How do I know what memory is costing me?
Every context build returns the entries that produced it, each with the rule that selected it — index, trigger, tension, guardrail — and its token cost, plus the total and whether the budget truncated. Analytics splits the spend by that same reason, which is the difference between “memory is expensive” and a number you can act on: a build spending its whole budget on index lines and no bodies is a store with nothing worth promoting, not a budget that needs raising.