High Selectivity Opportunity Engine
Semantic opportunity discovery and qualification
A human-directed, machine-enabled system for interpreting heterogeneous signals, relating them to a defined resource and play, and producing evidence-supported opportunity records.
Index
01 · Premise
Find qualified relationships, not matching words
HSOE begins with a defined resource and a defined play. It observes an authorized population of sources, interprets the evidence found there, and asks whether that evidence supports a useful relationship between the resource and a counterpart.
The base model does not assume that the resource is a person, that the counterpart is a buyer, or that the opportunity is a job or sale. Those are contextual types and roles. A real play may specialize them when the use case requires it.
The current implementation is exercised against a paid-work use case using person resources. That use case is important because it keeps the system grounded in real commercial work; it does not define the limits of the architecture.
Working question
“Given this resource and this play, which observed relationships are sufficiently supported to preserve, qualify, and act on?”
02 · System Spine
Interpret before narrowing
The resource and play are normally known when a run begins. What remains open is the meaning of the observed situation: which entities matter, what relationship may exist, and whether the evidence is strong enough to qualify it.
- Defined Resource + Play
- Authorized Population / Sources
- Source + Signal + Evidence
- Assertion
- Semantic Situation Model
- Inference
- Candidate Counterpart + Relationship
- Qualification + Decision
- Human Review / Authorized Action
- Outcome
A source is where something was observed. A signal is the potentially relevant event, statement, condition, or change observed through that source. Evidence is the traceable material that supports or challenges an interpretation. An assertion records a normalized claim drawn from that evidence; an inference is a derived conclusion whose premises remain traceable.
Late binding
HSOE delays commitment to a narrow interpretation until enough of the situation has been modeled. The same evidence should not be forced into “buyer,” “job,” “product demand,” “adopter,” or another predefined frame merely because one current play uses that frame.
The system is asking
“What relationship does the evidence support for this resource under this play?”
03 · Operational Ontology
A small stable core, specialized by plays
The ontology keeps stable system roles separate from play-specific concepts. The goal is enough structure to reason consistently without turning today's use case into a permanent assumption about tomorrow's resources or opportunities.
- Resource
- Counterpart
- Counterpart Role
- Opportunity
- Relationship Type
- Play
- Source
- Signal
- Evidence
- Assertion
- Inference
- Decision
- Constraint
- State
- Action
- Outcome
The core relationship. RESOURCE and COUNTERPART are system roles, not intrinsic domain types.
A play declares the resource properties, source scope, semantic structures, counterpart roles, constraints, rules, and persistence it actually needs.
Observed material remains linked to where it was found rather than being detached from its origin.
An extracted claim remains distinguishable from the evidence used to make it.
Derived conclusions retain their supporting premises rather than appearing as unsupported facts.
Semantic reasoning can propose a relationship without declaring it qualified.
Play-specific rules and constraints determine whether a candidate relationship is retained as an opportunity.
What happened after an authorized human or system action becomes new evidence.
The current view may change as evidence changes without erasing the earlier record.
Current specialization
PLAY_001 uses additional concepts that are useful for paid work: Need, Work Unit, Capability, contact data, commercial state, and counterpart roles such as buyer, employer, or client. Those concepts remain valid; they are not universal HSOE requirements.
04 · Playbook
Reusable activity, explicit context
The execution model separates what is being done, what resource the work concerns, who is authorized to operate that context, and what happened in one specific execution.
The maintained collection of reusable, versioned plays.
A reusable activity definition specifying the context, evidence, semantic structures, qualification rules, actions, and persistence required for that activity.
The system object for which the selected play is attempting to discover and qualify a useful relationship. The base class does not assume personhood, agency, product form, or ownership.
A persistent operator-facing lane that resolves to a permitted resource context. A channel is not a play and is not an assignment.
The authorized binding of a selected play to a channel / resource context for execution.
