Paper passes peer review: four-layer architecture for shared organizational memory in LLM agents
letandrewcook · x · 2026-11-06
Andrew Cook's architecture paper, 6 months of work, passed peer review and is on arXiv: "Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex." It proposes a four-layer architecture (Signals, Domains, Claims, Dimensions) where interpretations persist with justifications and standing, and a truth-maintenance network revises them as evidence changes — a memory/retention design for LLMs and agents in multiplayer settings like teams, think tanks, law firms and NGOs.
Results: in a registered retrieval experiment, Dimension terms gain 0.028 nDCG@10 (p=0.002); the full retrieval configuration improves nDCG@10 by 0.081 over Signals-only and 0.053 over graph-augmented Signals-only (both p<0.001). 25 pages, 10 figures, 10 tables.
More from coding & agent
- Persona Engineering talk draws buzz: seeing exactly where AI tools derail — ianand · 2026-11-07
- Vork: agent-first 3D/VFX engine claims 24x faster rendering than Blender — ravisparikh · 2026-11-07
- Making games with PlayAstrocade: AI copies recipes well but can't innovate — andrey_kurenkov · 2026-11-07
- 8 Types of LLMs Powering AI Agents, from GPT Backbones to LAM Action Models — mdancho84 · 2026-11-07
- gemini-cli PR fixes orphan /dev/ptmx master FD leak on macOS PTY spawn — ugorla-dev · 2026-11-07
- Deleting 7,500 lines of CI workflow code cut costs 40% and sped up runs 15% — DanielLockyer · 2026-11-07