All posts
2 min read

Member trust

Every member now has a trust level, from very low to very high. The bot is stricter with newcomers, and long-timers can be skipped.

Rules are equally strict with everyone, but a chat does not work that way: someone who has been reading it for six months and an account without an avatar registered yesterday deserve different treatment on their first questionable message.

Five levels#

Every member now has a trust level. It shows in the member list and on their card, and you can sort by it. The list finally became useful: suspicious accounts float to the top on their own.

Level Who ends up here
Very low an empty account with unexpired violations or a ban in another chat
Low a newcomer with no username or avatar who recently changed their name
Medium the ordinary state: someone showed up and writes, but is not yet familiar
High a filled-in profile and a decent record in the chat without violations
Very high a long-timer the chat has known for a while

What raises it and what lowers it#

Working upwards: a filled-in profile — username, avatar, Premium — and time spent in the chat. Downwards: an empty profile, a recent name or username change, violations on the platform, and an unexpired ban in any other Scarabeus chat.

Time in the chat deserves a note of its own, because it does not work the way you would expect: both time and messages are required. Someone who has been listed in the group for ages but stays silent remains at newcomer level — they have not shown anything. And the other way round, everyone starts out equal no matter how complete their profile is: trust builds up gradually, as a person writes without breaking the rules.

One special case: people who were in the group before the bot was connected. They count as long-timers straight away rather than newcomers.

The level is live — add an avatar or get a ban lifted and it updates at once, not a day later.

What it changes in moderation#

Trust on its own punishes nobody — it changes strictness.

With low-trust members the bot stops giving borderline cases the benefit of the doubt: what it is not entirely sure about counts as a violation. The same text from a long-timer goes through, from a fresh account without an avatar it does not.

From the other side there is a new "who to check" setting: AI moderation can skip members with high trust, or only those with very high trust. Regulars stop spending the balance, and checking stays where it does some good.

Read next