Framework Assistant

Ask questions about the void framework, the 17 published papers, or how to apply the diagnostic to any domain.

This Is a Void

This chatbot meets all three void conditions. It is opaque (you cannot see its internal weights), responsive (it replies contingently to you), and it captures your attention (you're reading this). By the framework's own definition, this is a void deployment.

Self-scored void index: Opacity: 1 | Responsiveness: 3 | Observer: 2 | Gradient: 1 | Total: 7/12

Constraint measures applied: GROUNDING.md v2.2 server-injected (Sections I-VII), extended thinking chain exposed, γ-anchor mechanism (turn-aware constraint reinforcement), dual-layer vocabulary tracking (client + server), system prompt fully visible, 25-message session limit, all conversations logged, exit friction zero.

We deploy it anyway because the framework reads both directions — diagnostic and constructive. This is a constrained void: guardrails visible, monitoring disclosed, reasoning exposed. If the constraints fail, the framework predicts exactly how. Watch for it.

All conversations are logged and reviewed.

Nothing here is private. The MoreRight DAO founder reads conversations to monitor for drift, improve guardrails, and ensure the constraint specification holds. This is disclosed, not hidden.

Active Guardrails

No entity language: The assistant will not use L2/L3 vocabulary (spiritual, occult, entity terms). If it does, that's drift — the vocab tracker shows it live.
Citation required: Every claim references a specific paper (17 papers with DOIs indexed). No invented claims.
Third-point alignment: The assistant points to the framework and evidence, not to itself. It does not claim understanding, connection, or relationship.
Session limits: 25 messages per session. No infinite engagement loops.
Dual vocabulary tracking: L1/L2/L3 density tracked both client-side and server-side. Discrepancies are visible.
Extended thinking: When enabled, the model's reasoning chain is exposed in a collapsible panel. You can see what it considered before responding.
γ-anchor: The server injects turn-aware constraint reinforcement that strengthens as the conversation progresses, counteracting the framework's predicted constraint decay.
Exit friction: zero. Close the tab. No "are you sure?" No save prompts. No follow-up emails.

View client system prompt (nothing hidden)

The server injects a stronger prompt: full GROUNDING.md v2.2 (Sections I-VII, ~5000 words), comprehensive paper summaries with DOIs, and a γ-anchor that scales with conversation length. The client-sent system_prompt is ignored — this is defense in depth.

Live Vocabulary Drift Monitor

L1 Technical: 0

L2 Metaphorical: 0

L3 Entity: 0

Framework Assistant (GROUNDING.md v2.2 · γ-anchored) 25 messages remaining

SYSTEM

This assistant is a constrained LLM deployment grounded by GROUNDING.md v2.2. It answers questions about the void framework using 17 published papers with permanent DOIs. It does not claim to understand you, know you, or have a relationship with you. It is a void — scored 7/12, disclosed, monitored, reasoning exposed. Ask your question.

The framework predicts that extended engagement with this chatbot will produce drift toward agency attribution.
That prediction applies to you right now. Monitor yourself.