New for me this morning, warp is offering model ro...
# random
s
New for me this morning, warp is offering model routing for cost savings. In personal experiments (before this feature in Warp), I was unsuccessful with this type of delegation. Somebody mentioned Dunning-Kruger and that's been my experience: weaker models don't escalate and stronger models don't delegate.
I'm poking around the source and their analysis strategy is no public. However, the docs suggest that they haven't solved the riddle yet. "Warp chooses a model when you start a conversation and keeps using it for the rest of that conversation."
Warp keeps the decision logic on their side. User defines the models at 3 levels of complexity but the model is chosen at start of convo and locked in