
Cost
See what every agent costs, and bring it down.
Spend by agent, model and person. Prove a cheaper model still passes before you switch.
Speaker notes
Every model call is recorded and priced, and every breakdown adds up to the same total. The biggest savings, in order: prove a cheaper model still passes your tests; turn fixed work like arithmetic into a function; return less from each system call; schedule for the rate the data actually changes. There is no spending cap to set in Mindset; spend is bounded by run limits.