4.0.2

• Fixed action limit: max_actions_per_game: 1 → 5 (enables multi-action sequences, which dramatically increases the amount of progress that can be made.)
• Enabled all learning features: Action intelligence, knowledge transfer, pattern recognition, meta-learning
• Eliminated runtime errors: Added missing Governor methods (record_action_result, analyze_performance_and_recommend)
• Stabilized architecture: Fixed missing ContinuousLearningLoop attributes (game_reset_tracker, training_state, rate_limiter)
• Updated ARC API integration: Real-time scorecard tracking with rate limiting
• Enhanced training scripts: Unified configuration across all training modules
• Added LLM Prompt Script - for automating the system. Note, you'll need to add this to the VSCODE "Rules" or whatever IDE you are using with a LLM copilot system. to keep the system working, you may need a macro script to keep the system "continuing" every hour or so in the terminal prompt.

4.0.1

Tabula Rasa v4 adds several key improvements over v3, focusing on enhanced memory systems, better training scripts, and improved ARC-3 integration.

🔄 Key Updates from v3

New Training Infrastructure

• 9-Hour Training Scripts: run_9hour_scaled_training.py and run_9hour_simple_training.py for extended training sessions
• Intelligent Resource Management: Automatic RAM detection and session scaling
• Windows Batch Files: start_intelligent_training.bat for easier Windows usage
• Enhanced Log Management: Automatic log rotation and cleanup

Memory System Improvements

• Cross-Session Learning: Persistent memory across training sessions with breakthrough detection
• Memory Decay System: Strategic memory prioritization to prevent flooding
• Hierarchical Memory Protection: 5-tier system for different achievement levels