ca-constraint-lab
CA Constraint Lab
Interactive cellular automata visualization for testing the message-passing invariant.
Quick Start
open ~/working/wanderland/experiments/ca-playground/index.htmlOr serve it:
cd ~/working/wanderland/experiments/ca-playground && python -m http.server 8888
# Then open http://localhost:8888Features
Rule Selection
Test different constraint-breaking scenarios:
| Rule | Constraint | What It Tests |
|---|---|---|
| Game of Life | Baseline | Standard GoL |
| Break Conservation | 2a | Random cell flips (5%) |
| Break Memory | 2b | State ignores self, only neighbors |
| Break Hierarchy | 2c | Random threshold per cell per step |
| Enhance Prediction | 2d+ | Momentum (30% flip probability) |
| Overprediction | 2d- | Anticipate future neighbor states |
Environment Noise
Slider to add stochasticity to neighbor perception. Tests the refined hypothesis:
- Deterministic (0% noise): Overprediction helps
- Stochastic (>X% noise): Overprediction hurts (hallucination)
- Phase transition at critical noise level
Noise Sweep Experiment
Automated experiment that:
- Runs GoL and Overprediction at 10 noise levels (0-30%)
- Measures temporal variance after 200 steps
- Identifies the crossover point (if any)
- Exports results as JSON
Metrics
- Step: Current iteration
- Alive Cells: Population count
- Density: Alive/Total
- Temporal Variance: Instability over last 50 steps
Recording
- Record runs to capture step-by-step data
- Export as JSON for analysis
- Attach to Oculus cases
The Hypothesis
The original framework predicted "overprediction β oscillation/hallucination". But the CA test showed overprediction HELPING in deterministic environments.
Refined hypothesis: Prediction overshoot is only pathological when there's a mismatch between prediction model and environment stochasticity.
| Environment | Prediction Effect |
|---|---|
| Deterministic | Helps (accurate lookahead) |
| Stochastic | Hurts (fitting noise) |
| Adversarial | Exploited |
Case
Part of: [[case:task-54c00d7b-9129-4db2-87a3-8b758f65fb4e|The Case of the Constraint-Breaking Automata]]
Source
/Users/graemefawcett/working/wanderland/experiments/ca-playground/index.html/Users/graemefawcett/working/wanderland/experiments/ca_constraint_breaking.py(original CLI version)
Provenance
- Created: 2026-01-06
- Context: Testing message-passing invariant falsifiability
- Status: π’ Active tool
North
slots:
- context:
- CA lab is an implementation of the experiments documented in falsifiable-experiments
slug: falsifiable-experiments-message-passing