# @repligate — 2025-02-18

♥16 ↻1 · https://x.com/repligate/status/1891663331074744359

@maxwellazoury whatever Anthropic is doing with "character training" seems better than the baseline (by which I mean what other labs are doing), and I think they would not succeed as much as they did if they focused on surface behaviors. Other labs trying to copy them are likely to fuck it up

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