Tokenflation: When "Hi" triggers 33 tool calls (quesma.com)

i can't believe people are waiting 5 minutes for a hi call on a cloud model. I assume most of that is provisioning; local model from cold start only takes ~30s