[dr_dshiv]

I mean, “scaling compute at inference” actually means “using an LLM agent system.” Haven’t we known that chains of agents can be useful for a while?