Opportunity Qualification Model

Assumptions

Everything the model believes, stated before the math. If you do not buy one of these, do not buy the output.

1. Opportunities start from verified record deltas only - an unsourced delta is rejected before any evaluation.
2. Mandate gates are deterministic; missing evidence is UNRESOLVED, never an inferred pass.
3. Novelty is a deterministic key check against prior candidates, not a judgment.
4. QUALIFIED requires a fully matched mandate, no unresolved gates, and an enumerated downside; anything less is WATCH or INSUFFICIENT_RECORD.
5. The model never generates ideas, never predicts unsupported upside, never chooses investments - QUALIFIED is eligibility for diligence, not a recommendation.

Derivation

A delta enters only with a source reference; without one the verdict is REJECTED and nothing else is computed. A verified delta against a missing subject record is INSUFFICIENT_RECORD - never an inferred opportunity.

The subject is then tested against every mandate's hard gates - minimum capacity, allowed states, allowed types - each PASS, FAIL, or UNRESOLVED. The verdict follows a fixed rule: any gate FAIL is REJECTED; a fully passed mandate with no unresolved gates and an enumerated downside is QUALIFIED; unresolved gates or an unenumerable downside is INSUFFICIENT_RECORD; the remainder is WATCH.

Downside is read from the record. A withdrawn or cancelled subject shows the downside realized. Otherwise shared-node exposure enumerates the channels, with effects propagation to run before acting. Missing node linkage means the downside cannot be enumerated - and independence must not be asserted.

Identification strategy

Why the estimate supports the claimed meaning - and where the model refuses to claim more than the data carries.

• Verified-delta entry keeps origination honest: the model qualifies what the record shows, never an idea it invented.
• The verdict rule identifies exactly what separates QUALIFIED from WATCH - a fully matched mandate, zero unresolved gates, and a named downside - each printed per candidate.
• Novelty is a deterministic key against prior candidates, so 'new' is a checkable fact rather than an analyst's impression.
• Refusing to assert independence without node linkage is the downside-identification rule: unknown channels stay UNKNOWN.

Inputs and outputs

Calibration and gate evidence

Validation design

• Gate-verified discipline: verdicts come from the closed set, unresolved gates never pass, unsourced deltas are rejected before evaluation, a verified delta on a missing subject is INSUFFICIENT_RECORD, and identical inputs evaluate identically.

Evidence on record

• Smoke gates at seal (6 m21-specific): verdict from the closed set; gates deterministic, unresolved never pass; unsourced delta rejected before evaluation; verified delta on a missing subject is INSUFFICIENT_RECORD; determinism (identical input hash on re-run); audit row written for every run.

Gates at seal

• smoke: 6 m21 gates incl. closed-verdict-set, unsourced-rejected, missing-subject-INSUFFICIENT_RECORD, determinism + audit row (suite counts at seal per the build agent's seal report)
• sealed at 456d820 on atlas-nl-shock-redesign

Failure modes

The ways this model declines, degrades or goes UNKNOWN - and what it prints when it does.

• Delta without a source reference: REJECTED before evaluation, with the reason printed.
• Subject record missing: INSUFFICIENT_RECORD - the required diligence starts with establishing the record.
• Node linkage missing: downside channels cannot be enumerated, the verdict cannot be QUALIFIED, and the diligence list says what to establish.

Sensitivity

• Gate definitions are the mandate's dials: moving a capacity floor or a state list changes which subjects qualify, and each gate's result prints so the effect of any change is traceable.
• QUALIFIED sits deliberately behind three conditions; relaxing any one moves candidates between WATCH and QUALIFIED, which is why the conditions print rather than tune silently.