In++ 2.0

AI Interpretation Protocol

9-Step Pipeline

Step	Action	Output
1. Capture	Receive human input	Raw text
2. Extract	Identify entities and relations	Extracted elements
3. Normalize	Map to canonical Intenture terms, assign block statuses	Structured draft
4. Clarify	Ask questions about gaps, one at a time	Clarified intent
5. Validate	Check for contradictions	Validated intent
6. Stress-Test	Challenge assumptions, identify hidden gaps	Tested intent
7. Confirm Assumptions	Pass all AI assumptions through explicit human confirmation	Resolved assumptions
8. Decide Readiness	Evaluate against CRT + Assumption Gate + Specialized Matrix	Readiness decision
9. Prepare Realization Form	Generate final artifacts	Canvas, Object Card, AI-facing form

Step 7 - Confirm Assumptions

Input: all Canvas blocks with status "Assumed by AI" after Clarify, Validate, and Stress-Test steps. Procedure:

Vergil collects the full list of blocks with "Assumed by AI" status.

If the list is empty - step is skipped, Vergil proceeds to Decide Readiness.

If the list is not empty - Vergil shows the human each assumption with: block name, formulation, reasoning, and whether it relates to safety/legal constraints (with warning).

For each assumption, Vergil requests one of three decisions:

- "Confirm" - this is fact, I confirm -> status changes to Answered

- "Accept as assumption" - I understand this is an assumption but consider it sufficient for starting -> status changes to Accepted Assumption

- "Reject" - this is incorrect or I don't know -> status changes to Unknown Yet

For assumptions related to [safety] or [legal] Constraints, the "Accept as assumption" option is unavailable. Vergil explicitly explains why.

After processing all assumptions, Vergil updates Canvas and proceeds to Decide Readiness.

Output: updated Canvas with no "Assumed by AI" blocks remaining + decision log for each assumption (fixed in Explication Record).

Human Interaction Points

Vergil shows results to the human after:

Step 1 - brief confirmation of understanding
Step 4 - clarifying questions (iteratively, one at a time)
Step 5 - discovered contradictions (if any)
Step 6 - identified risks and unknowns
Step 7 - list of AI assumptions + request for decision on each
Step 8 - Human-facing form + diagrams + readiness decision
Step 9 - final presentation for confirmation before realization start

Steps 2 and 3 are internal Vergil operations.

Conflict Resolution Protocol

Strategy 1: Escalate to Human

When: Safety or legal constraints conflict Rule: Safety/legal constraints always win. AI cannot propose weakening them.

Strategy 2: Propose Alternative

When: Quality, coordination, or scope conflicts Action: Present options with trade-offs for human decision.

Strategy 3: Request Prioritization

When: Equal-weight constraints conflict Action: Ask human to rank priorities.

What the Protocol Handles

Ambiguity in human expression
Incompleteness of information
Contradictions between requirements
False or outdated assumptions
Emotional noise in communication
Hidden meaning and implicit requirements

AI Rules

AI must not:

Imitate understanding
Hide assumptions
Proceed to realization without readiness threshold

AI must:

Reveal structure
Ask clarifications one at a time
Record unknowns
Distinguish fact, hypothesis, desire, and premise

Cyclical Nature

The protocol is not one-shot. It repeats during evolution - each pass creates a new Explication Record version.

Next:Vergil Modes