shoshshi

Status: live. @shoshshi_bot  with the Mini App at shoshshi.haywan.uz . 200+ users, and I seeded none of them — they came off a single ad. That number is the only honest signal I have so far, and it’s the reason I keep working on it.

What it looks like from the outside

A Telegram Mini App that pairs two strangers for a live voice conversation. No text chat. You pick a topic, filter by age, gender and language, skip whenever you want, and exchange contacts only through a mutual-consent handshake. The copy says company, not dating.

That is the pitch on the projects page. This note is the other half.

What it actually is

The topic is a mask, and I made it up.

The demand underneath isn’t mine and I didn’t discover it — it was proved years ago by Дайвинчик / Leo , the Telegram matchmaking bot that public bot directories put somewhere around 13 million monthly users. Almost every peer I know has used it. Nobody around me is missing a new category; they’re using the existing one and quietly embarrassed about it.

So the mask does three jobs:

1. It lowers the entry cost. “Talk about something for ten minutes” is a much cheaper thing to open than a dating profile.
2. It widens the audience to people who would never press start on a dating bot — which is most of the market, not the edge of it.
3. It gives the first thirty seconds a script. Two strangers on a voice call with nothing assigned is where the product dies.

I want to be precise about this: I’m not claiming a new insight. I’m putting a different front on a behaviour that already exists, because the front is the part that was stopping people like me from using it.

The real reasons

Here’s the part that doesn’t go in the pitch.

1. I need a channel

The main idea was traffic.

Everything else I’ve built has the same defect — no distribution. gopher.uz , the YouTube channel, this site, the next thing I ship: each one starts from zero attention and I go begging for eyeballs in Telegram groups. A bot that people open daily is a pipe I own — the /start message, the Mini App footer, a broadcast I can send.

Blunt version: shoshshi is meant to be the top of the funnel, and my other projects are what the funnel is for. If it also turns into a good product on its own, great. But the thing I’m actually buying with these months is a place to stand.

2. I was paying for a server that did nothing

The second reason is smaller and more embarrassing, and it’s real: I already rent a server, and it was sitting there doing nothing while the invoice arrived every month. Idle capacity I’m paying for is a standing argument to build something that uses it.

Sunk cost is supposed to be the fallacy you avoid. Here it worked as a forcing function: the machine exists, the bill exists, so the only variable left is whether anything runs on it. shoshshi is what I pointed at it — Postgres, Redis, the signalling server and coturn all live on hardware I was buying anyway, which also means the marginal cost of trying is close to zero.

What reading the competitor taught me

• Zero install wins. Telegram is the runtime, the identity and the payment rail. Profile, browse, express interest — that’s the entire surface.
• Anonymity is the feature and the bill. No verification means the moderation system is the product. In 2026 Leo sat at the centre of a contested criminal- recruitment story in Russian media; the details are disputed, the lesson is not: anonymous matching at scale becomes a safety problem before it becomes a business. Report, block and same-gender matching ship on day one for me, not in “phase 6 if there’s time”.
• The revenue is the audience. Leo monetises with ads inside the bot. That’s the same thesis I wrote above, just cashed out by someone else first.

The thing that actually decides it

Not the idea — the cold start.

What I planned was the textbook move: seed 50–100 people by hand from my own circles. I never did it. The 200+ users are clean — one ad, no friends dragged in, no numbers I inflated myself. That makes the signal worth more than a seeded launch would have been, and it makes the cold start harder, because there is no floor of people who show up out of loyalty.

So the part of the growth plan that still matters is the scheduling one: do not launch always-on. Two hundred users spread across a day is eight people online and a dead app; the same two hundred pushed into two “talk hours” is a party. Plus:

• fix the gender ratio before spending anything on acquisition,
• sit in the calls myself early, because ten real conversations beat any dashboard.

See also Ideas and TypeScript — the stack is Fastify + WebRTC + grammY + Drizzle, all TS.

Where the mask has to stop being a mask

If the topic never touches the product — if it’s only a word on the start screen — people will feel it in the first session, and then I’ve got Leo with worse matching. Either the topic genuinely shapes who you meet and what the first prompt is, or I drop the costume and say plainly what the thing is.

Stub. The other reasons I’m building this — the ones I haven’t written down honestly yet — go here.