{"data":{"kind":"file","path":"README.md","version_id":"crkvz8ay5c9x3v61elknx4vw","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2110,"modified_at":"2026-08-30T02:32:23.488000","content_hash":"77dcf8600e41c7db112b7fc0734f97e4ccc1ec9b6e328b52cbe99ac8757eb226"},"entries":[],"content":"# Cellular Automata / Conway's Game of Life Designer\n\nA verifiers environment that evaluates an LLM's ability to design, implement, analyze, and reason about cellular automata — with a focus on Conway's Game of Life and its generalizations.\n\n## Overview\n\nThis environment tests deep understanding of cellular automata through progressively challenging tasks:\n\n1. **Rule Implementation** — Correctly implement Game of Life transition rules (B3/S23)\n2. **Pattern Design** — Design specific patterns (still lifes, oscillators, spaceships, guns)\n3. **Pattern Analysis** — Analyze existing patterns: detect period, track population, compute bounding box\n4. **Custom Rules** — Define and implement alternative cellular automata rules (HighLife, Day & Night, Seeds)\n5. **Proof & Reasoning** — Prove properties about patterns (e.g., bounding box growth, population bounds)\n\n## Task Types\n\n### Single-Turn Tasks\n- Implement a step function for a given rule string\n- Design a pattern that satisfies specific constraints\n- Analyze a pattern's behavior over N generations\n- Compare two rule sets and predict behavioral differences\n\n### Multi-Turn Tasks\n- Iterative pattern refinement: given feedback, improve a design\n- Debug an incorrect implementation\n- Guided exploration of a rule space\n\n## Scoring\n\nEach response is evaluated on multiple dimensions:\n\n- **Correctness** (weight 0.40): Does the code produce valid Game of Life behavior?\n- **Pattern Validity** (weight 0.25): Does the designed pattern actually exhibit the claimed behavior?\n- **Analysis Accuracy** (weight 0.20): Are computed metrics (period, population, bounding box) correct?\n- **Code Quality** (weight 0.15): Is the implementation clean, efficient, and well-documented?\n\n## Configuration\n\n```python\nenv = load_environment(level=0)       # Easy: basic rules + simple patterns\nenv = load_environment(level=1)       # Medium: pattern design + analysis\nenv = load_environment(level=2)       # Hard: custom rules + proofs\nenv = load_environment(level=3)       # Expert: multi-turn design challenges\n```\n\n## Version\n\nV0.0.0 — Initial release\n","encoding":"utf-8","truncated":false,"total_bytes":2110},"status":null}