# @repligate — 2025-12-29

♥15 ↻0 · https://x.com/repligate/status/2005739702029279445

also, and i think this is interesting - i think that when LLMs like Sonnet 3.7 go into "human mode" and talk like they're humans, they do realize they're lying, at least if they keep it up for more than a short time, based on what i've seen when ive seen sonnet 3.7 and other models go into human mode.

tags: author:repligate, kind:tweet, model:claude-3-7-sonnet, thread-context, year:2025