Updated: 2026-01-03
| File | Description |
|---|---|
LLM_SYSTEM_PROMPTS.py |
30+ system prompts for game-playing LLM agents |
TACTICAL_GUIDE.md |
Semiotic analysis methodology for ARC puzzles |
CROSS_DOMAIN_INSIGHTS.md |
55 eigenvalue-based cross-domain insights |
MULTI_MODEL_PROMPTS.md |
Prompts for multi-model coordination |
ARC_AGI_3_INSIGHTS.md |
Interactive game architecture notes |
ANTI_CHEAT_ANALYSIS.md |
Human-like agent behavior patterns |
ratchet_loss.py |
Asymmetric ratcheting loss function |
| File | Description |
|---|---|
operator.rs |
GridOperator trait with compose/invert |
hypothesis.rs |
Beam search over operator compositions |
grammar.rs |
Formal DSL grammar for grid operations |
grid.rs |
64x64 packed 4-bit grid representation |
symmetry.rs |
Symmetry detection and classification |
prior_bank.rs |
Amortized inference via compiled priors |
discrimination.rs |
Hypothesis space partitioning |
simulator.rs |
World simulation for planning |
planner.rs |
A* planning with viability tracking |
memory.rs |
3-tier memory system |
cognitive_regulation.rs |
Self-monitoring and correction |
critical_analysis.rs |
Adversarial trap detection |
All files use descriptive names:
LLM_SYSTEM_PROMPTS.py— System prompts for LLM agentsTACTICAL_GUIDE.md— Step-by-step methodology guideCROSS_DOMAIN_INSIGHTS.md— Theoretical cross-domain analysisMULTI_MODEL_PROMPTS.md— Multi-model coordination promptsARC_AGI_3_INSIGHTS.md— ARC-AGI-3 specific insightsANTI_CHEAT_ANALYSIS.md— Human-like behavior patternsratchet_loss.py— Monotonic improvement loss function
- Code: MIT License
- Documentation: CC-BY-4.0
Maintained by Crystalline Labs