> complete independence of network connectivity and hence minimal latency. Does it matter that each token takes additional milliseconds on the network if the l

> complete independence of network connectivity and hence minimal latency. Does it matter that each token takes additional milliseconds on the network if the local inference isn't fast? I don't think it does. The privacy argument makes some sense, if there's no telemetry leaking data.