesson in pricing for cold traffic.
And aithumbnailsmaker.com, the $16 domain bought on camera? As of October 2026 it's a parked page. "This domain may be for sale."
No $100K victory video exists yet either. Starter Story publishes the running revenue numbers at labs.starterstory.com, which is more honesty than most build-in-public accounts ever offer.
So was Day 1 a failure? I don't read it that way. Day 1 energy costs nothing, and every founder I know has a graveyard of deployed hello-worlds that never charged anyone a dollar. What makes this series worth your time is that it filmed the months between the live welcome page and actual revenue instead of cutting them out.
Steal this, skip that
If you're starting your own version of this challenge this week, here's my filter after watching the whole arc:
Steal:
→ The five-minute one-pager. Problem plus edge, nothing else. If you can't name your edge in one sentence, you're building someone else's app.
→ Deploy before features. A live URL on hour one removes the scariest step while the stakes are zero.
→ Payment link before free tier. Especially when every user action costs you API money.
→ The public deadline. 4 hours for a skeleton, filmed, with witnesses.
Skip:
→ Buying the domain first. Pat says it himself while doing it. $16 is cheap, but the habit of buying artifacts instead of shipping is not.
→ Falling in love with the Day 1 idea. The parked domain is the proof. The challenge survived because the rules let the idea die without the builder quitting.
FAQ
What is the vibe coding challenge?
It's a public series from Starter Story Build where the builder commits to making $100,000 using only AI build tools like Cursor, with every working session filmed. Pat Walls launched it in April 2025 with three rules: no off-camera progress, no hand-written code, and no quitting until $100K or a total flop.
Did Pat Walls make $100K with AI tools?
Not yet, as of October 2026. The Day 1 app (an AI thumbnail maker) was eventually abandoned and its domain is now parked. The challenge continued through multiple pivots and builders, with live revenue numbers published at labs.starterstory.com. The first dollar alone took over 30 days.
What tools did he use on Day 1?
Cursor for coding and agent fixes, ChatGPT for step-by-step guidance, Ruby on Rails for the app, Heroku for hosting, a Google Doc for the one-pager, and a Stripe payment link for monetization. Total cash spent: about $16 for the domain.
Do you need to know how to code to run a vibe coding challenge?
Pat's bet is no, and he's a former software engineer deliberately not using that skill. But Day 1 shows the honest version: you still hit 404s, proc files, and database migrations. You don't need to write code, but you do need the patience to paste errors into the AI and ask it to explain what it's doing.
Want the stories after the hello-world high wears off?
Every week on the Profitable Founder Podcast I sit down with bootstrapped founders making $100K to $10M a year and get the numbers everyone else edits out: the pricing calls, the failed launches, the pivots that saved the company.
