A sports creator drove home from a friend's wedding with an idea stuck in his head.
By Monday night he had a working game.
Three months later, that game makes $40,000 a month with 700,000 monthly players.
That's the short version of GeoSports, the viral web game Frank Michael Smith built in about 8 hours with Claude. He can't code, and he spent zero dollars on marketing.
Starter Story just had Frank on, and I took notes like a maniac. Because this is the cleanest example I've seen of where building is going in 2026: a non-technical creator ships a product in a weekend, and distribution does the rest.
Here's the full story, real numbers included.
What GeoSports Actually Is
Five questions a day.
You get a sports clue ("Tailgating superfans throw themselves through tables in this NFL town"), and you tap the spot on a globe where it happened. The closer your tap, the higher your score. Five taps and you're done, usually in 60 to 90 seconds.
That's the whole game. Frank never had to teach a single person how to play it, and he considers that the main reason it worked.
The numbers behind that simplicity, straight from the interview:
- 700,000 monthly active users
- About $40,000 a month in revenue
- $30,000 to $35,000 of that is programmatic ads (around $1,200 a day)
- 1,800 Pro subscribers ($5 a month or $40 a year)
- A 53% organic share rate
Read that last one again.
On a day when 100,000 people play, 53,000 of them share their score. The product does its own marketing. I wrote about this pattern before in apps that market themselves, and GeoSports might be the purest case yet.
Built in 8 Hours With Claude, by a Guy Who Can't Code
Frank is not a developer. He's a sports creator with over 4 million followers, known for short-form commentary videos, not software.
His brother showed him MapTap, a daily geography game. Then, driving home from a wedding, the thought wouldn't leave: what if this had a sports twist?
So that Sunday he sat down with Claude and worked for about 8 hours. He dumped in everything he knew about the game he wanted, had Claude research MapTap, and built a spec. Then he canceled his Monday and kept going until it was good enough to send out.
Every time he got stuck, he took a screenshot and asked Claude what he was looking at. By his own count he took "a ridiculous amount of screenshots" wiring up hosting and domains. That was the entire engineering process.
The stack he ended up with:
- Claude Max for the building
- Vercel Pro for hosting (domain bought there too)
- MapLibre for the globe
- Upstash Redis for leaderboards and stats
- Google Sheets for the questions
- Stripe for Pro subscriptions
Google Sheets as the question database for a game that peaked at 150,000 players in a day. I love everything about that.
If you've been telling yourself you need a technical co-founder before you can ship, this should sting a little. It stung me. One person, one weekend, one LLM.
The Launch: One Reply Did 150,000 Players in a Day
The launch was quiet on purpose. Family first, then friends, roughly 100 people. He watched them share it without being asked, and that was the signal.
Then came the move I'd steal.
Frank posted a normal NBA tweet to his audience. Under it, he replied to himself: by the way, I made this game, would love if you checked it out.
The reply went viral. 40,000 players that day. The next day, a Saturday, GeoSports did 150,000 players in a single day.
The game was 5 days old.
From there it snowballed: Yahoo Sports' Kendall Baker plugged it in his newsletter, Sportico covered it, Jon Gruden posted a video of himself playing (and famously could not find Mississippi), and Frank ended up on national TV twice.
1 million unique players in under a month. Zero dollars of marketing spend.
"But He Has 4 Million Followers"
I know what you're thinking, because I thought it too.
Of course it worked. The guy has millions of followers.
Except Frank came with receipts: his own posts drive about 2,000 clicks a day. That's roughly 2% of GeoSports traffic. The other 98% is word of mouth, people getting a link from a buddy in a group chat with no idea who built the game.
And his audience never saved his earlier games. Five Card Draw, his first one back in 2020, got about 110,000 players and stayed niche. He's been shipping games for 6 years through his studio, Rhino Studios, which raised $500K at a $4M valuation in 2025. Most of them didn't hit.
His audience got the game its first look. The 53% share rate did everything after that.
So yes, build an audience (Frank says the same). Just don't use the lack of one as your excuse.
The Host Hack: Turning Big Accounts Into Free Distribution
This is the tactic I'd frame and hang on the wall.
Frank noticed sports media personalities and athletes with big followings were liking his GeoSports posts. Instead of asking them for a retweet (low value, easy no), he DM'd them and offered a role: be the host.
Hosting means they write the five daily questions. That's it.
When their edition goes live, Frank announces them as the day's host. And almost every time, they retweet it to their entire audience. He says it "still feels like a cheat code" that people are happy to do this, and it doubles as a networking engine with the exact people he wants to know.
→ The steal: find the people already engaging with your product and give them a role in it. People say yes to contributing way more often than they say yes to promoting.
Frank's 3 Rules for Building in 2026
Near the end, Gus asked him what advice he'd give someone building their first project. Frank gave three rules.
1. Build for web first. GeoSports went viral because there was zero friction. People found it on X, played it without leaving X, and shared it without leaving X. No download, no signup wall, no app store review. If your product can live at a URL, start there.
2. Make it sharable by design. The score card formats itself for sharing. That single feature took GeoSports to 1 million unique players in under a month without an email campaign. Decide on sharability the day you design the product, not after launch.
3. Block 8 hours and start. Skip the coding course and the co-founder hunt. Block one uninterrupted day, and when you get stuck, ask your LLM and keep moving. That's the entire playbook he used.
And underneath all three, the lesson that hit me hardest: stay in the game. Frank started making YouTube videos in middle school, around 2007. Nothing really worked until 2020, about 13 years later. He started building games that same year, and they exploded in 2026, another 6 years on.
The overnight success took two decades of staying in the water. I see the same arc in almost every solo game developer story worth telling.
FAQ
What is GeoSports?
GeoSports is a free daily web game that mixes sports trivia with geography. Each day you get five sports clues and tap the location on a globe where each moment happened. The closer your tap, the higher your score. It runs in the browser at geosports.app, with spin-offs like Geo History and Geo Footy under the Geo Games umbrella.
How much money does GeoSports make?
About $40,000 a month as of September 2026, per founder Frank Michael Smith on Starter Story. Roughly $30,000 to $35,000 comes from programmatic ads (around $1,200 a day), and the rest from about 1,800 Pro subscriptions at $5 a month or $40 a year. The game has around 700,000 monthly active users.
Did Frank Michael Smith really build GeoSports in 8 hours?
The first playable version, yes. He built it with Claude over roughly 8 hours on a Sunday, then kept refining it the next day before sending it to friends. He's not a developer: he worked from a spec Claude helped him write, screenshotted anything confusing, and asked questions until it worked.
Do you need an audience to make a viral web game work?
It helps less than you'd think. Frank has over 4 million followers, but his own posts drive only about 2% of GeoSports traffic. The other 98% comes from word of mouth, powered by a 53% organic share rate. His earlier games had the same audience behind them and stayed niche.
What should I build my web game with if I can't code?
Frank's stack is a good template: Claude for building, Vercel for hosting, MapLibre for maps, Upstash Redis for leaderboards, Google Sheets for content, and Stripe for payments. Total setup cost is a rounding error. Tools stopped being the constraint. Starting is.
Hear Founders Like This Every Week
Frank went from idea to 1 million players in under a month, but the setup took 19 years of shipping.
That's the part most people skip, and it's exactly what I dig into on the Profitable Founder Podcast: bootstrapped founders sharing the real numbers behind wins like this one.
