How LLMs are used at Oxide

I love Bryan Cantrill’s framework for LLM use at Oxide, grounded in the company’s values. He observes that there’s a social contract between readers and writers about the amount of work that has gone into a piece, and that LLM-generated prose breaks this contract. It maps to my own belief that LLMs should be used to make work better, not necessarily faster.

https://rfd.shared.oxide.computer/rfd/0576