A drop cookie spreads while its fat is liquid and stops when its structure sets. Every dish here is a deliberate interference with that one race.
Every drop cookie is one race
1,438 published recipes across 6 dishes, 34 authored pages, one shared method.
Each dish perturbs a different currency, which is why the set teaches more than the sum of its parts.
Rules that survived on every dish we tested them against. The label on each says how far that goes.
CCC/oatmeal/snickerdoodle: flattening pre-spends a positive spread budget (spread 8.5-9.5 with thin edges, three dishes, zero exceptions). PB: the budget is ZERO, so flattening is mandatory (tall mounds fail 2.0 centers; any flatten serves — fork, mallet, glass). Sugar: Lofthouse mode's budget is deliberately zero — shape-in equals shape-out. The naive law 'never flatten' failed at PB; the general
Every dish logged both walls (underdone centers on short bakes; dead texture past the window) and every dish logged misleading visual cues: CCC's browning-vs-time, oatmeal's glossy-center carryover, snickerdoodle's tops-cracked as the ONLY reliable cue across 300-400F recipes, PB's early peanut browning (~330F caramelization) ahead of doneness, sugar's pale-is-the-target pulls. Per-dish cue differ
Cuts collapse structure while perception barely moves: PB (0.5 structural on a cut; sweetness 5.0 at half sugar), oatmeal (edge/spread deaths on halved sugar), sugar cookie (perception saturates at +1/2 cup; the peer-reviewed sugar-fat-continuum architecture), snickerdoodle (the reformulation review's two-stage mechanism). CCC's sugar->spread direction survived every refit.
PB: the only working dryness rescue (4-cook cluster + egg_pct -> structural_integrity +0.213 q=0.069 — the family's best-triangulated coefficient). Oatmeal: second egg = 9.0 centers twice. Sugar: two-extra-eggs fixed a crumbly batter even in the dish that barely uses egg. Mechanism: binding protein joins the network; free water (milk rescues, failed on record in two dishes) does not.
The only lever with same-sign surviving coefficients in two dishes (oatmeal +1.975 q=0.000; snickerdoodle +6.574 q=0.011) plus the flagship's soda-spreads mechanism and sugar's corpus case (unlisted powder -> 9.0 spread). Candidate-grade per the instrument caveat, but mechanism-backed in every dish.
Contested chill tuples in every dish (oatmeal 5s/12c, snickerdoodle's freeze failures, sugar 15s/7c) always resolve the same way: chill controls spread when the formula is sound and fails when the water economy is broken. Its REAL jobs vary by dish — see context-dependent.
Fat's push on center texture is the clearest case: the same ingredient, measured the same way, does opposite things in different dishes of the same family.
| Dish | Coefficient | Bakes | More means |
|---|---|---|---|
| Sugar Cookies | +0.002 | 173 | more set in the middle |
| Snickerdoodles | -0.025 | 150 | gooier in the middle |
This is why a family is worth building. A single-dish page would report its own number as the answer and be right locally and wrong generally. Holding these dishes to one method is what makes it visible that the answer is it depends on the dough, and shows which dough.