“I call it cost per thought… total cost divided by the trace… it looks at the entire trace and figures out how efficient it is”

“Measure the return… might be cost savings, might be revenue growth, divided by both the labor cost and the token cost associated with that project.”

“I would do cost by app… or cost by model… I track spend and efficiency by model, so devs can look and see what’s the best model per their workload”

“You have to optimize the whole stack all the way from the power… down to the GPUs… to really reduce your cost per token”

“I like to do is find out how much this event is going to cost and then double that amount of money as your profit… cost $500 per person → $1,500 ticket”