BSP Documentation
Bitset System for Prediction: A CPU-friendly approach to continuous learning without Transformers.
🚀 Start Here
- CONCEPT-001 New The Anti-Transformer Manifesto Why bitsets? Why CPU? Understanding the philosophy behind the architecture.
- GUIDE-001 New The Lifecycle of a Turn A visual journey of data through the engine, step-by-step.
- QUICKSTART Running the Demo Get the system running on your local machine in minutes.
🧠 Core Specs (The Engine)
- DS-001 Core Architecture System design, components, and high-level data flow.
- DS-002 Data Structures Bitsets, groups, inverted indexes, and memory layout.
- DS-003 Learning Algorithms Online learning mechanics, surprise minimization, and updates.
- DS-004 Deduction Engine Temporal links and multi-hop reasoning logic.
✨ Features & Behaviors
- DS-009 Sequence Generation From abstract groups to ordered text output.
- DS-013 Context & Memory Maintaining coherence across conversation turns.
- DS-005 RL & Feedback Modulating learning with rewards and importance signals.
- DS-010 Sleep & Consolidation Merging redundant concepts during idle time.
🛠️ Operations & Dev
- OPT-001 New Optimization Journey Experiments, results, and future work on compression performance.
- DS-006 Server API HTTP interface specifications.
- DS-007 Serialization Saving and loading session states efficiently.
- DEV-001 New Tuning & Debugging Practical guide to configuration flags and troubleshooting.
- DS-017 Optimization & Flags Performance tuning and system configuration options.