INTELLIGENCE IN NATURE
Tissue-section study — working title: INTERSTICES
13 October 2026

THE WORK
A computational tissue develops inside a fibrous environment. The cells
enlarge and divide; neighbours compete for space; surrounding elastic chains
are displaced and exert a reaction on the cells. The resulting state is
interpreted as a coloured tissue section.

The narrative connection to "Intelligence in Nature" is interpretive:
large-scale organisation can arise from many local responses, while the
developing system changes the environment that constrains it. This is not a
claim of cognition, learning or consciousness in the model.

Art'n'Sci Collective supplied the aesthetic reference for colour, dense
material detail and section-like composition. No source photograph was used,
traced, composited, or used as a target image. No image-generation model was
used. This is an assistant-implemented study under the user's creative brief;
it should not be described as hand-painted or as a validated biological model.

WHAT IS ACTUALLY SIMULATED
- Closed deformable cell boundaries with 24 membrane samples per cell.
- Area elasticity, membrane springs, local smoothing and contact exclusion.
- Adhesion varying between matching and different cell types.
- A diffusing nutrient field with local uptake and replenishment.
- Nutrient-dependent enlargement and explicit cell division with parent IDs.
- Elastic fibre chains, initially arranged by an authored curved field.
- Cell-to-fibre displacement and a scaled reaction from fibres to cells.

WHAT IS AN ARTISTIC RENDERING CHOICE
- The initial inoculation regions and initial fibre field are authored.
- Colours, crop, pigment absorption texture and microscopic granulation.
- Nuclear marks are graphical markers linked to cell position and shape;
  nuclei are not independently simulated mechanical bodies.
- Multiple fine strands around each coarse simulated fibre are material
  detail, not additional simulated agents.
- Fibre strain influences the tint of the fine strands.
- No actual histological staining chemistry or microscope optics is claimed.

IMPORTANT SIMPLIFICATIONS
Cells are constrained to star-shaped boundaries. Membrane samples are
redistributed, and support-interval contact projection prevents appreciable
interpenetration. This is an original simplified off-lattice model, not the
implementation or reproduction of Gardiner et al., Nematbakhsh et al., Andy
Lomas, or any other published solver. Fibres are coarse elastic chains with
weak background anchoring, not a molecular collagen model. The world and
simulation steps have no calibrated biological units. Two initial cell types
are assigned rather than differentiated by a gene-regulatory network.

FILES
tissue.py                    Complete simulation.
render_section.py            Material renderer, separate from the dynamics.
validate.py                  Numerical and control checks (requires shapely).
final_state.npz              Selected cells, fibres, nutrient and lineage IDs.
runs/final/                 Checkpoints 0, 450, 900, 1350 and 1500, plus records.
runs/final/history.json      Population, area and fibre-displacement record.
runs/final/divisions.json    Explicit parent/daughter division events.
runs/final/parameters.json   Model settings for the saved run.
runs/final/validation.json   Measured results and their limited scope.
runs/final_no_growth/       Control checkpoint and records.
runs/final_no_reaction/     Control checkpoint and records.
layers/                      Separate ground, pigment, nuclei and cell-mask PNGs.
Interstices.png              Selected artwork, 2400 x 2688 pixels.
Development.png              Four saved states at the same view and scale.
requirements.txt             Dependency versions.

REPRODUCE
Python 3.12; numpy 2.3.5; scipy 1.17.0; Pillow 12.3.0.
The optional geometric verification uses shapely 2.2.0.

pip install -r requirements.txt

To render the supplied state (without rerunning the simulation):
python render_section.py final_state.npz --out render --size 2400 --width 88 --cy -3 --layers
python validate.py runs

To regenerate the simulation and then render:

python tissue.py --out runs/final --steps 1500 --matrix-reaction 0.65 --snapshot 150
python render_section.py runs/final/state_1350.npz --out render --size 2400 --width 88 --cy -3 --layers

Controls used for validation:
python tissue.py --out runs/final_no_growth --steps 300 --growth 0 --matrix-reaction 0.65 --snapshot 150
python tissue.py --out runs/final_no_reaction --steps 900 --matrix-reaction 0 --snapshot 150
python validate.py runs

The artwork selects step 1350 from a 1500-step run. This keeps more visible
space between tissue regions. Numerical checks also found less than 0.1%
residual cell-area overlap at the selected state, compared with approximately
1.27% in the denser step-1500 state. The selected state is saved as both its
numbered checkpoint and final_state.npz. "Final" means selected artwork state,
not the last simulated step. This limitation is retained in validation.json.

ARTISTIC CONTROLS
No coding is needed to recolour or composite the supplied image layers in an
image editor. Keep the colour layer and nuclear layer separate when testing
palettes. The mask isolates cellular regions from the surrounding matrix.
The final image also has a very light optical grain added after compositing.
The continuous pigment variation is artistic material, not a simulated field.

The renderer's --width, --cx and --cy control the view without changing
the simulation. Its --seed changes the material interpretation only.
The simulation's --seed changes the growth realisation and --growth changes
nutrient-dependent enlargement. --matrix-reaction changes mechanical coupling.
The main palette and material coefficients are near the top and cell-render
sections of render_section.py. These edits do not change the simulated state.

BACKGROUND REFERENCES (NOT CODE DEPENDENCIES)
Art'n'Sci Collective: https://art-n-sci.art/
Gardiner et al. (2015), Discrete Element Framework for Modelling Extracellular
Matrix, Deformable Cells and Subcellular Components:
https://doi.org/10.1371/journal.pcbi.1004544
Nematbakhsh et al. (2020), Epithelial organ shape is generated by patterned
actomyosin contractility and maintained by the extracellular matrix:
https://doi.org/10.1371/journal.pcbi.1008105
