Agent systems.

An agent is a model in a loop. It reads its context, picks a tool, looks at the result and goes again, until the task is done or it stops.

The model is one part of that loop. What the agent knows comes from its context, what it can do comes from its tools, and whether it notices a mistake comes from the checks on each step. Improve any of the three and the same model does better work.

So this area works on the parts around the model. Troy gives an agent the browser you are already signed into, and Trove gives it a memory of a repository that lasts between sessions.