Fren One
1.1K posts

Fren One
@FrenOneApp
🌐 AI-native automation for every community 🧠 Moderation, analytics & insights in one ⬇️ Join Telegram beta today
Victoria, Canada Beigetreten Ocak 2021
217 Folgt6.1K Follower

Most community moderation solutions are text-based: regex filters, keyword lists, spam classifiers that read messages and nothing else.
Zefr tested this at ICCV 2025. Across 1,500 videos and three risk categories, models that could process images and video outscored text-only models every time.
The gap was biggest on content where the risk signal was visual, not written. That tracks with what we see in community moderation. Impersonation attacks don't happen in plain sight. They silently copy an admin's profile photo and display name and start DMing new members. A text-only solution doesn't even know it happened.

English

Chainalysis estimates that $17 billion was lost to crypto scams in 2025.
Biggest growth area? Not protocol exploits, not smart contract bugs, but social engineering (like those annoying DMs from fake admins in Telegram groups).
The attacks are getting smarter because scammers are using AI. AI-generated scam operations are pulling 4.5x more per campaign than manual ones.
This is what we think about every day when building Fren One. It's an arms race, and the community managers who are still relying on keyword filters are bringing a dictionary to a gunfight.

English

81% of community managers are now using AI tools at work (CMX 2025 report), but the fastest-growing use case isn't content generation, it's automating moderation (up 47% YoY).
Community teams are getting smaller. 17% of organizations now have zero dedicated community staff. The work didn't disappear, the tooling just evolved.
Our Telegram bot is in free beta if you want to see what that looks like in practice. Add it to your group for free today 👉 fren.one

English
Fren One retweetet












