[INIT] Mounting Research Pipeline...
The Researcher's Ledger
[PHILOSOPHY]
This ledger tracks the transition of technical theory into local software.
Each implementation is grounded in peer-reviewed research or documented validation scars.
| Research Anchor | Theoretical Logic | Acme Implementation | Git Link | Status |
|---|---|---|---|---|
| FS-Researcher (2602.01566) | Dual-Agent: Context Builder + Report Writer | Foundation: `nibble.py` is the Context Builder . | field_notes/nibble.py | 100% |
| Agentic-R (2601.11888) | Learning to Retrieve: Utility-based ranking. | Memory Bridge: `ArchiveMemory` utility ranking . | nodes/archive_node.py | 85% |
| TTCS (2601.22628) | Test-Time Curriculum: Synthesizer + Solver. | Quality: Mitigates repetition loops via Synthesize-then-Solve patterns . | field_notes/ai_engine_v2.py | 100% |
| Apple CLaRa (2511.18659) | Semantic Compression: 16x-128x density. | Optimization: `SemanticCondenser` compresses raw logs . | field_notes/ai_engine_v2.py | 100% |
| Liger Kernel (2410.10989) | Efficient Triton Kernels for LLM Training. | Efficiency: Maximizing 8B residency on 11GB silicon . | nodes/loader.py | 100% |
| RLM | Context as a Code-Readable String. | Discovery: `peek_related_notes()` logic . | nodes/archive_node.py | 100% |
| TTT-Discover (2601.16175) | Test-Time Discovery: RL-based optimization. | Automation: Planned: RL loops for bug reproduction. | N/A | 0% |
| WideSeek-R1 | Width Scaling: Parallel subagent orchestration. | Orchestration: Planned: Parallel contexts for extraction. | N/A | 0% |
| MiMo-V2-Flash | Multi-Objective RL: Post-execution feedback. | Observability: Planned: Use Grafana feedback to reward agents. | N/A | 0% |
| Dreaming (2603.04257) | Subconscious Compression (Memex). | Consolidation: `dream_cycle.py` moving logs to Wisdom . | dream_cycle.py | 100% |
| Internal Debate (2603.00142) | Moderated Consensus (Byzantine ToM). | Consensus: `delegate_internal_debate` facilitates reasoning . | logic/cognitive_hub.py | 100% |
| AT2QA (2603.01853) | Autonomous Exploration Pivot. | Autonomy: Pivot-query logic pass for thin results. | logic/cognitive_hub.py | Design |
| Agentic FS (2602.20478) | Context as a File System. | Grounding: Validates our "Static Synthesis" architecture. | field_notes/mass_scan.py | 100% |
| 13-Param Reason (2602.04118) | Extreme parameter efficiency. | Sentinel: Optimization logic for 1B "Sentinel" mice nodes. | nodes/lab_node.py | Active |
| REDSearcher (2602.14234) | Long-Horizon Search Scaling. | Discovery: framework for deep technical history discovery. | nodes/archive_node.py | Design |
| VibeThinker-3B (2606.16140) | Verifiable Reasoning / Spectrum-to-Signal. | Sanity: Unified local base model for Phase 10 . | nodes/loader.py | Active |
| MCompassRAG (2606.18508) | Metadata as semantic compass. | Discovery: Planned: Metadata-guided paragraph RAG. | nodes/archive_node.py | Design |
| Banter Decay | Frequency-based idling. | Replaced by Metabolism of Presence . | Replaced by passive wakefulness logic. | DEFEATURED |
| Persona-Locked Dispatch | Rigid isolation. | Rigid isolation prevented collaborative synthesis . | Bleed prevention replaced by shared Identity Bedrock. | DEFEATURED |
| Thinking to Recall (2603.09906) | Uses CoT as a "computational buffer" for factual priming. | Implemented via TTCS logic. | field_notes/ai_engine_v2.py | 100% |
| AutoHarness (2603.03329) | Synthesizing Python guardrails for agent self-verification. | Grounds future Verifier Synthesis. | N/A | Planned |
| Stochastic KV Routing (2604.22782) | Adaptive depth-wise cache sharing (Apple MLR). | Optimization pattern for Liger. | nodes/loader.py | Active |
| ARIS (2605.03042) | Adversarial multi-agent collaboration for research. | Theoretical base for Internal Debate. | logic/cognitive_hub.py | 100% |
| CodeTracer (2604.11641) | Traceable agent states & failure onset localization. | Enhanced Forensic Ledger visibility. | debug/test_forensic_logging.py | 100% |
| PersonaVLM (2604.13074) | Proactive memory extraction & response alignment. | Core logic for Subconscious Dreaming. | dream_cycle.py | 100% |
| TriAttention (2604.04921) | Trigonometric KV compression (10x reduction). | Future optimization for VRAM efficiency. | nodes/loader.py | Active |
| Bonsai-ML | High-fidelity fine-tuning for structured reasoning. | Grounds the Qwen Pivot strategy. | nodes/loader.py | Active |
| BKM-015: Bedrock | Prefix-cache stability through shared identity strings. | Shared IDENTITY_BEDROCK implementation. | nodes/loader.py | Live |
| Stoic Strategist | Filler stripping ("Certainly!", etc.) & strategic focus. | Persona constraint enforcement. | debug/test_persona_bugs.py | Live |
| VRAM Guard | Deep Sleep failover on memory exhaustion. | Attendant VRAM characterization. | debug/test_vram_guard.py | Live |
| Hierarchical Mind | Specialized Architect node for BKM generation. | Semantic map generation. | debug/test_architect_flow.py | Live |
| Nightly Recruiter | Automated resume mapping to active job listings. | Recruiter query matching. | debug/test_recruiter.py | Live |
| Strategic Amygdala | Input sentinels gating voice/typing console modes. | console barge-in sentinels. | debug/mic_toggle_audit.py | Live |
| Iron Moat | Prevent persona bleed / clear visual panels on hello. | Persona Isolation boundaries. | debug/test_persona_logic.py | Live |
| Barge-In interrupts | Cancel long execution steps via voice/hotkey patterns. | ear_poller stream cancellation. | debug/test_barge_in_logic.py | Live |
| Zombie Port Recovery | Monitor WS socket status & restore engine on lockups. | Foyer port watchdog. | debug/test_zombie_recovery.py | Live |