Emergence is what happens when many copies of a simple rule, each acting only on local information, produce large-scale order that nobody wrote down. A bird reacts to its nearest neighbors, and the whole flock turns as one. A cell responds to the chemicals touching it, and the animal grows stripes. In every exhibit below, the entire rule set fits in a paragraph, and the text tells you what it is. The pattern is not stored anywhere; it appears when the rule runs.
Everything runs live in your browser. Tap the presets to jump between notable configurations, drag on the canvases, and edit any number directly when you want an exact value. To save battery, each simulation sleeps while it is scrolled off screen, and stops computing once its picture has settled; touching any of its controls wakes it.
The simplest possible world: a row of cells, each on or off. Every tick, a cell looks at itself and its two neighbors. That is eight possible situations, and a rule is just a table saying what to do in each one. Eight yes/no entries means there are exactly 256 such rules, so this entire family of universes can be numbered 0 to 255. Each row of pixels below is one moment; time flows downward.
Most rules are boring: they die out or freeze into stripes. A few are not. Rule 30 boils into froth so unpredictable that Mathematica used it for years as a random number generator. Rule 110 can, in principle, compute anything your laptop can; that was proven, not conjectured. The interesting question is why these particular tables misbehave while their 250 siblings sit still. Nobody fully knows.
A starling murmuration has no choreographer and no leader. In 1986, Craig Reynolds showed that three private urges per bird are enough: separate (don't crowd your neighbors), align (fly the way they're flying), and cohere (drift toward their center). Each bird sees only the birds within a small radius. None of them knows the shape of the flock; the shape is what the three rules look like from far away.
Rebalancing the urges changes the species. Strong cohesion collapses the flock into a dense, wheeling ball. Remove alignment and it becomes a cloud of midges, milling but never traveling. Your cursor or finger is a hawk: birds flee it, and the escape wave that ripples outward is the same one you see in films of real flocks under attack.
In 1952, two years before his death, Alan Turing proposed that a pair of chemicals spreading through tissue, one feeding a reaction and one starving it, could paint spots, stripes, and labyrinths onto skin. No gene stores the leopard's pattern. The genes only tune two numbers, a feed rate and a kill rate, and the mathematics does the rest. Biologists dismissed the idea for decades; it is now textbook.
This is the Gray-Scott model, the standard modern form of Turing's equations, computed live pixel by pixel. Every regime below differs only in those two numbers, and the gap between coral, dividing cells, and worms is a change in the third decimal place. Draw on the canvas to inject chemical and watch the pattern absorb the intrusion.
A sunflower places each new seed at a fixed angle from the previous one, a little further from the center. That is the whole algorithm. The choice of angle is everything: any rational fraction of a turn eventually stacks seeds into wasteful spokes with gaps between them, so the plant needs an angle that never repeats. The number hardest to approximate with fractions is the golden ratio, and the corresponding angle is 137.508 degrees.
The presets show how sharp the optimum is. At exactly 137.5, three thousandths of a degree short, the packing unravels at the rim. At a rational angle like one seventh of a turn, the spokes appear instantly. Evolution found a theorem in number theory and grew it in a field.
One pendulum is the physicist's favorite pet: tame, periodic, solved for centuries. Hang a second pendulum from the tip of the first and you get one of the simplest systems known to be chaotic. Below are two double pendulums drawn on top of each other, released from angles that differ by an amount you can set. The default gap is one ten-thousandth of a degree.
For a while they agree so exactly that you can only see one pendulum. Then, abruptly and permanently, they part. Try the gentle preset for contrast: at small swings the identical system is orderly and the twins never separate at all. Chaos is not randomness; it is the same deterministic physics amplifying any error in your knowledge of the starting point. This is also why weather forecasts decay. The physics of the atmosphere is known, but today can never be measured precisely enough to deserve next month.
Drop grains of sand one at a time onto a grid. A cell holding four grains topples, giving one grain to each of its four neighbors. A neighbor pushed to four topples in turn, and so on until everything settles: sometimes after one topple, sometimes after thousands. Grains that fall off the edge are gone. That is the entire model.
Left alone, the pile tunes itself to the edge of stability, where the next single grain can do almost nothing or almost anything, and avalanche sizes follow a power law, the same statistical signature as earthquakes and forest fires. The physicists who found this called it self-organized criticality; nobody steers the pile there, falling sand is enough. The colors count grains (0, 1, 2, 3), toppling cells flash rose, and the mandala that forms under a steady central drip is stored nowhere. Tap the canvas to pour sand wherever you like.
The botanist Aristid Lindenmayer wrote grammars for how plants grow: start from a short word and apply the same substitution to every letter, generation after generation. A turtle then reads the final word as instructions, drawing a segment for F, turning for + and −, and using [ and ] to remember and return to a branching point. The full rule for each plant below is printed under its canvas, and each is one line long.
Nothing in that line says fern. The overall shape, the taper, the way fronds curl, all of it is the rewriting rule echoing through six generations of itself. Change the turning angle by a few degrees and you get a different species, which is roughly what mutations do to real developmental programs. This last exhibit is the closest to an actual organism: a plant is a program that runs on sunlight.