Few rules, many forms

A cellular automaton advanced step by step makes the distance between an easily written rule and a hard-to-predict pattern observable.

Technology and computing
A stamp places a square in front of a growing mosaic.
A local gesture and a larger pattern. AI-generated metaphor: the mosaic is not a rule 30 output.

A cell looks at itself and its two neighbours. From these three squares it derives its next state. The task is small; the sheet forming below, row after row, can become much harder to describe. Complexity appears without every detail having been drawn in advance.

For this workshop I built an elementary automaton using rule 30. It begins with a single active cell in the middle of a row of forty-one. Beyond the boundary, cells remain inactive; we observe twenty steps. These declared, finite conditions let us return to the same result instead of asking the word “generative” to do all the work.

Eight cases suffice

The three observed cells can form eight combinations. The rule assigns an output to each: from 111 to 000, the output sequence is 00011110. Each new row is calculated from the entire previous row, without using newly updated cells. Changing that calculation order would produce a different system.

One rule, twenty steps

41 cells, rule 30, one active cell in the centre; zero beyond the edges. Each row derives from the entire preceding row. No randomness or automatic start.

Read every row: 1 active, 0 inactive
  1. 00000000000000000000100000000000000000000
  2. 00000000000000000001110000000000000000000
  3. 00000000000000000011001000000000000000000
  4. 00000000000000000110111100000000000000000
  5. 00000000000000001100100010000000000000000
  6. 00000000000000011011110111000000000000000
  7. 00000000000000110010000100100000000000000
  8. 00000000000001101111001111110000000000000
  9. 00000000000011001000111000001000000000000
  10. 00000000000110111101100100011100000000000
  11. 00000000001100100001011110110010000000000
  12. 00000000011011110011010000101111000000000
  13. 00000000110010001110011001101000100000000
  14. 00000001101111011001110111001101110000000
  15. 00000011001000010111000100111001001000000
  16. 00000110111100110100101111100111111100000
  17. 00001100100011100111101000011100000010000
  18. 00011011110110011100001100110010000111000
  19. 00110010000101110010011011101111001100100
  20. 01101111001101001111110010001000111011110
  21. 11001000111001111000001111011101100010001

The control advances one step at a time; reset returns to the initial display. There is no automatic start, and the complete evolution is also available as a text list. Wolfram’s Atlas of Simple Programs presents rule 30 and other examples of its evolution. Our grid is calculated here with its own dimensions and boundaries, rather than importing an image from the atlas.

A short rule can need many steps to show its consequences. We need not give the cells intentions. Each follows the same table; the whole preserves the effects of preceding responses.

Seeing time in the sheet

In the static image all rows are present together. In the system, one comes before another. The vertical coordinate records a succession: moving down a row means applying the rule again, not moving to another part of the same instant.

The step control makes this difference visible. We can stop where a band divides and ask which three squares produced each output. Predicting the whole pattern is not always easy, but every update can be reconstructed. The opacity of the whole does not require its procedure to become mysterious too.

Two paper strips unfold into different mosaics from small initial marks.
The initial condition enters the history of the form. AI-generated illustration, distinct from the calculated, reproducible grid.

The AI-generated illustrations speak of rules and initial conditions metaphorically. Their mosaics are not rule 30 results. The mathematical evidence is the calculated grid and its list of states: confusing the two roles would make the image more evocative and the workshop less controllable.

The boundary is a decision too

Forty-one cells are not an infinite line. We chose to treat everything beyond the ends as inactive. The two boundaries could be connected, as if the row formed a ring; we could start from a different group of cells. These variations deserve their own comparisons because they change the system’s conditions.

Here I kept a single initial condition to avoid mixing questions. Within the twenty observed steps, the pattern expands to the row’s ends. Continuing without discussing its boundaries would bring a new constraint into the pattern we are reading. Reproducibility therefore requires an explicit limit to the comparison too.

The pattern does not absolve the rule

An unexpected result can be interesting, but visual interest does not prove that the rules are good for every use. In a composition we can choose a system for the variety it produces. In a process distributing resources or deciding access, we would also need to judge what its outputs entail, for whom and with which opportunities for correction.

Describing a rule does not mean judging all its consequences. The workshop leaves a small, concrete responsibility: before admiring the pattern, preserve a way to reconstruct it. That trace lets us move from surprise to a design question.

Two paper strips unfold into different mosaics from small initial marks.
The initial condition enters the history of the form. AI-generated illustration, distinct from the calculated, reproducible grid.