Pseudpocalypse

Your secret online identity may be cooked — and commenters are split between panic and sarcasm

TLDR: The article argues that your writing style may soon be enough to reveal who you are across supposedly anonymous accounts. Commenters split between doom, skepticism, and jokes about using artificial intelligence to remix their words and dodge the digital detectives.

The big scary idea in Pseudpocalypse is simple: your writing style may be as unique as a fingerprint, which means posting under different names might not stay secret for long. The author argues that as language tools get better, even a small chunk of text could be enough to connect your anonymous account, your old blog, and your “definitely not me” burner into one very awkward bundle. And the comments? They immediately turned into a mix of privacy panic, skeptical eye-rolls, and DIY chaos energy.

One camp basically said: welp, anonymous writing is dead. But even the doomers had limits. One commenter argued that if written anonymity is collapsing, society should at least protect other private spaces, like confidential political donations and masked protesting. Another crowd was not ready to surrender and fired back with the obvious modern fix: can’t artificial intelligence just rewrite your words and save your fake name era? That sparked the thread’s most delicious tension — are we watching the end of online secrecy, or just the start of an arms race where bots hide humans from other bots?

Then came the skeptics, who treated the whole thing like a dramatic trailer without receipts. “Has it happened?” one commenter asked flatly, puncturing the apocalypse mood in six words. Another pushed back on the math itself, dropping a Nature paper like a scholarly subtweet. The vibe was pure internet: half existential dread, half “source?” and half jokes about shuffling your prose until your digital ghost loses the scent.

Key Points

  • The article argues that online writing carries a statistical fingerprint that can link texts published under different names to the same author.
  • It says public tools for broad author-linking may not yet exist, but claims such capabilities are likely possible and may soon become easier to build.
  • The article cites improved LLM performance, including an example involving Claude 4.8 identifying the author from 1,000 words of draft text.
  • It extends the idea beyond text, arguing that many high-bandwidth behaviors or physical signals could become identity-revealing even when privacy measures are used.
  • A thought experiment in the article claims that about 29 bits would likely be enough to uniquely distinguish a currently living person in the Anglosphere.

Hottest takes

"So anonymity of written speech is toast" — pmdulaney
"can't we just ask AI to sufficiently shuffle our words" — erelong
"Has it happened?" — inigyou
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.