PROGRAMMATIC GRAPHICS STACK

how a sentence becomes a picture — no image model, just engines

RENDER ENGINES — deterministic, seeded Gabriel's brief "make a diagram of X" Data / dataset CSV, JSON, metrics Topic / concept math, papers, systems SVG + HTML diagrams · dashboards · vectors p5.js / Canvas noise · particles · shaders · 3D Manim + LaTeX animated explainers · proofs Python raster PIL · matplotlib · headless Live in browser interactive · animated PNG / SVG file print-ready · any resolution GIF / MP4 narrated explainers vector / browser generative / canvas math / animation raster / python input / external dashed = secondary path

DETERMINISTIC

  • • same seed, same output — always
  • • every label is text, every shape is math
  • • edit the source, re-render, done

RESOLUTION-FREE

  • • vectors scale infinitely; raster renders at any DPI
  • • 4K print or favicon from one file
  • • no "upscale" step needed

NO IMAGE MODEL

  • • no GPU, no diffusion, no prompt roulette
  • • data stays exact — charts show real numbers
  • • this very diagram proves the pipeline