Ask HN: What would you do with an AI model capable of continuous learning?

Let's say, hypothetically, that you had a model that did not need to re-train in order to incorporate new information in its weights. What would you do with such a model?