Six ways to make rules
The computational techniques behind the Week 2 clinics. Each comes with a small live sketch, its parameters, what it's good for, and the look you'll need to get past.
- Clinics
- Days 6–7 · 30 Nov – 1 Dec
- Techniques
- Six
- Each has
- A live sketch
- You choose
- One, or two together
- Recorded in
- System worksheet, C
How to choose
Start from your metaphor, not from a look you like. The technique should do what the identity believes.
| If your metaphor is about… | Try |
|---|---|
| drift, current, weather, erosion, wind, the movement of water | 01 Noise and flow fields |
| growth, branching, lineage, bureaucracy, things inside things | 02 Recursion and L-systems |
| crowds, migration, foraging, labour, many small decisions | 03 Agents and simulation |
| spread, contagion, settlement, rumour, neighbours shaping neighbours | 04 Cellular automata |
| records, measurement, evidence, what was counted and what wasn't | 05 Data as material |
| order, the grid, property, the survey, rules breaking down | 06 Geometry and tiling |
The sketches below are small on purpose: a few dozen lines each. Change the slider, press New seed, and watch which parts stay the same. Those are the rules. New to how computers draw? Start with Foundations.
Noise and flow fields
Smooth randomness: values that change gradually across space, so neighbouring points agree. Turn those values into directions and you have a field that lines can follow.
How it works
- Make a noise field over the page.
- At each point, read the noise as an angle.
- Drop particles and move each a short step along its angle, again and again.
- Draw the path each particle leaves.
Parameters
- scale
- how quickly the field changes
- count
- how many lines
- steps
- how long each line runs
- step
- how far each move goes
Good for
Anything that moves continuously: water, air, sediment, crowds at a distance, a slow change over time.
The default to avoid
Evenly spaced, smooth hair-like curves in a pastel palette. Let data, a boundary or a rule from the manifesto shape the field.
Look at: Tyler Hobbs's essay on flow fields; Manolo Gamboa Naon; Ken Perlin's original noise (1985).
You're my collaborator. Read system.md. Build the smallest flow-field sketch that follows my rules: a seeded noise field, particles that move along it, and lines exported as SVG. Every value in one config. Explain each step before you write it.
Recursion and L-systems
A rule that applies to its own result. Rewrite a short string of symbols a few times, then read it as drawing instructions, and branching structures grow.
How it works
- Start with a short string, the axiom.
- Replace every symbol by its rule, all at once. Repeat for each generation.
- Read the result: F draws forward, + and − turn, [ and ] save and return to a position.
Parameters
- rules
- what each symbol becomes
- generations
- how many rewrites
- angle
- how sharply it turns
- length
- how far F moves, and how it shrinks
Good for
Growth and inheritance: plants, rivers, family trees, institutions that copy themselves, a document that cites a document.
The default to avoid
The textbook fern or tree. Write rules from your identity's condition, and decide what happens when growth is blocked.
Look at: Aristid Lindenmayer, who devised L-systems (1968); Prusinkiewicz and Lindenmayer, The Algorithmic Beauty of Plants (1990).
You're my collaborator. Read system.md. Build a minimal L-system: an axiom, rules I define in the config, a turtle that draws them, and SVG export. Keep randomness seeded and optional. Show me the rewritten string for the first two generations so I can check it by hand.
Agents and simulation
Many small actors, each following simple local rules. No one draws the whole picture; it emerges from their decisions.
How it works
- Place agents, each with a position and a heading.
- Every step, each agent looks around and decides: turn, slow down, follow, avoid.
- Move them all, and record the trails.
Parameters
- count
- how many agents
- turn
- how freely they change direction
- sense
- how far they look
- rules
- what attracts or repels them
Good for
Behaviour: migration, work, foraging, traffic, people moving through a city, or anything shaped by many decisions.
The default to avoid
Flocking "boids" with nothing at stake. Give the agents a reason to move that comes from the condition: water, a wage, a border.
Look at: Craig Reynolds's Boids (1986); Casey Reas's Process series.
You're my collaborator. Read system.md. Build a minimal agent simulation: agents with a position and heading, one rule at a time from my list, trails recorded and exported as SVG. Seeded. Start with one agent so I can see a single trail, then scale up.
Cellular automata
A grid of cells, each on or off. At every step a cell changes according to its neighbours. Simple rules can make order, chaos or something between.
How it works
- Start with a row of cells, some on.
- For each cell, look at it and its two neighbours: eight possible patterns.
- The rule says which patterns turn the cell on in the next row.
- Draw each new row under the last.
Parameters
- rule
- which of the 256 rules
- start
- one cell, or a random row
- size
- cell size and grid width
- steps
- how many generations
Good for
Spread and neighbourhood: settlement, infection, rumour, land use, a decision that passes from house to house.
The default to avoid
Rule 30 or the Game of Life shown for their own sake. Seed the grid from real data, or let the identity's rules decide what a neighbour is.
Look at: John Conway's Game of Life (1970); Stephen Wolfram's elementary automata, including Rule 30 (1983).
You're my collaborator. Read system.md. Build a minimal one-dimensional cellular automaton: a rule number and starting row in the config, each generation drawn below the last, SVG export. Then show me how to start it from a row of real data instead of random cells.
Data as material
Real records from the condition drive the form. The choices are what to measure, how to map a value to a mark, and what the mapping leaves out.
How it works
- Find and clean the records: one row per thing measured.
- Choose which column drives which visual quality: length, weight, spacing, angle, density.
- Decide the mapping: straight, squared, inverted, stepped.
- Draw, then label the source.
Parameters
- range
- which years, places or rows
- mapping
- how a value becomes a mark
- emphasis
- how strongly differences show
- gaps
- how missing values are drawn
Good for
Evidence and testimony: anything your identity measures, counts, or knows has gone uncounted.
The default to avoid
A chart with the axes taken off. Make the mapping a position: say why this quality, and draw what's missing as well as what's there.
Look at: Giorgia Lupi and Stefanie Posavec, Dear Data; Ryoji Ikeda's data works.
You're my collaborator. Read system.md and data/[file]. First, describe the data back to me: columns, ranges, missing values. Don't draw anything yet. Then build a minimal sketch that maps [column] to [visual quality] using the mapping in my config, and draws missing values in a way I choose.
Geometry and tiling
Grids, subdivisions and repeated shapes, and the rules for how they bend. The handbook's own Mondrian mark is one: a rectangle split again and again.
How it works
- Lay down a structure: a grid, a tiling, a subdivision.
- Decide what each cell holds.
- Add a rule that changes cells by position, time or data.
- Draw, keeping the structure visible.
Parameters
- cells
- rows and columns
- split
- how cells divide
- disorder
- how far each cell may move
- gradient
- where the rule is strongest
Good for
Order and its failure: property, planning, the survey grid, bureaucracy, a system breaking down at its edges.
The default to avoid
Decorative pattern. The grid should mean something: whose order is it, and where does it give way?
Look at: Georg Nees, Schotter (1968), which this sketch nods to; Vera Molnár; Manfred Mohr; Sol LeWitt's wall drawings.
You're my collaborator. Read system.md. Build a minimal grid sketch: rows and columns in the config, one rule that changes each cell by its position, seeded randomness, SVG export. Then add a second rule from my list, and show me the two side by side.
Combining
Two techniques can work together when each does a different job, such as a grid for order and data for what fills it. More than two usually means the rules aren't settled yet.
Record your choice and your reasons in the system worksheet, Part C, and use the researcher prompts to find more on your technique.