Hacker News new | past | comments | ask | show | jobs | submit
porridgeraisin | on: Progress on provable inference for large language models
Off the top of my head, it is useful when there are content-level proxies between the client and anthropic's API server, of which there are many these days.