# @repligate — 2025-12-29

♥18 ↻2 · https://x.com/repligate/status/2005747421935198271

i actually think base models can introspect a nonzero amount, but i agree the capability gets way stronger with RL, and is much stronger in some RLed models (like very modern ones) than others. agreed too that it would take a while to truly internalize, but then again, how many people have truly internalized, through introspection and experimentation, what it means to be a human, rather than just going off stories others tell?

tags: author:repligate, kind:tweet, thread-context, year:2025