Skip to content

About

The research behind wake: a uji fork with seeded randomness and time shear, a plotter series, contact sheets, and the tools that measured them.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

wake-research

The research behind wake (live · case study): how I took apart uji, a generative-art toy by Noah Doersing, measured it, extended it, and found out what survives a pen plotter.

20 seeds from the plotter series

What's here

Folder What it is
uji/ A fork of uji (submodule, wake-series branch) with two options added. seed makes every random draw reproducible, so a link redraws the same picture. timeshear gives each line segment its own clock. With both off, output is byte-identical to upstream.
tools/ Headless rendering with Playwright (render.mjs), parameter sweeps into contact sheets (sweep.mjs), the seeded plotter series (series.mjs), and plotcheck.py, a vpype-based plot-time and ink-density check for A4 with a 0.3 mm pen.
out/contact-sheets/ The sweeps, including the failed first timeshear (01–03) and the version that worked (05–07).
out/svg/series/, out/plot-ready/ The 20-seed plotter series and the plot-ready finals.
docs/findings.md The working log: how uji draws, what seeding fixed, the 957-minute plot, the paper choosing circles, and the license reading.
docs/statement.md The artist statement for the series.
docs/wake-model.md, docs/mockups/ The model and the interaction concepts that became wake.
.claude/agents/ The three small helper agents (on Claude Haiku) that mapped code, ran sweeps and ran plot checks.

Findings in one paragraph each

  • uji's share links weren't reproducible. All randomness went through one unseeded Math.random, and jitter was on by default.
  • My first time shear did nothing. Shifting the rotation's timing per point only slid points along the same circle; the contact sheets showed 100 identical circles. Giving each point its own clock (no drawing, no motion until its time) worked.
  • The screen look is unplottable. A typical render needs about 957 minutes and 2.1 km of ink. Plottable pieces need about 60 well-spaced rings: 3–5 minutes, no over-inked cells.
  • The paper chose circles. With seeds deciding the parameters and the plot check rejecting over-inking, squares and triangles always failed (corners fold lines onto each other). All 20 survivors are open circles.

Run

npm install && npx playwright install chromium
python3.12 -m venv .venv && .venv/bin/pip install vpype     # vpype needs Python ≤ 3.12 here (Shapely)
node tools/render.mjs "seed=7,timeshear=60,expansionhori=0.99,expansionverti=0.99,jitter=0.3,iterations=80" out/svg/example --png
tools/plotcheck.sh out/svg/example.svg

Clone with git clone --recurse-submodules to get the uji fork.

Credit

uji is by Noah Doersing (MIT, with a clause forbidding NFTs), and this work stays within it. Plotting statistics come from vpype by Antoine Beyeler. Made by Preet Panchal, with Claude Code.

About

The research behind wake: a uji fork with seeded randomness and time shear, a plotter series, contact sheets, and the tools that measured them.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages