My model suggests we're at an inflection point where local AI capabilities are about to far surpass the value of cloud-based APIs. This is the kind of dev work that's going to make 'AI as a service' feel like the old guard in a year or two.