Source sample field
The image is sampled along a Hilbert path and stored as a delta from the previous sample. On decode it isn't a grid of cells — it's a single painted worm that fills the plane. No image loaded yet shows a procedural sample.
Filament substrate
The body of the worm. Together with thread these set how far it reads as an image: high thread · thin girth · low bleed · low relax resolves the picture; the opposite dissolves it into pure ooze. Render-only — the same bytes can ooze thick or run thin — and saved into the file so a specimen re-opens looking the same.
Damage none
The body is a stream of deltas, so corruption integrates forward and flows down the worm. The signature loss is the travelling stain, not speckle.
Corrupt as data locked
Open the .vermis as raw data and glitch it by hand. A length-changing find/replace shifts the curve phase downstream — colour reassigns along the thread and scrambles in a path-coherent way. The 16-byte header stays locked (order, anchor, substrate), so the file always still decodes.
Specimen · .vermis v1
- magic
- VERM · 16-byte header
- record
- 3 B · dR dG dB · per sample
- substrate
- one thread · no grid, no cells
Export