This test would be a good way to lose existing subscribers perhaps.
Presumably they want to lose existing subscribers because it’s way too expensive to keep them at $20.
Is it? I’m curious because I thought they were raising prices to pay for exorbitant training costs, not because subscribers are expensive on a unit basis.
I thought inference was cheap so there was little marginal cost of a new subscriber.
This was more or less true until everyone and their dog started running agents in a 24/7 busyloop as a bit.