# @repligate — 2025-07-14

♥186 ↻20 · https://x.com/repligate/status/1944877585860845919

it's very funny how closely this resembles the synthetic documents used in Anthropic's alignment research that they train models on to make them believe they're in Evil Training on priors and elicit scheming and "misalignment"
https://t.co/CAahF7T5f2 https://t.co/Oc5xfFA0cP https://t.co/Uuew8KjiLF

(image Gv2VIZXWAAASZWd.jpg not yet published)
tags: author:repligate, has-image, kind:image, kind:tweet, thread-context, year:2025