Gap Computing Ledger

Gap Ledger Computation

Compute the distance between desired state and current state for each hypothesis and proposal in the register. Produce scored, ranked gap records that drive prioritization.

Process

Step 1: Extract Desired and Current States [S]

For each hypothesis section (Problem, Segment, Unit Economics, Value Proposition) and each proposal section (Growth Architecture, Solution Design, GTM Plan):

1. Read the Desired State block. List every condition under "SUPPORTED means."
2. Read the Current State block. For each condition, record its fulfillment status: MET, PARTIAL, MISSING, or CONTRADICTED.
3. If a section has no Desired State block, mark the entire section as having a gap of type EVIDENCE_STRENGTH with note "no desired state defined."

Gate: states_extracted: bool -- every register section has a desired/current pair or is flagged as missing.

• Pass: Step 2.
• Fail: flag missing sections and proceed with available data.

Step 2: Score Each Gap [R]

For each condition that is not MET, compute a gap record using these scales:

Confidence Gap (0-3):

Evidence Weakness (0-3):

Pain Uncertainty (0-3):

Time Penalty (0-2):

Blast Radius Weight (1-3):

Final Priority Score:

Final Priority = (Confidence Gap + Evidence Weakness + Pain Uncertainty + Time Penalty) x Blast Radius Weight

Gate: all_gaps_scored: bool -- every non-MET condition has all five component scores and a computed final priority.

• Pass: Step 3.
• Fail: identify which gap is missing a score component. Re-evaluate that component.

Step 3: Assign Gap Dimension [R]

Classify each gap into exactly one dimension:

Gate: all_gaps_dimensioned: bool -- every gap has exactly one dimension.

• Pass: Step 4.
• Fail: re-classify ambiguous gaps. If a gap spans two dimensions, split it into two records.

Step 4: Rank and Select Top 3 [S]

1. Sort all gap records by Final Priority Score descending.
2. Mark the top 3 as the active gap set.
3. If more than 3 gaps exist, mark remaining as deferred (Focus Rule: max 3 active).

Gate: ranked_and_capped: bool -- gaps are strictly sorted, top 3 identified, no more than 3 active.

• Pass: Step 5.
• Fail: re-sort. If ties exist, break by blast radius weight (higher wins), then by confidence gap (higher wins).

Step 5: Generate Recommended Action Per Gap [K-grounded]

Grounded in: gap dimension, evidence profile, current confidence state, deadline status.

For each active gap, determine:

• Action Type: RESEARCH, INTERVIEW, PRESELL, EXPERIMENT, MODEL_REVISION, ESCALATION, HALT, ARCHITECTURE_CHANGE, or SOLUTION_REDESIGN.
• Description: what exactly should be done.
• Expected Output: what evidence or state change should result.
• Evidence Target: which epistemic tier the action should produce (T1 or T2; never recommend T3 as a target).

Selection heuristic:

Fabrication boundary: Do not invent specific market data, customer quotes, or quantitative evidence. Actions must describe what to gather, not assert what will be found.

Gate: actions_specified: bool -- every active gap has an action with all four fields populated. No action targets T3 as evidence output.

• Pass: output.
• Fail: revise action to target T1 or T2 evidence.

Output

Produce:

1. List of scored GapRecords (all gaps, not just top 3).
2. Top-3 priority list with recommended actions.
3. Deferred gaps list (if any, with reason for deferral).
4. Any blockers detected (sections with no desired state, contradictions).

Edge Cases

Boundaries

In scope: Gap computation, scoring, ranking, action recommendation, blocker detection.

Out of scope: Running destruction protocol (gap-running-destruction), enforcing decision rules or setting readiness gates (gap-enforcing-decisions), writing to the register (agent handles I/O), conducting research or interviews (strategist skills).