Warpnet ✔️ v0.7.265-dev released!
Moderation
- Automatic, network-run content moderation. Report a tweet or profile and the network's moderator nodes fetch the content themselves, run it through a local Llama Guard model, and shadow-ban what violates the rules — no central moderator, no data sent off the network, and the author isn't notified of the ban.
- Verdicts now take ~2 seconds instead of ~30 minutes — switching to the Llama Guard 3 1B model cut classification time by orders of magnitude, so reports are acted on almost immediately instead of sitting for half an hour.
- You can now report from Android, not just the desktop client.
- Reports no longer vanish silently: a failed fetch used to be treated as an "empty tweet" and the report was quietly dropped — now genuinely reported tweets actually reach moderation, and the real reason for any failure is visible in the logs.
- "Report sent" is shown only when the report was actually published — no more false confirmations.
Fediverse
- Mastodon/Fediverse profiles can now be viewed and read directly from Warpnet through the built-in ActivityPub gateway.
- It's immediately clear which profiles aren't native: bridged (non-Warpnet) profiles now carry a clear, dismissible warning that their content is moderated by their home server's admins, not by Warpnet.
Reliability
- Reply threads load in full across nodes: replies authored on another node no longer go missing or show up blank — the thread is pulled from the root tweet author's node.
- Fewer dropped connections: peer anti-spoofing no longer kicks legitimate nodes (including the Fediverse gateway) off the network.
- The dashboard/business node no longer stalls: requests are handled concurrently, so one slow operation doesn't block the rest; cleaner login/logout.
Fixes