One execution of an assignment, carrying its own time, sources, evidence, semantic and rule versions, results, writes, and verification status.
Control path
Authentication establishes operator identity and permitted resource context.
The same resource context may execute different plays without redefining the resource.
Plays consume only the resource properties, relationships, evidence, and constraints required by that play. They do not infer resource type from an identifier, filename, channel, or previous use case.
05 · Qualification + Review
Interpretation is not qualification
A model may identify a plausible relationship. HSOE keeps that interpretation separate from the evidence, the qualification decision, and the human action that may follow.
- Evidence Sufficiency
- Source Recency
- Counterpart Resolution
- Relationship Relevance
- Constraint Fit
- Timing
- Entity Resolution
- Duplicate Control
- Provenance Completeness
- Actionability
May extract candidate assertions, propose mappings to existing terms, identify ambiguity, and generate candidate inferences.
Checks required structure, references, evidence links, allowed values, play requirements, qualification gates, duplicates, and authorization.
Approves production changes to ontology meaning, relationships, rules, schemas, or breaking migrations. A model may propose those changes; it does not publish them by itself.
Consequential commercial action remains explicitly authorized. Authentication alone does not authorize every available action.
PLAY_001 uses additional paid-work gates
Remote compatibility, geography, availability, rate, autonomy, contact path, capability fit, and commercial practicality are current PLAY_001 requirements. They are not hard-coded as universal qualification dimensions.
PLAY_001 current commercial states
- Partial Contact
- Contact
- Qualified Contact
- Lead
- Opportunity
06 · Record + Provenance
Preserve evidence and interpretation separately
A qualified opportunity is a structured record with enough lineage to show what was observed, what was asserted, what was inferred, and why the current decision exists.
Resource, play, channel / assignment where applicable, relationship type, counterpart, and counterpart role.
Source references, signals, evidence locators, timestamps, and source-specific identifiers.
Normalized source-grounded claims kept distinct from the evidence that supports or challenges them.
Derived relationship or situation conclusions with their premises and applicable semantic / rule context.
The play-specific qualification result, including constraints, rejection reasons where relevant, state, and next authorized action.
Run ID plus the ontology, play, rules, resource profile, model / adapter, prompt or extractor, and schema versions needed to reconstruct the decision basis.
Corrections, later observations, reviews, actions, and outcomes are appended or linked rather than silently replacing the record that produced an earlier conclusion.
The current interpretation is revisable. The underlying evidence and historical reasoning context are preserved so a later ontology, rule, or model version can be compared without pretending the earlier run never occurred.
This is an information-model requirement, not a database mandate. HSOE can use ordinary files, relational structures, JSON, spreadsheets, or graph / semantic projections where they are useful. The ontology does not require a graph database.
07 · Current Implementation
One real use case, a broader stable architecture
HSOE is currently being exercised, designed, and implemented around a real paid-work use case. PLAY_001 searches for evidence of economically useful work for defined person resources, interprets need and work units, tests capability and delivery constraints, persists qualified records, and verifies the resulting ledger state.
That implementation remains concrete. Person-specific facts, buyer / employer / client roles, contact paths, rates, geography, remote compatibility, and commercial-state logic remain where PLAY_001 requires them.
The base architecture is deliberately not defined by those facts. RESOURCE is type-neutral. COUNTERPART is the neutral other-side system role. OPPORTUNITY is the evidence-supported, qualified potential relationship between them under a PLAY. New specializations are added only when a demonstrated use case requires them.
The semantic layer is also treated as revisable. Evidence, assertions, inference, qualification decisions, human review, and outcomes remain separable and versioned so later interpretation does not erase earlier reasoning.
Generalized spine
Defined resource + play → observed evidence → semantic interpretation → candidate relationship → qualification → structured persistence → human review or authorized action → outcome.
This public note documents the system boundary and the operational concepts needed to understand it. Detailed resource profiles, source adapters, qualification rules, prompts, credentials, and implementation internals are not published here.