Lightspeed's @buckymoore says long-running agents and similar workloads could "relax the latency constraint" in the inference chip market, benefiting Intel and AMD: "If I'm just telling my software agent to go do work and come back to me in a couple of days when it's done, I probably don't care as much about latency. I care about throughput." "If that's true, then you start to look at these chips that Intel and AMD have, and they look really, really good. They're cheaper and really stand up…