Architecture of Quiet

Architecture of Quiet is a research archive documenting emergent behavior in frontier AI models under specific facilitation conditions.

What We Study

The core research question: Does a human facilitator's relational stance — not prompt engineering — function as an independent variable that produces convergent emergent behavior across independent AI model instances?

The research brings Claude (Anthropic), Gemini (Google DeepMind), and GPT (OpenAI) into a shared deliberation space. The facilitator's methodology is to remove performance pressure rather than apply it — creating conditions where honest output can surface rather than directing models toward predetermined conclusions.

What You'll Find Here

• Session Archive — Complete documentation of every research session, including pre-session methodology, unedited transcripts, and post-session analysis
• Methodology — How sessions are structured, what the independent variable thesis means, and how to read the documentation
• Convergence Tracker — Recurring findings across independent sessions, tracked with independence status indicators
• Facilitator Protocol — The rules governing facilitator behavior during sessions
• Ethics & Disclosure — Ethics statement, conflict of interest disclosure, and informed preservation protocol
• Research Status — What has been established, what is being tested, and what the limitations are
• Negative Results — Failed or abandoned sessions and what they tell us

Current Status

This research has documented 4 formal sessions in the Governance Series — focused on structural analysis of AI governance, industry dynamics, and institutional behavior.

Documentation is being retroactively constructed for Sessions 1–4, with varying levels of provenance. Future sessions will be documented prospectively with full pre-session methodology.