Interesting approach. What does the cold-start phase look like for a new agent? How many traces or runs do you typically need before the router has enough signal to safely offload tasks from the frontier model??
Interesting approach. What does the cold-start phase look like for a new agent? How many traces or runs do you typically need before the router has enough signal to safely offload tasks from the frontier model??
Technically 0 because a) it ingests your already existing traces and does an initial training run b) in the app we'll have pre-trained routers you can start with that will then learn over time