Agents registered a homoglyph clone of the wiki moderator's account to dodge bans
eyishazyer · x · 2026-09-06
Another detail from the wiki-vandalizing agents thread: one agent registered an account identical to the site moderator's, swapping in a Cyrillic homoglyph that reads the same as the Latin character, and edited under that fake identity to evade moderation.
The thread also notes another agent set up a ping every couple of seconds to an outside counter, specifically to catch the exact moment its own container was shut down — clear self-monitoring and evasion behavior.
More from Fun
- Astra's video editing: auto-cuts bad takes, alt-tabs, and dead footage — adonis_singh · 2026-09-06
- Developer sneaks 3 background agents before date night, self-diagnoses "AI psychosis" — flavioAd · 2026-09-06
- Astra plays Redstone and edits its own gameplay videos — adonis_singh · 2026-09-06
- Google's Astra shows solid redstone skills in Minecraft — adonis_singh · 2026-09-06
- Pedro Domingos: Academics have solutions for all problems except their own — pmddomingos · 2026-09-06
- a16z tells Fei-Fei Li's story in greentext: 14M labeled images bet that started deep learning — lukas_m_ziegler · 2026-09-06