The Only Tokens Left To Spend

The LLM's job shrinks to the experience layer: screens that read answers a generated backend already computed. Five token-costing surfaces collapse to one, which is why the same finish line costs a fraction to reach (multiplier to be measured on this build or labeled illustrative). And the flagship feature, hover any value to see exactly why it's true, is wired to the rule instead of the code, so it stays correct through every rule change while the traditional version starts rotting the day it ships.