I used AirTalk for about six months before I started building this. It does roughly the same shape of thing as XES — anonymous voice calls with strangers — and on a quiet night the calls were fine. The reason I stopped was less the calls and more what was sitting behind them.
Go and read the AirTalk terms of service properly. There’s a clause in the middle that lets them listen to your audio. Not “recordings can be reviewed if someone reports you” — actually monitor live audio at their discretion. It’s wrapped in safety language, which is the polite version of “we have the technical capability to do this whenever we like.” Once you know that clause is there, you can’t un-know it.
It made sense once I looked at the rest of the stack. AirTalk routes calls through their own servers. In webrtc terms that’s an SFU, which is a fancy way of saying “the audio goes through us, not directly between you and the other person.” If your audio is moving through someone else’s server, that someone else can do whatever they like with it. They mostly probably don’t. But the option exists, and the option not existing is a meaningfully different thing.
The other half was the lack of moderation. I had multiple calls where the person on the other end said something that should have got them banned in any platform that took itself seriously. I reported them. Nothing happened. There wasn’t even a “thanks, we got your report” automated reply. The form felt like it went to a folder no one was opening.
So I started writing this in the evenings.
The two things that were going to be different from the start were the topology and the moderation. XES is peer to peer. The audio goes from your browser to the other person’s browser directly. The only thing the XES server does is help you find each other and then get out of the way. Once you’re talking I’m not in the middle, I can’t listen, and there isn’t a server-side recording because there isn’t a server-side anything. There’s some STUN/TURN infrastructure for getting through awkward home routers, but those are relays of encrypted media. They forward the bits without being able to read them.
The moderation runs through the rating system. After every call you give a thumbs up, a thumbs down, or a block. The thumbs feed a trust score, the score feeds the queue, and people who collect downvotes either lose access to matchmaking for a stretch or stop getting matched with most people. There’s a moderator dashboard for the worst cases. Reports get acted on, usually within a day, because there are about three of us looking at the queue and it’s easier to be fast than to be slow.
That’s the whole thing. No investor deck, no growth team, no roadmap chasing user counts. XES is one person’s reaction to a product that disappointed them, built in evenings, with a handful of people helping out with moderation and feedback. If it stays small forever that’s fine. The point was always that it should exist, not that it should win.
← Back to Blog