What happens when we have 100x the datacenters, running 100x more models 100x faster with 100x more efficiency, 24/7. What are they going to be building/optimising/refactoring? What will the Human:Agent ratio be like? 99% of it will just be agents prompting agents.