LLM Spam Bots Evade Detection With 7-10 Word "Haiku" Tweets, Investor Says
StewartalsopIII · x · 2026-10-10
Investor Stewart Alsop III observes crypto spam accounts running LLM-generated tweets in his mentions, refined to an art form: operators (apparently from a content farm in Delhi) learned that 7-10 word AI tweets are nearly undetectable — their profiles repeat the same short "haikus" over and over. Pangram-style detectors need longer text to make a call, though he notes watermarks are starting to play a role, possibly via partnerships between detection firms and foundation labs. "This poetry should be studied for the slop it is," he writes.
More from Fun
- Someone modeled a Calabi-Yau manifold in Blender — burny_tech · 2026-10-10
- Satire about "outraged mathematicians" shows some people actually believe it — burny_tech · 2026-10-10
- 10M factory workers losing jobs is a statistic; a mathematician's solved problem is a tragedy — Yamapama · 2026-10-10
- Teachers demand handwritten essays to block AI—students: "No problems" — burny_tech · 2026-10-10
- Schmidhuber renews attack: none of core AI algorithms invented by Hinton — SchmidhuberAI · 2026-10-10
- Hating requires taste: for every take you trash, recommend one you love — floguo · 2026-10-10