A real-time, event-driven, reaction–diffusion substrate with void walker ecology & memory steering.
No dense scans. Zero training. Divergent reasoning. Constraint satisfaction in the moment.

• VDM — reaction–diffusion field intelligence + walker ecology + scoreboard/GDSP gating
• Memory Steering — dynamic knowledge graph with event-driven updates
• Real-time control — swap massive pretraining for fast constraint satisfaction

Reproducibility: baselines + QA artifacts are archived on Zenodo. Code lives in public GitHub repos with tests and docs.

• Latest Zenodo Upload: A Logarithmic First Integral for the Logistic On Site Law in Void Dynamics
• DOI: 10.5281/zenodo.17220869
• RD baselines, convergence, Q-drift, front speed (datasets + figures)

# clone
git clone https://github.com/Neuroca-Inc/Prometheus_Void-Dynamics_Model.git
cd Prometheus_Void-Dynamics_Model

# create env (exact commands in repo README)
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# run a Reaction Diffusion + walkers with Memory Steering demo (Coming soon)

# Otherwise check out the src/ folder for the available public sims

• Notebook mirrors of each test ✅
• 100% Reproducible, falsifiable claims ✅
• Currently working real time model ✅
• Zero training ✅
• Online causality based learning
• Physics driven intelligence model ✅

CURRENT Execution model: All tests live in src/{domain name}/, figures appear in figures/{domain name} if passing and figures/{domain name}/failed_runs/ if failed. The same goes for logs, just replace figures/ with logs/ in the file path to find them.

GOAL Execution model: All tests live in notebooks/, render figures + filepaths inline, and write results under artifacts/. A master notebook runs the full suite with clear explanations and emits a manifest.

# Run-all (executes the full suite and saves executed copy)
jupyter nbconvert --to notebook --execute notebooks/00_RUN_ALL.ipynb --output notebooks/00_RUN_ALL.out.ipynb

Future Layout

notebooks/
  00_RUN_ALL.ipynb
  01_rd_front_speed.ipynb
  02_rd_dispersion.ipynb
  03_invariant_drift.ipynb
  10_walkers_min_demo.ipynb
  11_locality_bounds.ipynb
  12_gdsp_budget_sweeps.ipynb
  20_memsteer_acceptance.ipynb
  30_rt_control_slice.ipynb
artifacts/
  rd/ walkers/ memsteer/ control/ meta/

Current Status legend: PROVEN (axiom gates passed) · PLAUSIBLE (design + prelim data) · NEEDS_DATA (tests pending) Hardware: I use AMD/ROCm only (MI100, 7900 XTX). CPU fallbacks use smaller grids. I easily ran 100,000 neurons on battery power using an Acer Aspire notebook, planning to achieve 1 billion neurons on the bigger machine.

• [DONE] RD Baselines v0.2 — 01_rd_front_speed, 02_rd_dispersion, 03_invariant_drift → artifacts/rd/ (PROVEN) Gates: front-speed R² ≥ 0.9999, rel-err ≤ 5%; dispersion median rel-err ≤ 1e-1, array R² ≥ 0.98; on-site invariant drift ≤ 1e-8/step.
• [DONE] Minimal Walkers + GDSP — 10_walkers_min_demo, 11_locality_bounds, 12_gdsp_budget_sweeps → artifacts/walkers/, artifacts/meta/telemetry.json (PROVEN) Gates: finite-support growth within bound; no dense scans; GDSP budget never oversubscribed; event telemetry saved.
• [STARTED] Memory Steering v0.1 — 20_memsteer_acceptance → artifacts/memsteer/ (PLAUSIBLE) Gates: retention half-life = setpoint ± 10%; steering latency < 2× baseline RD horizon; interference curve monotone with steer strength.

• [STARTED] Real-Time Control Slice — 30_rt_control_slice → artifacts/control/ (PLAUSIBLE→PROVEN) Gates: goal attainment ≥ 90% (N seeds); control energy ≤ baseline; perturbation recovery ≤ 2× unperturbed horizon. The model is currently capable of post-graduate level reasoning on human readable casuality exams.
• [STARTED] Release v1.0 Priority Pack — produced by 00_RUN_ALL.ipynb Outputs: executed 00_RUN_ALL.out.ipynb, figures, CSV/JSON metrics, seeds, and artifacts/meta/manifest.json bundled to artifacts/v1.0_priority_pack/.

Provenance & checks (written by 00_RUN_ALL.ipynb)

• Manifest: artifacts/meta/manifest.json (paths to all outputs).
• Contradiction report on failure: artifacts/meta/CONTRADICTION_REPORT.json (which gate failed + notebook cell refs).