Notes
What the weekly measurement turns up, and the engineering decisions behind it.
Index
All notes
| 16 Sept 2026 | 51 channels hold 38 per cent of the audience
The largest one per cent of the index is 51 channels, and they hold 38.5% of every subscription we measure. The median channel has 15,982. |
| 16 Sept 2026 | The channel whose username is an empty string
Ask for @durov and the username field comes back as "". Not null, not missing — the six real handles moved into an array, and only one of them belongs in a link. |
| 16 Sept 2026 | The comments on a Telegram post are not on the post
They live in a linked discussion group with their own message ids, and a channel can have them on for one post and off for the next. |
| 5 Sept 2026 | A health check that returns 200 while everything else fails
The liveness probe answered fine for an hour while every real endpoint returned 503. That is worse than having no probe. |
| 5 Sept 2026 | The biggest channels are not the most recommended ones
Ranking 5,177 channels by subscribers and by inbound recommendations produces two almost disjoint lists. |
| 5 Sept 2026 | What this index deliberately does not collect
Public channel metadata and post timestamps. Not message contents, not member lists, nothing from private groups. |
| 5 Sept 2026 | Why a Telegram bot cannot read the channel you want
The Bot API is a permission boundary, not a missing feature. No amount of work gets you around it. |
| 5 Sept 2026 | Telegram shows you today's subscriber count and never yesterday's
Telegram publishes today's subscriber count and never yesterday's. Growth is not a field you can read: it exists only where somebody stored the earlier reading, which is what this index does every week. |
| 5 Sept 2026 | Telegram has no directory. We built one out of its own recommendations
Search returns ten results and stops. The recommendation graph returns 5,177 channels and keeps going. |
Filter