I agree, I have tried it for making motion graphics using Argon, in middle of code generation maximum tokens got used and auto switched to 3.1 pro

At some point human labor will be competing with AI simply by charging less than the models charge per hour for the same task

Automated model routing is just immature right now. Given how fast things are moving, though, I'd expect these resource allocation pitfalls to be ironed out soon.

Argon is a 4$in/20$out model; how far do you expect a 20 dollar a month plan to subsidize?

Get yourself an OpenCode Go sub for $10/month, it will go much much further

How did you use it? In agy cli?