Skip to main content

​
Step 1. Let the Decision Black Box listen

DBB runs on the client layer. It watches chats, docs, and notes, extracts decision pivots, and freezes reasoning before the outcome is known. Backend escalation is optional and used only when external knowledge is required.
  • Timestamped reasoning (context, motivators, assumptions)
  • Confidence level and predictions
  • Alternatives you considered
  • Emotional state and open questions
Entries become immutable once locked. Outcomes are append-only so original reasoning remains intact.

​
Step 2. Read the Decision Surface

Decision Surface is the client dashboard. It shows open loops, drift, pending reviews, and signals derived from DBB records. It does not read raw chat.

Open Loops

Unresolved decisions, conflicted assumptions, and unanswered questions.

Drift

Evidence that a belief changed after the decision was captured.

​
Step 3. Resolve a card

Click a card, review the DBB draft (decision, context, evidence, questions), and choose:
  • Confirm decision to make it immutable.
  • Edit if wording is off (original reasoning stays preserved).
  • Reject when the draft is noise, not a real decision.

​
Step 4. Track outcomes and learn

When the outcome is known, append the result and compare prediction vs. reality. The client layer surfaces calibration signals, timing patterns, and principle adherence.

​
Where the backend starts

When the client needs external knowledge or escalation, the request goes to the backend layer (KCG + DoD + gateways):

KCG & gateways

Backend graph logic, gateway consensus, and Arweave storage.

Technical notes

End-to-end pipeline from client capture to verified knowledge.