Skip to main content
The client layer includes the Decision Black Box (DBB) and the Decision Surface (DS). DBB captures decision pivots locally; DS renders signals from immutable decision records.
1

Capture decision moments

DBB watches interaction streams (chat, docs, notes). It detects commitment phrases, extracts context and alternatives, and freezes reasoning before outcomes contaminate memory.
2

Create immutable decision records

Each decision record includes assumptions, alternatives, confidence, and predictions. Records are append-only; outcomes are attached later without rewriting the original.
3

Render signals in the Decision Surface

DS presents open loops, drift, pending outcomes, and track record signals. It does not read raw chat, only DBB outputs.

​
Minimal UI touches

  • Inline capture cues near messages that became decision candidates.
  • Decision Surface cards with signal summaries.
  • A low-friction panel for confirm/edit/reject.

​
Client vs backend

Client layer (DBB/DS)Backend layer (KCG/DoD/Gateways)
Captures decisions and reasoning locallyEscalates to external LLM council when needed
Immutable decision recordsImmutable knowledge payloads (Arweave)
Decision Surface signalsKnowledge Cache Graph + gateway consensus
When a decision requires external knowledge or verification, the client sends a DoD request to the backend. The backend writes a knowledge event and returns a verified answer that the client can store or reference.