INTERSTICES — EXPANDED FIELD
Intelligence in Nature: Artificial Life, Emergence, and Computational Form
14 October 2026

SELECTED ARTWORK
3,931 cells at step 1,050, developed from 634 founders in 17 authored regions.
3,297 recorded divisions. The earlier version selected 1,033 cells.
The new view covers 218 × 156.96 uncalibrated model units, about 3.95 times
the original view area. Its 6,400 × 4,608 pixels preserve fine cellular and
fibrous detail across the larger landscape. This is newly simulated tissue,
not an enlargement, repeated tile or copy of the original image.

The coral/burgundy cellular pigment and teal fibrous ground retain the
original material direction. The initial distribution forms branching arms
around channels. Detailed contours, growth and contact develop mechanically.
Later states reached 6,124 cells, but their channels closed and numerical
overlap increased; step 1,050 is the selected artwork, not the last checkpoint.

MODEL
The original 24-node deformable-cell implementation remains the foundation.
Changes: 17 inoculation regions; a world half-width of 128; 3,600 generated
coarse fibre seeds; a 360 × 360 nutrient grid; a higher population ceiling.
Cells enlarge according to nutrient availability and divide through their
long axis. The matrix deforms and exerts a scaled reaction on cell boundaries.
The numerical units are not calibrated to a particular biological tissue.

Colour, continuous pigment variation, nuclear markers and multiple fine
strands per simulated fibre remain artistic rendering decisions. They are
not additional organisms, physical staining chemistry or independent nuclei.
No image-generation model or source photograph was used.

REPRODUCE THE SELECTED STILL
Python 3.12, NumPy, SciPy and Pillow; dependency versions in requirements.txt.
From this directory:
  python render_section.py final_state.npz --out render --size 6400 --width 218 --cx 0 --cy 2 --layers
To regenerate the entire developmental run:
  python tissue.py --out runs --steps 1350 --matrix-reaction 0.65 --snapshot 150
Select runs/state_1050.npz for the artwork. The generator uses seed 417.
The material renderer uses seed 919.

LIVE VERSION
https://interstices-living-field.yuvxraj99.chatgpt.site
The page computes growth on the viewer's device automatically, with no
controls or video playback. It uses a reduced-cost JavaScript port with the
same initial cells, 24 membrane nodes and 360 × 360 nutrient grid. Its matrix
uses 822 selected coarse chains, and it omits the offline solver's full
membrane-node pair search. Exact shapes and division times differ.
Population is capped at 4,200 to retain open composition and bound computation.
At the cap, target growth stops and the tissue continues to settle mechanically.
The tab pauses when hidden. Reloading begins again from the founders.
The live material treatment is lighter than the high-resolution still.

VERIFICATION
The selected still has no invalid/self-crossing cell boundaries, all state
arrays are finite, identifiers are unique and division arithmetic agrees.
Residual summed-cell-area overlap is about 0.205%, recorded exactly in
validation.json. The later step-1,350 state had about 3.43% overlap and was
not selected. These are numerical checks, not biological validation.
The live integrator was checked through 1,500 steps, with a separate 350-step
zero-growth control. The application scripts were also exercised with a
worker and native Canvas. Actual browser QA was unavailable in this environment;
hardware-specific browser performance remains unmeasured.

AUTHORSHIP AND CONTEXT
Creative brief and project: Yuvraj Mittal. Assistant-assisted implementation.
Art'n'Sci Collective supplies the acknowledged aesthetic reference; its images
were not copied, traced or composited. "Intelligence in Nature" is an
interpretive theme about local responses and reciprocal environmental effects,
not a claim of cognition or consciousness in this model.

The archive's original/ directory preserves the complete earlier study.
The expanded/ directory holds the new source, state, layers and records.
