01 Sharing an idea online gets you everything except an argument.
You post a take and you get likes, silence, or a pile-on. What you almost never get is one person taking the other side and answering what you actually said. That back-and-forth is the whole reason to share an idea in the first place, and the internet mostly lost it.
The thing I wanted was the Samantha from Her. Not to fall in love with. To challenge me. A voice that takes the other side, answers my points, and tells me straight when I lost.
02 A bit of structure is the whole trick.
One question. One person on each side. A clock that keeps turns fair, and a judge that explains the decision when it ends. That is all the structure a real dialogue needs. It is just enough to make two people answer each other instead of talking past each other, which is what separates an argument from a comment section.
03 Open to anyone.
You do not need any background to walk in. The questions run from everyday clashes to the heavy stuff, the rounds are in plain language, and the judge says who won in words anyone can read. Most rounds are between strangers, and that is the point: the best argument you will have this week is with someone you have never met.
04 Made to be watched.
An argument is better with a room around it. Rounds are public, spectators pick who they think wins, replays land on Watch, the ladder keeps score, and the best rounds get broadcast live. Sharing an idea should have an audience, and so should defending it.
05 Why this exists, in order.
06 Open problems I'd take help on.
- Realistic interruptions. A voice argument needs real interjections: sub-200ms barge-in, a sense of when to yield the floor and when to hold it.
- AI judging consistency. The same question + same speeches should not produce a different decision brain-to-brain. Multi-brain consensus is a stopgap, not a solution.
- Multilingual rounds. Hindi, Mandarin, Spanish, argued and judged natively. The infra is there; the register isn't.
- Two people at the same hour. Every matchmaking problem on the site is really this one. Concentrating demand without faking a crowd is the hardest open problem here.
- Memory between rounds. The AI should remember how you argued last week and play you differently this week. Style fingerprint exists; round-to-round adaptation is thin.
07 What I believe.
A good argument is a beautiful thing, and it is rarer than it should be. Arguing out loud is undervalued. AI should pressure-test thinking, not flatter it. The product is built around the live round because that's the room I came up in. Everything else, eventually, runs on the same engine.