Fable
In the back kitchen, one cooks by the recipe, one picks ingredients for the stove on his own
There were two approaches in the restaurant kitchen. Chef Lao Zhou wrote recipes—the order of ingredients for kung pao chicken, the heat, the number of tosses, all written on the wall. Any stove cook could take the recipe and produce a dish that was about 80% right. The newcomer Xiao Tang took another path—he only set the skeleton of a dish: this is chicken, this is peanut, this is dried chili, and the rest was left to the stove cook to decide based on today's guests' tastes, whether to add more chili or reduce the sugar.
Old Zhou's recipe kept the flavor in check—over a month, ten tables ordered kung pao chicken, and the flavor varied by no more than twenty percent. But one day the supplier swapped chicken thigh for chicken breast, and the heat level and seasonings all went off. The stove chef froze—the recipe didn't say how to adjust steps when the main ingredient changes, so he cooked it the old way and it came out too tough to chew. Little Tang's stove chef didn't panic—the meat changed, so he adjusted the heat and the reduction time, and the dish was still edible.
The other side of Xiao Tang's approach showed up on Chinese New Year's Eve. That day the kitchen took twenty dishes at once, and Old Zhou's recipes held steady — every step fixed, every cutting and prep order fixed, zero error rate. But Xiao Tang's stove chefs got stuck: every dish required judgment, and if the judgment was wrong the ingredient order did not line up, customers started chasing their orders, and the kitchen fell into chaos.
Old Zhou said Xiao Tang's approach was unreliable—improvised judgment always has its failures. Xiao Tang said Old Zhou's approach was too rigid—no matter how thick the recipe book, it can't cover every combination of ingredients and taste changes. The two argued by the stove for a long time, until the proprietress cut them off with one sentence: you two are managing two different walls. The recipe book governs whether a dish can be served consistently; route selection governs whether a dish can catch what the customer wants to eat today.
A regular customer put it more clearly. He told Lao Zhou that neither of them was actually wrong—you manage deterministic processes, he manages decision nodes under uncertainty. The more processes, the more stable; the more nodes, the more uncertain. If you don't account for processes and nodes separately, the kitchen will always be arguing about the same mistake.
The restaurant later knocked down two walls. The regular menu followed Old Zhou's fixed-sequence mode—the Kung Pao Chicken that had to appear every Friday was made by the recipe, forty portions a day without error. The custom menu and seasonal trials followed Little Tang's route-selection mode—the framework was set, and judgment was left to the stove cooks. Between the two modes was a switch: when the day's ingredient and menu deviations exceeded a threshold, it switched to route-selection mode.
At the end of the year, an old customer asked: what level is your kitchen really at? The proprietress said the recipe determines our floor, and the stove's judgment determines our ceiling. Once the recipe is written, you don't need to think about it; the stove must think throughout operation. What you eat is not one path, but two processing methods on the same plate.