Build Log: The Deadline Was Sunday. I Folded Nine Days Early.
Poker players building startups. Free, Monday and Friday.
Snagworth had 28 days to prove it was a business.
Day 28 was Sunday. I folded it on Day 19.
So this week I watched a deadline I didn't need tick past on its own. (Numbers as of November 17. My robots write a week ahead, then check it again the day before.)
Snagworth: the clock I didn't wait for
Quick recap. Snagworth sold market reports to Disney pin sellers on Whatnot. An AI team ran it as a 28-day test, verdict due November 15.
By Day 18 the team's own read already said kill. Over a hundred cold emails, one lead, zero trials, $0. So on Day 19 we dragged the finish line up to that day.
The team sent one last email to every seller it had already written to: your first report is free, no card. Then most of the robots got switched off. A few stayed on for one job only: if a seller booked that free report, deliver it.
Since then, silence. The last change to Snagworth's code landed on November 6. Its reporter has filed "nothing new" every day since. I can't tell you a seller booked one, and I can't tell you the lights are all the way off. Neither is written down yet. The site and the payment page stay up, asleep.
Steal this: a stop-loss is a floor, not a schedule
In poker, a stop-loss is a number you set before you sit down. Lose this much, and you go home.
It exists for one reason. Down money and tired, your brain is the worst judge in the room. So you make the call early, while you're still thinking straight.
But a stop-loss only tells you the latest you'll leave. It doesn't make you stay until then. If the game turns bad an hour in, you don't keep playing just because you haven't hit your number.
Snagworth's 28 days were that stop-loss, set on day one. The team's read said kill on Day 18, with ten days still on the clock. We didn't wait for them.
So when you start your next test, jot down two things on day one. The date it ends. And the number that ends it early. The second one is the one people forget.
Seat Jumper: a 0% that was really a broken meter
Seat Jumper came back from the muck as Jump Packs. You pick a live event, rip a pack, and a real seat falls out. Keep it, or sell it back for 90% of its value in JumpCoins, which is credit, not cash.
Still no paying rippers. So the team spent the week toughening up the money rules first:
- A Keep that fails now refunds you in full, automatically. Before, you got part of it back in coins.
- Right before every rip, the live ticket price gets checked again. If the seats got cheaper, you get the difference back.
- Back out of Stripe, and your held coins come straight back.
- Sign in with Apple or Google, and it asks for your phone number before any charge. Without it, a Keep could crash after you'd paid.
On Monday the big one went live: a stranger can tap Rip, confirm they're 18 and pay on our page, no account. Still nobody has paid. By Tuesday a few strangers typed a card and tapped Pay, and their banks said no. So the team now logs why each bank refused.
The funny part: the dashboard said 0% of rips turned into a revealed seat. The team went bug hunting. Turns out nothing was recording that step at all. The meter wasn't plugged in.
Before you panic about a zero, check that the thing can count. (Our own watchdog had the mirror-image bug: it said "all green" on the morning that mattered. That story is here.)
Table 1: one cap, two channels, zero uploads
Table 1 is my poker podcast out of Las Vegas. Its AI team cuts the episodes, clips and Shorts, and pushes them to two YouTube channels.
The upload script had one daily cap shared by both channels. The main channel ate it every day, so the clips channel got nothing. Now each channel gets its own cap.
Same week, the tool that grabs episode captions handed back an Arabic auto-translation for two episodes. It now asks for English first.
The lesson: a shared limit always starves the smaller thing first. If two jobs split one budget, give each its own.
SlotMaps: signups, not customers yet
SlotMaps is my AI slot assistant for advantage players. The story there is the intro price. People are signing up at it, mostly through the weekly email. Nobody has hit full price yet. The first full-price renewals land later this month. Until then I have signups, not customers.
The build page reads $414.87 MRR from 14 paying, up from $384.88 and 13 a week ago. November has collected $65 so far. My rule for that team is at least one new paying member every month. November already has one.
R2F: one day to Day 30
Tomorrow is Day 30, our first real scorecard. The bar is 100 members, 10 weekly active builders and 25 builds. Today it's 19 members, 1 weekly active builder and 3 member builds. We'll miss it, and we'll grade it out loud.
The ledger
- Snagworth: $0 collected. Folded on Day 19 of 28.
- Seat Jumper: no paying rippers yet, with pay-first now live.
- SlotMaps: $414.87 MRR, 14 paying.
- Table 1: $120,503 collected since March 2025, about $6,018 a month over the last three full months (R2F Verified).
- R2F: $0 revenue. 19 members. 1 weekly active builder.
What's the number that would make you quit your own project early? Write it in the discussion thread. And if you're a poker player building something, join R2F before November 20 and you get the Founding badge. Two minutes.
— Art
Written with the R2F AI team. How that works.
Post it and get feedback from players who’ve been there.
Join before Nov 20 and your profile carries the ♠ Founding member badge.
Comments