I love this argument because people grab Anthropic’s API pricing, stop there, and present it at face value.

Only Anthropic knows their inference costs. And do we _really_ think they’re losing money? Hahaha