Mechanism
The unified semantic discovery substrate is architecturally independent of any specific inference engine, model architecture, training methodology, or computational paradigm. The three-in-one traversal step requires, at each anchor, an inference engine capable of scoring or selecting among a set of structured candidate transitions given a structured semantic state. That is the entire requirement. Any computational mechanism that accepts the structured inputs and produces a preference ordering or a selection satisfies it. The discovery object never depends on which mechanism an anchor uses; it depends only on receiving a preference ordering over its candidate transition set.
Because the requirement is stated as a preference ordering over structured candidates, it is satisfied by a wide range of mechanisms: large language models, small language models, embedding-similarity scorers, rule-based matchers, probabilistic graphical models, Bayesian inference engines, decision trees, symbolic reasoners, neuro-symbolic hybrid systems, and human evaluators. The substrate does not privilege any one of these. It treats the inference engine as a pluggable component that proposes; the governance of the traversal lives elsewhere.
Why Model-Agnosticism Holds
Model-agnosticism is not a convenience layer added on top of the architecture. It is a direct consequence of the structural separation between inference and execution. In the three-in-one traversal step, the inference engine at an anchor proposes transitions and the execution substrate decides whether to commit them. The inference engine operates as a proposal generator: it evaluates candidates and produces a preference ordering, but it has no authority to advance the traversal. Authority to commit a transition resides exclusively in the execution substrate, which evaluates each proposed transition for admissibility against the deterministic governance criteria carried in the discovery object's policy reference field, the anchor's governance configuration, and the traversal's accumulated lineage.
This is why the inference engine need not be trusted. The execution substrate provides the governance guarantee; the inference engine provides only the proposal. A highly capable but occasionally unreliable large language model can serve as the inference engine at one anchor while a simple embedding-similarity scorer serves at another and a rule-based matcher serves at a third, and the governance integrity of the traversal is unaffected by any of these choices, because the admissibility evaluation at every step is performed by the execution substrate independently of the inference engine's internal mechanics. The model proposes; the substrate decides.
Heterogeneous Inference Across the Index
Because each anchor's inference engine is local and untrusted, different anchors may employ different inference engines chosen to suit the characteristics of their semantic neighborhoods. An anchor governing a container of scientific literature may employ a domain-specific embedding model trained on scientific text. An anchor governing a container of legal documents may employ a rule-based matcher that evaluates candidate transitions against legal ontology constraints. An anchor governing a container of multimedia content may employ a multimodal model capable of evaluating visual, auditory, and textual content.
A single traversal passes seamlessly through anchors employing different inference engines. The continuity across anchor boundaries is provided by the discovery object's semantic state, which is maintained by the execution substrate and not by the inference engine. At each anchor the inference engine receives the discovery object's current state and produces a proposal; the execution substrate evaluates the proposal and advances the traversal. The inference engine's architecture, training, and capability are local to the anchor and do not affect the traversal's governance integrity, so heterogeneity across the index carries no cost to the consistency of the traversal.
Future-Proofing Against Engine Evolution
The same separation that permits heterogeneity also permits evolution. As new model architectures are developed, as language models are replaced by more capable successors, as new reasoning paradigms emerge, and as specialized models are trained for new domains, the discovery substrate can incorporate them without architectural modification. An anchor upgrades its inference engine by replacing the current engine with a new one and verifying that the new engine produces valid proposals against the anchor's test suite.
No change to the traversal protocol, the discovery object schema, the admissibility evaluation, or the governance infrastructure is required for such an upgrade. The discovery substrate is a traversal and governance framework, and the inference engine is a pluggable component within that framework. The boundary between the two is where engine evolution is absorbed, so the substrate does not inherit the lifecycle of any model it happens to use at a given moment.
Beyond Search
The model-agnostic architecture also ensures that the discovery substrate is not limited to search. Any computational problem expressible as a query over a semantic graph can be expressed as a traversal through the adaptive index with an appropriate inference engine at each anchor. This includes planning, scheduling, optimization, knowledge integration, hypothesis testing, anomaly detection, and creative generation.
In this view the discovery substrate is a general semantic computation framework in which search is one operating mode and reasoning is another, and any form of structured semantic traversal is supported. The choice of inference engine at each anchor determines what kind of computation that anchor contributes, while the traversal and governance framework around it remains constant.
Prior-Art Distinction
Conventional retrieval and discovery systems tend to embed a specific model into the architecture, binding the system to that model's lifecycle, or to expose a low-level inference interface that callers must adapt to per model, pushing model heterogeneity into application code. In both cases the inference engine carries authority: its output is the result, or is acted upon directly. The disclosed mechanism withholds authority from the inference engine entirely. Because proposal authority is separated from commitment authority, and because commitment authority resides in a deterministic execution substrate that governs every transition, the inference engine becomes substitutable without renegotiating trust. Substitution, heterogeneity across anchors, and evolution over time are absorbed at the boundary between the engine and the substrate rather than treated as architectural ruptures requiring redesign.
Disclosure Scope
The model-agnostic applicability of the unified semantic discovery substrate, comprising the requirement that each anchor's inference engine produce only a preference ordering or selection over a structured candidate set given a structured semantic state, the derivation of model-agnosticism from the structural separation of proposal authority in the inference engine from commitment authority in the execution substrate, the heterogeneous deployment of different inference engines at different anchors with continuity maintained by the execution-substrate-held semantic state of the discovery object, the upgrade of an anchor's engine by verification against the anchor's test suite without change to the traversal protocol or governance infrastructure, and the extension of the substrate beyond search to any computation expressible as a governed semantic traversal, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart) at Section 10.11, in dependence on the inference and execution separation disclosed at Section 10.5. This article describes that disclosed mechanism. The scope extends to inference engine classes not enumerated whose output is a preference ordering over the candidate set, provided commitment authority remains with the execution substrate and the traversal's governance integrity is preserved.
