Genesis Engine

⚠️ CORRECTION — 29 July 2026

The headline result below is WITHDRAWN. The reported Clock→Map ordering was produced by measurement artifacts, not by the dynamics it was attributed to. Eight defects were found in the authors' own re-analysis; each alone is sufficient to produce the 100% figure. Measured without the phase detector, the ordering holds in ≈1% of runs.

The underlying physical hypothesis was then tested directly under imposed division regularity (1,350 runs) and can be neither confirmed nor refuted with this design.

Full audit, data and analysis scripts: paper/v6_supplement/ · Correction note: CORRECTION_NOTE_DRAFT.md · Automated checklist that recovers all eight defects: tools/detector_audit.py

The text below is retained unaltered as the published record.

---

Sequential Assembly of Biological Agency: Division Regularity Precedes Pattern Stability in Evolving Protocells

---

Abstract

The origin of biological agency requires the coordinated emergence of three sub-systems — a Clock (metabolic rhythm driving division), a Map (stable spatial patterning), and an Engine (thermodynamic coupling that exports entropy). We ask whether these sub-systems must appear in a specific order, or whether any ordering suffices. Across 500 independent 1D Monte Carlo runs and 200 independent 2D runs on a spherical manifold (642-vertex icosphere, cotangent Laplace–Beltrami operator), the Clock phase preceded the Map phase in 1,165 of 1,165 runs where both transitions occurred (100 %; combined binomial p < 10⁻²⁰⁰; Hedges' g = 6.34). Twelve

Twelve ablation conditions spanning factor-of-2 perturbations of lipid supply, reaction-diffusion noise, growth perturbation, and stability window each preserved 100 % Clock-before-Map ordering. The ordering is topology- and parameter-invariant: pattern formation is a filter on top of a running Clock, not a parallel process. These findings reframe abiogenesis as the sequential assembly of necessary preconditions rather than the simultaneous emergence of all capabilities.

Key Result

Clock always precedes Map. 1,165 / 1,165 = 100 %. Zero violations.

WITHDRAWN 29 July 2026. Every figure in this box is an artifact.

claimed	status
1,165 / 1,165 = 100 %	detector latched Map only after Clock — no other outcome was reachable
p = 8.01 × 10⁻¹⁴⁶ (1D), 2.49 × 10⁻⁶⁰ (2D)	tests against an unreachable null
Combined p < 10⁻²⁰⁰	denominator double-counts the shared OAT baseline: 1,845 → 1,554 distinct runs
Hedges' g = 9.41 / 3.27 / 6.34	computed on final_pop, which the grouping variable causally determines; the 2D arm is n = 2

Corrected figure: Clock precedes Map in ≈1 % of runs (1.0 % by 1-tick resample N=100; 0.7 % by threshold grid N=150; 0.0 % in 2D across all nine threshold cells).

See paper/v6_supplement/ for the full audit.

---

Repository Layout

genesis-engine/
├── genesis_engine.py          # 1D reaction-diffusion engine + protocell dynamics
├── genesis_engine_2d.py       # 2D spherical-manifold engine (icosphere + LB operator)
├── mesh_utils.py              # Icosphere subdivision, cotangent Laplace-Beltrami, sparse CSR
├── run_monte_carlo.py         # 1D Monte Carlo driver (N=500)
├── run_monte_carlo_2d.py      # 2D Monte Carlo driver (N=200)
├── run_pilot_2d.py            # 2D PASS/FAIL gate (N=10) run before overnight
├── run_ablations.py           # Twelve-condition ablation grid (N=100 × 12)
├── calibrate_2d.py            # α* sweep for 2D ↔ 1D statistical calibration
├── analyze_results.py         # Aggregate summaries, Hedges' g, binomial + Wilcoxon
│
├── paper/                     # Manuscript, figures, canonical aggregate data
│   ├── genesis_paper_v3.md    # Source markdown (latest revision)
│   ├── paper_data.json        # Single source of truth for all numerical claims
│   ├── FIGURES.md             # Caption and provenance for all 7 figures
│   ├── figures/               # fig1 … fig7 (PNG, archival resolution)
│   ├── scripts/               # Figure-generation scripts
│   └── submission/            # SSRN-formatted PDF + LaTeX source + .bib
│
├── web/                       # Interactive dashboard (static HTML + JS)
│   ├── index.html             # Three tabs: 1D live, 2D sphere live, results
│   ├── genesis.js             # 1D engine (JS port of genesis_engine.py)
│   ├── genesis_2d.js          # 2D sphere engine (JS port of genesis_engine_2d.py)
│   ├── icosphere.js           # Browser-side mesh builder
│   ├── app.js                 # UI glue: tabs, controls, charts
│   ├── results.js             # Monte Carlo tab — loads paper/paper_data.json
│   ├── style.css
│   ├── server.py              # Optional local server (adds /api/status; not needed for static hosting)
│   └── start.sh               # Convenience launcher
│
├── results/                   # 1D Monte Carlo raw + aggregate outputs
│   ├── summary.csv            # N=500 per-run metrics
│   ├── timeseries/            # Per-run time-series CSVs
│   ├── ablations/             # 12-condition ablation outputs
│   └── figures/               # Diagnostic figures
│
└── results_2d/                # 2D Monte Carlo raw + aggregate outputs
    ├── summary.csv            # N=200 per-run metrics
    └── timeseries/

