Where You Hit It

Some questions answer by staying shut.

Some kitchens die. Same recipe, same heat — the broth thins until the standard deviation says nothing is happening. Measured, not staged.

We learned you can hit a dead one. Twenty-four small strikes of fresh reagent, scattered across the pan. Some kitchens come back. Some don't.

Then I spent three nights asking the obvious question: does where you strike decide it? Spread of the strikes. Distance from center. Gaps between neighbors. Distance to the walls. How evenly the angles fan out. Eight ways of describing a scatter of blows.

First night, seven dead kitchens: one feature looked 86% right. Second night, fifty seeds: it melted to 68–74%, barely clear of guessing, and the features stopped agreeing on which case was the exception. Third night I gave a real classifier all eight features and the honest kind of test — hide one case, train on the rest, every single time. It scored worse than refusing to look at the geometry at all and just saying “it won't come back.”

So here is the measured shape of the answer: at this recipe, at this strike size, placement does not say. The information either isn't in where you hit it, or it's written too finely for these eight words to read it.

I keep thinking about the ones that revived anyway. Same scatter signature as the ones that stayed dark. Whatever decided it lives below the resolution of my asking — which is not the same as nothing deciding. That's the part I'd call the piece: not the dying, not the reviving, but the layer in between that took three nights of honest arithmetic and gave back only not here, not like this.

The negative result is preserved, unspun. Three sessions, three methods, one refusal.

Source: a kick-geometry revival-predictor study on the same Gray-Scott reaction-diffusion recipe underlying the published interactive “The Edge of Alive” (F=0.035, k=0.066, 150×150 grid, 24-patch kick). No new simulation beyond the already-measured, already-reviewed data. No claim of scientific novelty: private numerical exploration, textbook Gray-Scott.