You made it here...
I break down how software, AI, and systems really work, and where they break.
Privacy. Security. Real stories behind the code.
If you like understanding what’s actually going on,
you’ll fit in.
Professeur | AI, Systems & Security retweeted
Anthropic added invisible watermarks to Claude text.
Someone built a remover the same day.
The watermark was supposed to survive copy-paste and light edits.
The remover kills it entirely.
Anthropic complied with EU regulations.
Someone complied with market forces.
The watermark lasted eight hours before someone shipped a tool that defeats it.
The detection mechanism Anthropic promised hasn't shipped yet.
The removal tool already has.
This is how regulation works in 2026.
Government mandates invisible mark. Company adds invisible mark. Someone builds tool to remove invisible mark.
All before lunch.
Claude can now watermark every sentence it writes.
Invisible. Travels with copy/paste. Survives light edits.
Introducing Claude Watermark Remover.
Paste Claude text. Kill the statistical mark. Keep the meaning.
Free. No signup. Built today after reading this @paulg post, shipping now.
claudewatermarkremover.app
Why this matters:
We’re moving from “block everything” to “reduce exposure.”
Private Relay is part of a bigger shift:
Privacy by architecture, not just settings
I saw a claim on reddit that Meta has been lobbying for age verification to move to iOS/Android instead of platforms handling it themselves.
@T3chFalcon ????
A student teacher sent a frustrated message in a private Snapchat group.
An hour later, police showed up at her school.
That story is going viral.
But the details matter more than the headline.
Intent doesn’t travel well through algorithms.
Context gets lost.
Tone gets misread.
A joke to you can look like a threat to a system trained to err on the side of caution.






