intelligance
An intelligence layer above agents. Not an agent and not a framework — an intermediate layer that takes a task in natural language and returns a verifiable artifact: a typed graph of meaning with computed confidence.
Intelligence is the ability to ask the next question.
The knowledge is already inside the model; what the model lacks is the discipline of interrogating itself. So here the model does not "think" — it answers. Intelligence is moved outside: into an explicit interrogation procedure and into the graph where answers accumulate. Context lives in that structure, not in the model window — task size is bounded by a question budget, not by context size.
A verifiable artifact, not an answer to trust.
| Field | Contents |
|---|---|
| graph | A typed object model of the reasoning — flat, diffable, mechanically validated. |
| conclusions | Findings with the chain that produced them and a computed confidence. |
| validation | A mechanical integrity check — ok plus the list of issues. |
| trace | Every question asked and every answer received, in order. |
| stats | Graph size, questions spent, stop reason, token cost. |
The layer does not manage execution: it hands the artifact over, and other agents execute it. Determinism is the point — the same inputs produce the same reasoning path, and an error localizes to a specific node.
The answer is nowhere in the prompt.
The reference test asks what happens around a wall-mounted air conditioner. The conclusion ("mold is likely on the surface under the unit") is hardcoded nowhere: the mock provider only answers local questions and does not know the task. The engine assembles the conclusion itself — with a computed confidence and a full trace of how it got there.
TECHNICAL DETAIL — THE ONTOLOGY, THE OPERATORS AND THE UPDATE MECHANISM — IS WITHHELD PENDING IP REVIEW.