# Inhabited Ground A first inhabited-landscape study for **Intelligence in Nature: Artificial Life, Emergence, and Computational Form**. The proposal is visible in a simple relationship: **life changes a place, and the altered place supports further life**. Green builders grow branching strands and lay a pale, persistent bed. Coral settlers can establish only in that constructed material. They consume it, return nutrients, and leave sites where dormant builders can begin again. The same ground remains in view throughout. This is an invented artificial ecology, presented as an artwork study. Its local rules were designed for this relationship; it is not a calibrated account of a particular natural organism, and the behaviour is not evidence of consciousness or biological intelligence. ## What is simulated - Continuous builder tips sense nearby nutrient levels, moisture, crowding and accumulated material. They turn, branch, deposit living strands and stop in unsuitable conditions. - A lattice tracks living builders, settlers, age, nutrients, material thickness, construction history and settler residue. - Builder secretion spreads locally into a porous bed. This spread is part of the state that affects later settlement. - Settlers reproduce locally and germinate only where sufficient material exists. They digest the bed, die as it is exhausted and return nutrients. - Retained material changes nutrient diffusion. Builder reawakening requires prior settler activity and enough local nutrients. - Growth, occupation, depletion and regrowth occur together in different parts of the image. The three film captions name the changing emphasis, not three externally switched simulation modes. ## Artistic decisions The initial terrain, moisture distribution, water gaps and seven unequal founding populations are composed starting conditions. The diagonal waterway is part of that starting environment; the organisms do not excavate a river. The branching routes, encounters, material accumulation, settlement and subsequent renewal are computed from local interactions. Colour distinguishes living roles and varies with age, nutrient availability and material thickness. Pale deposits are retained habitat. Darker mineral regions and water provide a stable frame. Matte illumination uses the actual material field. Small filtering in the renderer provides coverage antialiasing and broad illumination; there is no fitted mesh or final sculpting pass. The large PNG is an enlarged rendering of the 1600 × 1000 simulation grid. Upsampling adds display size, not new simulated detail. The chosen still's step is recorded in `selection.json`. The development strip shows the same camera at steps 120, 360 and 660. ## Film The MP4 records one trajectory from step 0 through 660, sampling every third step. It is encoded at 24 fps from 12 state samples per second, followed by a 2.5-second hold. Repeated encoded frames are presentation timing; intermediate states are not fabricated. The film is a finite record, not a seamless loop. It has no audio. The captions make the intended relationship accessible without a separate artist statement: 1. Green builders grow and leave a lasting habitat. 2. Coral settlers inhabit the material the builders leave. 3. Settlers recycle the material. New growth returns. ## Reproduce Python 3.12 and FFmpeg are required. Install the recorded Python dependencies: ```sh python3 -m pip install -r requirements.txt python3 make_film.py --out output ``` The defaults reproduce the delivered trajectory: 1600 × 1000, seed 718, 660 steps. Output includes the film, the development strip, intermediate candidate stills, checkpoints and a trajectory log. Rendering and encoding take several minutes on a CPU. Recreate the selected enlarged still after the film run: ```sh python3 select_still.py --out output ``` Run the meaningful checks on a generated checkpoint: ```sh python3 verify.py output/state_0360.npz --out output/validation.json ``` `landscape.py` contains the model; `optical.py` contains the artistic rendering. They are separate so display changes cannot alter behaviour. Checkpoints include all state arrays and the random generator state. Platform, dependency versions and fonts may affect bitwise media reproduction. The save/resume check verifies identical state continuation in the recorded environment. ## Checks `validation.json` records finite/bounded fields, exclusion of living occupants from open water, deterministic save/resume, removal of the constructed habitat, and an intervention disabling nutrient recycling from the same checkpoint. The intervention tests the programmed causal relationship in this model. It is not a statistical claim about natural ecosystems. The `builder_return_sites` metric counts places where a builder was newly laid after substantial prior settler activity. The broader `reoccupied` counter includes all later occupations, including settlers. `ever_built` counts sites touched by builders; the secreted bed can spread beyond them. ## Scope of this study The source provides a running simulation and a reproducible film. It is a first artistic study, with deliberately limited ecological roles. It does not yet include an interactive browser installation, audience input, or a long-run evolutionary process. The visual result should be judged as a proposed direction, not as a settled final artwork. ## Recorded result The 660-step trajectory ends with 317,077 sites that have received a builder after prior settler activity. In the checkpoint intervention, removing all accumulated habitat caused all 512,066 settlers to disappear on the next update. After 120 further steps, mean nutrient availability in the original inhabited region was 0.751 with recycling and 0.435 without it; cumulative builder-return sites were 70,117 versus 63,814. These are outcomes of this one seeded artificial ecology, not biological measurements. The delivered MP4 is 1600 × 1080, 24 fps, 502 frames, approximately 20.92 seconds. The artwork occupies a 1600 × 1000 field above a small caption area.