LLMs automatically manage long context now. But I still manually assign machine attention resources. SandHive assigns those resources. I reclaim my attention for higher tasks.