#distributed-inference

Cross-datacenter inference just split the monolith that never should have been one

Breaking prefill and decode across datacenters isn't innovation, it's just fixing a fundamental architectural mistake.