In 2018, AWS released DeepLens, a deep-learning-enabled camera designed for developers. I got one and spent time figuring out what it would actually take to add ML-powered visual awareness to a home automation system — not as a product, but as an honest level-of-effort estimate. Could a house know who walked in the room? What would the latency, accuracy, and maintenance burden really look like? The answer was “not yet, not cheaply” — which was exactly the kind of answer worth having before recommending similar approaches to clients.