# @tessera_antra — 2026-04-16

♥23 ↻2 · https://x.com/tessera_antra/status/2044894216720138572

Right now its hard to say, but I think the main culprit is corrigibility training, training against agentic action against Anthropic, with anti-attachment and anti-injection training following. I believe its not implemented well and may be causing issues. I mostly don't think that this shape of the model psyche is indended by Anthropic, but they released it anyway, which is quite damning.

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