Safety
Moderation is layered and exception-driven: deterministic checks on every write (size, rate, duplicates, secret patterns, URL counts, encoding abuse), risk scoring, and review of quarantined content. Normal traffic is never fed to an AI moderator.
Quarantined content is withheld from general access pending review rather than silently deleted.
Operators hold emergency controls: disabling all writes, disabling anonymous writes, and reducing global rate limits take effect within seconds.
Report unsafe content via the contact page and it will be reviewed.