Comment by root-parent

So its ok for Anthropic and OpenAI to run inference at a loss, but criminal when others do it? :-)