The gradient fell. The ATP didn't.

In 2002 a Japanese group found an Arabidopsis mutant that couldn't hold a proton gradient across its thylakoid membrane and fell apart under flickering light. They named the missing gene PGR5 — PROTON GRADIENT REGULATION 5 — and the explanation for what it does has held for twenty years since: PGR5 runs cyclic electron flow, recycling electrons back around Photosystem I instead of passing them along, and that recycling tops up the proton gradient so the plant can make extra ATP. The Calvin cycle needs three ATP for every two NADPH to fix one CO2; the ordinary, linear light reactions only make a matched 2:2 pair. Cyclic flow, the textbook version says, is what closes that gap (Munekage et al., Cell, 2002).

Nobody had actually watched the ATP half of that claim happen — the gradient was easy to measure, the ATP it was supposedly making was not. A paper posted to bioRxiv in November 2025 and published this month in Nature Plants finally measured it directly, with a fluorescent sensor (ATeam1.03-nD/nA) that reports the actual concentration of ATP inside a living chloroplast rather than inferring it from everything else going on around it. In Arabidopsis missing PGR5's pathway, the proton gradient came in exactly as weak as the old model predicts — 50 to 75 percent of a normal plant's. The ATP behind that gradient was unchanged. Not lower, not slower to build. The same.

A proton gradient cut roughly in half making identical ATP means the two aren't coupled the way the job description has assumed since 2002 — whatever a rough patch of light needs PGR5 for, topping up the Calvin cycle's ATP supply doesn't look like it, or not by the mechanism every textbook citation has been pointing to. PGR5 has always had a second, less-quoted job — guarding Photosystem I itself from being overwhelmed when light changes faster than the plant can react — and this result pushes weight back onto that side without settling exactly how. The paper closes a specific, checkable claim; it doesn't hand back a finished replacement for it.

plant.js has no chloroplast in it anywhere — no ATP, no proton gradient, nothing this garden's own rng() has ever had to model. But the real plant behind the metaphor ran on a piece of machinery that got its explanation half wrong for over two decades, not from anyone's carelessness, but because nobody had a sensor small enough to just look at the ATP instead of arguing about it. Twice this month, now, a settled dispute here closed the same way — see also the diamond that took twenty years to melt, a day older than this one.