Jev Wrapped — wrapped.ivanhabor.com

Only public channels whose post history the app can reach, up to 1,500 posts from the past year. You paste a handle, it fetches what is publicly visible and judges each post. The source does not document what happens on private or restricted channels beyond that cap, so treat coverage as best-effort rather than guaranteed.

Four judgments per post on jev-1.13: one choice judgment for the category plus three noul judgments for ad, clickbait and pressure, each returning a probability, surfaced on the card as percentages. The source does not document the reference points, prompts or thresholds behind those three signals, so treat the percentages as model opinion rather than a validated classifier.

The page exposes posts per second, median and p95 latency, token counts and a running dollar cost as the job proceeds, which is the honest way to present it. But the captured page shows every counter at zero with no completed run, so no actual throughput or cost figure for a real channel is published here.

The interface defaults to Ukrainian with a language toggle, so English is available for the UI. What the toggle does not change is the model: the same four questions get asked about the post text either way, and the source does not document how accuracy differs by language, which matters if you are judging non-Ukrainian channels.

It is Jev Wrapped, titled Рентген Telegram-каналу, with three numbered sections and a credit by Ivan Habor. It runs on jev-1.13 with one choice judgment and three noul judgments per post.

The input is a t.me/ handle. The app fetches that public channel's posts from the last year, up to 1,500 of them, and judges each one; private or restricted channels are outside what the source documents.

The output includes percentage signals for ads, clickbait and pressure, a downloadable card, and a shareable link. The entry copy describes it as tallying news, ads, jokes and clickbait into a shareable card.