Underlay is a technical show about digital infrastructure, with networking as its center of gravity. Its recent episode excerpts point to a catalog concerned with packet-level truth, automation, telemetry, SD-WAN, WAN optimization, and the messy work of keeping production systems understandable. The show does not treat infrastructure as an abstract layer. It keeps coming back to what people can see, measure, configure, and troubleshoot. Packet captures become a way to settle uncertainty. Operational state data becomes a way to ask whether the network is actually behaving as expected. gRPC and gNMI are discussed alongside NetConf, RESTConf, YANG, and telemetry as part of a wider search for better network management patterns. SD-WAN is handled through performance limits, path selection, latency, application sensitivity, and the tradeoffs that appear when private circuits cannot simply be removed. The register is conversational, but the subject matter is precise. Guests and speakers explain their background, compare tools, and work through practical examples rather than staying at the level of slogans. Some episodes come through Network Collective formats, including community roundtables and Network to Code conversations. Others carry the Riverbed Technical Podcast frame, with packet analysis and application acceleration front and center. The show also has an unusually self-aware thread about format. In “New Year New Us?”, the speakers describe a history of unscripted conversations with smart people and consider more edited, narrative approaches built from multiple perspectives. That matters because it matches the material. Networking is full of competing claims, vendor language, and inherited habits, and Underlay is strongest when it slows those down into concrete operational questions. What evidence proves the fault? Which protocol model fits the task? Where does automation help, and where does it merely rename old work? The result is a show for people who want infrastructure explained by people close to the machinery. It is not polished futurism. It is working technical conversation, grounded in packets, state, tools, and the humans responsible for making them line up.