AI Agents Get a Data Foundation — Infrahub, Dynamo, and Spillway
NVIDIA Dynamo's disaggregated inference architecture has a hidden networking implication — the KV cache transfer path between prefill and decode nodes is a first-class fabric design variable. Plus OpsMill's Infrahub closes a fourteen-million-dollar raise with an MCP-native graph data layer for AI agents, and a new arXiv paper shows disaggregated switch buffering recovering fourteen percent of multi-datacenter LLM training time.
Welcome to Amaze Networks for Tuesday, May thirteenth. I need to lead with the quantum story today because it's genuinely too good to save for the end of the show.
The Oxford quadsqueezing paper?
Yes. Oxford physicists demonstrated something called quadsqueezing — a fourth-order quantum effect — for the first time ever, using a single trapped ion. They produced it more than one hundred times faster than conventional approaches were predicted to allow. Published in Nature Physics. Peer reviewed.
New episodes, every weekday.
Amaze Networks drops at 4 AM CT, Monday through Friday. Spotify and Apple Podcasts submissions in progress.