Scoreboard · argue and defend
You Know Ball
Sports takes, scored honestly, by something that actually watches.
Web demo is public. Builds 26 and 27 were accepted by Apple, but reached no tester
Verified 2026-07-20 · 3 sources
- Builds 26 and 27 were accepted by App Store Connect
- Apple's own 409 responses: the 2026-07-16 upload log reports previousBundleVersion 26, and the 2026-07-19 log reports 27 — Apple naming builds it had already accepted
- No delivery receipt or tester assignment survives
- No success log, no App Store Connect record on this machine; processing state and group assignment unverified
- The iOS work is unmerged and unbacked-up
- banter-bot-content-expansion has no git remote configured; main sits at cf74ac4
The human problem
Sports takes are the most passionate opinions most people hold, and they evaporate into group-chat noise within an hour. Nobody keeps score. Nobody has to defend anything. The most fun argument you had this week left no trace.
The thesis
The fun isn't in being told you're right. It's in being made to defend a position by something that knows ball and doesn't flatter you. No participation trophies.
Where it actually stands
- The web demo is playable right now, in a browser, with no account. Drop a take, the debate engine counters, your argument gets a transparent score.
- The engine is deterministic and mechanically neutral across five sports — no model deciding who wins, and a score a player can reconstruct.
- On iOS, builds 26 and 27 were uploaded and accepted by App Store Connect. I never confirmed they finished processing, never assigned them to a tester group, and nobody has installed either one.
- None of the recent engine work is merged to main, and the repository has no remote — it exists on this machine and nowhere else.
- Builds Apple accepted
- 2
- Reached a tester
- 0
- Outside players
- 0
- App Store review
- None
Decisions and tradeoffs
Made the engine deterministic instead of generative.
A scoring system you can't audit isn't a score, it's a vibe. If a player can't reconstruct why they lost, they stop caring about winning.
Tuned it until blind play loses badly.
Early on, someone who knew nothing could win often enough that the score wasn't measuring anything. Closing that gap is the whole product, and it's the work I'm proudest of and least able to show you.
Betting guardrails from day one.
I spent three years in sportsbook operations. I know exactly which sentence turns a game into something I don't want to have built.
What building it taught me
I uploaded two builds to Apple and then never took the last step of putting either in front of a person — and until I went looking for evidence, I'd have told you confidently that nothing had ever been uploaded at all. Not knowing the state of your own release is its own kind of failure.
What isn't true yet
- Never confirmed past Apple's processing step, and never assigned to a tester group.
- Not submitted to the App Store.
- No outside player has installed the iOS build. Confirmed testers: zero.
- The engine numbers I'd want to quote here — cohort win rates, tournament results — I can't currently point at an artifact for, so I'm not quoting them.