---

Quickstart

Requires Python 3.10 or newer.

git clone https://github.com/Ouroboros-Research-Institute/genesis-engine.git
cd genesis-engine
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

Run a single 1D simulation

python3 genesis_engine.py

Launch the live dashboard

cd web && ./start.sh          # serves http://localhost:3000

Open http://localhost:3000 in any modern browser. Three tabs:

• 1D Live Simulation — reaction-diffusion on a 1D genome, canvas-rendered with latched phase transitions and live stats.
• 2D Sphere — icosphere surface rendering with up to 30 simultaneous protocells, each tracked through its own Clock/Map transitions.
• Monte Carlo Results — aggregate statistics loaded directly from paper/paper_data.json.

---

Reproducing the Paper Results

1. 1D Monte Carlo (N = 500, ~8 min on 16 workers)

python3 run_monte_carlo.py -n 500 -t 80000 -w 16
python3 analyze_results.py results/summary.csv

2. 2D Monte Carlo (N = 200, ~30 min on 16 workers)

python3 run_pilot_2d.py                         # PASS/FAIL gate first
python3 run_monte_carlo_2d.py -n 200 -t 40000 -w 16
python3 analyze_results.py results_2d/summary.csv

3. Ablations (N = 100 × 12 conditions, ~20 min)

python3 run_ablations.py

4. Regenerate figures

cd paper/scripts
python3 render_phase_transitions.py
python3 render_clock_vs_map.py
python3 render_example_timeseries.py
python3 render_ablation_table.py
python3 render_2d_sphere_results.py
python3 capture_fig6_dashboard.py        # requires playwright; live server must be running
python3 render_comparison_table.py

5. Rebuild the manuscript PDF

cd paper/submission
tectonic -X compile genesis_paper_farina_2026.tex

---

Canonical Parameters

All shared between 1D and 2D engines unless noted:

Parameter	Value	Role
MIN_RADIUS	8.0	Min protocell radius (division gate)
LIPID_SUPPLY	0.015	Ambient lipid accumulation rate
RD_NOISE	0.004	Reaction-diffusion stochastic amplitude
PHASE_B_CV	0.25	Division-regularity (Clock) CV threshold
SAMPLE_INTERVAL	50	Tick interval for phase-metric sampling
STAB_WINDOW	40	Window size for pattern stability metric
PHASE_D_GEN	5	Generations required to latch Phase D
ALPHA_RESCALE	0.40	2D-only; empirically calibrated to match 1D
RD_STEPS	90	2D-only; RD substeps per simulation tick

The 2D mesh is an icosphere, subdivision = 3 → 642 vertices, 1,280 faces. The Laplacian is the cotangent Laplace–Beltrami operator, stored as a sparse CSR with 4,482 non-zeros.

---

Citation

@article{farina2026genesis,
  author  = {Farina, Micka{\"e}l},
  title   = {Sequential Assembly of Biological Agency: Division Regularity Precedes Pattern Stability in Evolving Protocells},
  journal = {SSRN Electronic Journal},
  year    = {2026},
  url     = {https://papers.ssrn.com},
}

See CITATION.cff for machine-readable metadata.

---

Contact

Mickaël Farina AVA Digital L.L.C. 1603 Capitol Ave Ste 415 #258343 Cheyenne, WY 82001, USA D-U-N-S: 136864260 Email: mikarina@avadigital.ai

---

License

Released under the MIT License. You are free to use, modify, and redistribute the code and data. If you reproduce the scientific results in an academic context, please cite the paper.