The cell is the same one from /qwop/, imported rather than copied: four detuned cilia on Q W O P, hold to beat, tap in rhythm to bundle, drive one side to turn. What is new is a bucket. Beating burns ATP, ATP is what you are out here to collect, and beating is the only way to collect it. Fill the bucket and you divide, which is the score and also the only thing a cell is actually for.
Light falls off exponentially with depth — Beer-Lambert, and the one piece of environment here that is not invented. The depth where the light you can photosynthesise exactly cancels what it costs to stay alive is the compensation depth, a real quantity in ocean biology, drawn on screen as a dashed line. Above it a motionless cell slowly gains. Below it a motionless cell slowly dies, however patient it is.
You start below it. That is the whole design: moving is not an optional flourish for the greedy, it is how you get to the light at all. An earlier build started the cell above the line, and a rock — a cell that never pressed a key — out-grew every strategy I could write. That is the paper's 96.6% arrived at for entirely the wrong reason, and a game with nothing in it.
Everything pushes on how much you beat.
| Beating… | costs you |
|---|---|
| burns ATP | the dominant term in the budget; idling is nearly free |
| is heard | the dashed ring is how far. Predators concentrate in the lit water near the surface, which is exactly where the food is |
| scatters your dinner | swarmers feel the flow of a hunting cell and bolt — the louder you are, the further out they start running |
Which means you cannot run a swarmer down. It flees at almost your own speed, so a pursuit closes at a crawl and costs more than the meal is worth. They have to be ambushed: sit quiet until one drifts inside burst range, because they only start running when they hear you. Motes do not flee and can simply be collected.
In /qwop/ the paper's headline — a real Pterosperma is stopped 96.6% of the time — is a fact printed on a panel. Here nothing sets it. The selftest runs a family of scripted strategies against the same oceans, sweeping how long each rests after a dash, and reports which one banks the most growth. The answer it finds:
| strategy | grew | divisions | quiet | lived |
|---|---|---|---|---|
| never moves | −0.55 | 0 | 100% | 126 s |
| sprints constantly | −0.31 | 0.25 | 25% | 31 s |
| rest 4 s between dashes | −0.26 | 0 | 39% | 105 s |
| rest 12 s | +0.07 | 0 | 62% | 135 s |
| rest 96 s | +0.30 | 0.25 | 87.5% | 150 s |
Sit-and-wait wins, at 87.5% quiet against the paper's 96.6% — the right shape, not the same number, and I have not pretended otherwise. Growth rises monotonically with rest and then flattens; both extremes lose badly, the motionless cell because it can never climb to the light and the sprinter because it is eaten and broke. I expected the curve to turn over at the quiet end and asserted that it would; it does not, it plateaus, and the test now says what the data says instead.
The scripted forager is a floor, not a ceiling — it is a crude thing that cannot really ambush. A player who parks in the light and waits for a swarmer to drift into burst range should do considerably better than a quarter of a division per run.
The cell is measured; see /qwop/ and /flag/ for which parts. The ocean around it is not. Pterosperma is a prasinophyte — an alga, a phototroph — so the photosynthesis half of the budget is in character and the light curve is real, but grazing is a what-if: mixotrophy is common among flagellates and is not something this paper, or as far as I know anyone, has shown for this genus. The prey, the predators, and every number in the energy budget are the game's. The budget constants in particular were not chosen, they were searched — tuned until the experiment above produced an interior optimum instead of a dominant rock.
The four keys turn up twice more with no biology at all, just the same question of what four buttons should mean: /qgol/ on a cellular automaton, and /griddle/ on a pancake griddle.
Source for everything about the swimmer itself: Embodied behavioural complexity in a ciliated microorganism, Nature Communications 17, 8445 (2026).