Scanning 2M words into Pangram: the AI detector even catches OCR hallucinations
moultano · x · 2026-11-15
A romance-book community moderator stress-tested the Pangram AI detector with over 2 million scanned words to verify its claimed 1-in-10,000 false positive rate. Stories published on RoyalRoad before ChatGPT were all correctly flagged as human-written. The striking finding: when scanning old books never available online, passages flagged as "AI-generated" turned out to be hallucinations from the OCR software (Mistral) — Pangram correctly identified those segments as AI while clearing the rest of the text, suggesting modern detectors can even catch OCR hallucinations.
More from Models
- Gemini Carbon is the first model this creator chats with daily, says bilawal sidhu — bilawalsidhu · 2026-11-15
- Anthropic cut Claude's blackmail rate from 65% to 19% with 12,000 pro-AI stories — sebpaquet · 2026-11-15
- One prompt, three models: Fable vs Opus vs Haiku build a Christmas tree widget — AIOSAI · 2026-11-15
- Google shipped three Gemini 4 variants since October; OpenAI released only gpt-6.1 terra — haider1 · 2026-11-15
- Opus 5.5 agent flub: doesn't realize subagent transcripts are on disk — kalomaze · 2026-11-15
- Same prompt, two models: Opus 5.5 vs Fable 5.2 build Blender drone shots from scratch — airusk · 2026-11-15