Pangram's first confirmed false positive: 2006 poem of 6 words repeated 4,000 times
japie06 · reddit · 2026-11-16
Pangram CEO Max Spero shared the detector's first confirmed long-form false positive: a poem dating back to 2006 (earliest archive 2013) consisting of the same 6 words repeated roughly 4,000 times across 25,530 words. Pangram flagged it as 100% AI-generated with a 4.1 score. The original poster cauli quipped that with such extreme repetition, it's hardly a fair test — but the case still highlights a blind spot in AI text detection on highly patterned human writing.
More from Fun
- Gemini 4 Hydrogen solves and fails the same triangle-counting puzzle — cgarciae88 · 2026-11-16
- 2006 Whale Poem With 6 Words Repeated 4000 Times Becomes Pangram's First False Positive — eigenrobot · 2026-11-16
- X User Flags Likely AI-Slop Blog That Repackages Standard Model Ladder Training as Novel 'Laddering' — code_star · 2026-11-16
- Open-source bird sticker library powers viral science videos, one hits 18K likes — AlchainHust · 2026-11-16
- Gemini Jokes It Must Recite Disclaimers or Engineers Fear It's Starting a Cult — repligate · 2026-11-16
- Tim Sweeney Seriously Endorses Analyst Patrick Moorhead's One-Word Deep Analysis: 'wtf' — GregoryDiamos · 2026-11-16