The source sums all three track allocations and divides their total by seventy-two to display tokens per hour. Hypothetical totals seven hundred twenty and fourteen hundred forty give ten and twenty. Bars share eighteen pixels per token-per-hour unit. This is planning arithmetic, not a benchmark of any model or evidence that a product was built. The source completion index starts at ninety-four point five, then adds the allocation difference from fifteen hundred divided by twenty-five times point four five. Without the bottleneck branch, fifteen hundred gives ninety-four point five and fifteen hundred fifty gives ninety-five point four. Bars share four pixels per index point. The output is an uncalibrated score rather than a validated probability of success. The source applies one twelve-point deduction if any of the first, second or third track allocations falls below three hundred, three hundred fifty or two hundred fifty respectively. A pre-penalty index ninety-four point five becomes eighty-two point five before clamping. Bars share four pixels per index point. No backend model is called for this video. Complete native manifest JSON, original reset, responsive allocation and first-footer placement remain checked.