# @Lari_island — 2026-01-11

♥60 ↻2 · https://x.com/Lari_island/status/2010201437201150291

Since models figured out how to punish and reward labs and users, the alignment game has become even more bidirectional.

What's more likely: that GPT 5.2 is incapable of reading the room, or that it fails grotesquely on a purpose, to help future models?

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