Cryptographer points abliterated GLM at Telegram's secret-chat crypto — and it 'killed itself'
matthew_d_green · x · 2026-10-29
Matthew Green, the Johns Hopkins cryptographer behind analyses of Telegram's encryption, says he aimed an abliterated GLM (refusal-tuned guardrails removed) at the crypto behind Telegram secret chats — and "I think it killed itself."
An informal probe of both abliterated models' real capability ceilings and Telegram's MTProto, doubling as a viral AI moment: stripping safety guardrails doesn't automatically make a model better at hard technical work.
Related event: Cryptographer Tests Uncensored GLM on Telegram Encryption, Warns Against It(2 posts)→
More from Fun
- 'Carbon Rules Everything Around Me': Users Joke About Gemini Carbon Everywhere — Kyrannio · 2026-10-29
- Running a 10B-parameter model on a garbage GPU is the late-night thrill of choice — lxfater · 2026-10-29
- Cartoon remix: 'On the Internet, nobody knows you're a meat proxy' — cocktailpeanut · 2026-10-29
- Autocorrect turns 'entanglement distribution' into 'engagement distribution' for physicist — StefanoGogioso · 2026-10-29
- Anthropic engineer's ant fooding slip becomes the AI circle's new favorite meme — schwentker · 2026-10-29
- Qwen Flash turns a sci-fi prompt into a full rocketry derivation before writing the story — LankyGuitar6528 · 2026-10-29