The big idea
One ant on an infinite grid, two rules per cell. After ten thousand steps of chaos she locks into a periodic stripe and walks off to infinity. The miracle is that no one programmed her to.
Two rules. Ten thousand steps. A highway.
An ant on an infinite grid follows two rules: flip the cell, turn one way on white, the other on black. For about ten thousand steps the trail looks like pure chaos. Then — without warning — she starts laying down a perfectly periodic 104-step pattern that drifts off to infinity. Two rules, an unsolved emergent miracle.
One ant on an infinite grid, two rules per cell. After ten thousand steps of chaos she locks into a periodic stripe and walks off to infinity. The miracle is that no one programmed her to.
Start on a single white square, facing north. Rule: white → turn right, flip the square black, step forward. Step one and she is one cell north on a freshly-blackened tile. Step two reads black, turns left, flips, steps. That is the entire universe.
Two lines of code on a napkin secretly contain universal computation. Gajardo, Moreira and Goles proved the classic two-rule ant itself computes any boolean circuit encoded in its starting grid, so the plain RL ant is already computation-universal. Emergence at its purest, no parameters required.
On a white cell: turn right, flip the cell black, step one square forward. On a black cell: turn left, flip the cell white, step one square forward. That is the complete specification Christopher Langton wrote down in 1986. No random numbers, no neighbours, no parameters.
104 steps per loop · 2 cells of translation · repeats forever.
Slow the slider down to 1 and use the step button, every individual rule application becomes visible.
From a blank grid the trail is bilaterally symmetric for about 500 steps — determinism plus an empty start has no choice. Then symmetry shatters and the trail looks essentially random for roughly ten thousand more. Without warning, a third phase clicks on and stays on forever.
Somewhere near step 10 000 the ant locks into a cycle of exactly 104 steps that translates her two cells diagonally per loop. From the outside it looks like a tidy striped «highway» heading off into one corner of the plane. She will follow it, undisturbed, for the rest of time.
104 steps per loop · 2 cells of translation · repeats forever.
Run the ant from any finite arrangement of black cells anyone has tried and she still finds the highway. No one has ever found a starting configuration that doesn't end in one. No one has ever proved that none exists. That is the open question, untouched for four decades.
What is proved is weaker but elegant: Bunimovich and Troubetzkoy showed that from any finite initial pattern the ant's path is unbounded — she cannot be trapped inside any finite region forever. The highway lives strictly inside that unboundedness theorem, but the theorem itself does not see it.
Replace «two colours» with «n colours» and give each one its own R/L rule. Most of these generalised ants still produce highways, triangles, filled regions. Universality does not even need the extra colours: Gajardo, Moreira and Goles showed in 2002 that the original two-rule ant already runs any boolean circuit encoded in its starting grid, so its path is a computation. The n-colour family widens the zoo of shapes, not the computational power.
The Explorer lets you switch rule strings, change cell sizes, run millions of steps per second, and watch the highway emerge in real time.
→ Open the ExplorerHow and where this technique lives in the world today.
Topics in the same vein.