Telegram

Depth: shipping. Most things I build end up inside Telegram. Not a strategy — it’s just where people already are. Ask someone here to install an app and you’ve lost half of them; send a bot link in a group and you haven’t.

What I’ve shipped into it: shoshshi (Mini App + grammY bot, phone gate, admin, broadcast), a downloader bot on yt-dlp, a tiny bot that posts one dhikr a day and has run untouched for months, chat verification for a streaming room, resume delivery for nerd.uz because “send it to me on Telegram” is what people actually ask for.

What it gives you

• Identity, free. initData is signed — validate it server-side and you have a real user. No signup form, no password reset, no email deliverability.
• Sharing that works. A t.me link opens the thing. No store review, no install prompt.
• Local payment rails. Payme, Click, Uzum — and no 15–30% store cut, which changes which prices are even possible.
• A notification channel you own. Rarest item on the list.

What it takes back

• Rate limits shape your UX. A progress bar becomes “edit one message every three seconds”.
• Captions cap at 1024 characters, so the design becomes caption + follow-ups.
• Ten photos sent in a loop is ten notifications; send a media group.
• The web platform isn’t your browser — viewport quirks, a theme you must respect, a back button whose meaning you inherit, and whatever WebView the phone shipped with.

One small thing

Two days into a Mini App on React 19, npm install refused: @twa-dev/sdk pinned react@^18.3.1 as a peer. I didn’t want --legacy-peer-deps in a project’s install instructions, so I filed twa-dev/SDK#47  instead of working around it quietly. Closed now.

Telegram tooling outside the Russian-speaking core is thin. The gap you hit is usually the gap everyone hits.

Still learning: payments end to end with real refunds, webhooks at a size where polling stops being fine, and keeping bot state sane once “just a bot” has become an application.