Graphglyph

Encode text as reversible blue-and-orange unit-distance graph images.

The visual style is based on the finite illustration from OpenAI's unit-distance paper: points of the form

z = a + b i + c rho + d i rho
rho = exp(2 pi i / 3)

with edges drawn between projected points at Euclidean distance 1. See Planar Point Sets with Many Unit Distances.

How It Works

• Text is normalized with Unicode NFKC and encoded as UTF-8.
• Long payloads are zlib-compressed.
• A packet header stores magic bytes, version, flags, seed, length, and CRC32.
• Payload nibbles are distributed through seeded graph cells.
• Each cell encodes four bits by making one edge in each candidate edge pair stronger than the other.
• The same text seed also varies the coefficient window, point count, basis family, edge weights, and payload placement.

The SVG and JSON outputs are decodable. PNG outputs are presentation previews.

Usage

python3 graph_cipher.py encode "you are loved immensely" \
  -o examples/you_are_loved_immensely.svg \
  --json examples/you_are_loved_immensely.json

python3 graph_cipher.py decode examples/you_are_loved_immensely.svg

Colors are presentation-only and do not affect decoding:

python3 graph_cipher.py encode "goblins" -o goblins_dark.svg \
  --edge-color "#7c8cff" \
  --node-color "#ffd166" \
  --node-stroke-color "#d08700" \
  --background-color "#0b1020"

Use --variant-strength 0 for the exact finite algebraic box. The default is 0.75, which gives more visible variation between texts.

Examples

You Are Loved Immensely

SVG | PNG

Goblins

SVG | PNG

Meditations On Moloch

Encoded from Scott Alexander's Meditations On Moloch.

PNG

The large meditations_on_moloch.svg is intentionally not included.

All code written by GPT 5.5.