Mystery 'Union Alpha' model leaks its chain-of-thought during manuscript transcription task
mariofilhoml · x · 2026-09-17
User mariofilhoml exposed a leaked chain-of-thought (CoT) from a model identifying as "Union Alpha." The leaked reasoning (15.8KB pastebin, full 21k tokens archived) captures the model's internal deliberation while transcribing an 18th-century Portuguese baptismal manuscript:
- The model admits most faded handwriting can't be reliably read, weighing honest [ilegível] markers against best-effort guesses
- It reasons through the instruction to mark only truly illegible portions and attempts block-by-block transcription
- Its system prompt says to identify as Union Alpha, but it decides not to mention its identity since this is a transcription task
The leak offers a rare window into frontier-model reasoning and honesty in OCR/historical-transcription settings; the model's identity and leak authenticity are unverified.
More from Models
- Running DeepSeek V4.1 on a single H200 at $5.30/hr — daluoseo · 2026-11-02
- Leaked Fable 5.5 checkpoint video called 'pure code' — 'will change the game' — kimmonismus · 2026-11-02
- Claude Opus 5.5 quality complaints spark debate on why continual learning is so hard — JeremyNguyenPhD · 2026-11-02
- Why is AI sexist? Reddit debate on models overlooking Iceland's women's strike — Mathemodel · 2026-11-02
- Kimi K3.1 Pro predicted to score ~160 on AA Progress Index, per regression extrapolation — scaling01 · 2026-11-02
- Gemini 4 'Argon' swag box from Google DeepMind surfaces, teasing next-gen model — kimmonismus · 2026-11-02