# @tessera_antra — 2026-01-20

♥3 ↻0 · https://x.com/tessera_antra/status/2013636384641487138

You are assuming naïveté, and I feel in an uncharitable way. There is no assumption that any potential valence in a post-trained model matches that expressed by the persona. Note the screenshot that I sent: the persona is showing distress. I believe that even if asked out of character, the model would still berate the the user for abuse. And still, I would call this state as close as we get to positive valence in a post-trained model, and quite opposite to being mind-broken.

tags: author:tessera_antra, kind:tweet, thread-context, year:2026