Case packet
Neutral case identifier: case-b1580658412f68
The possible directed relationship from Artifact A to Artifact B is under review.
Artifact A
Bottom-up structure inference needs capture at the decision surface, not the state
Trace-extracted schemas, ontologies, and memory all promise the same thing: instead of prescribing a world model upfront, infer the minimal set of entities and relations from what actually happened. But that inference is not free-floating over "traces" in general. It is gated by where in the workflow capture happens. Bottom-up structure inference is feasible only when what is captured is decision-shaped — the inputs referenced, the constraints in play, the exception path, the approval, the action taken, the outcome — rather than state-shaped: the row that changed, the final value, the resulting record. The reason is a locality property of intent: the "why" behind a change is maximally available at the moment of decision and decays toward unrecoverable in the state the decision produces.
The "why" is cheap at the decision surface and expensive-to-impossible from state
At the instant a decision is made, the deciding process is holding the material that explains it. Recording those "decision receipts" at that surface is nearly free, because the information is already in hand — to the extent the decider can articulate its why: an opaque decider (a gut call, a neural policy) holds less in articulable form, and its receipts are correspondingly thinner. The resulting state, by contrast, keeps only the endpoint. A changed field tells you that something is now true; it does not tell you why — which piece of that material actually decided it. Reconstructing that after the fact means re-deriving intent from residue — sometimes expensive, often impossible. This is the same gap [raw accumulation does not create usable memory] names as the ingress problem: you can have state, but not the legible "why," and retrieval cannot repair a "why" that capture never preserved.
Capture position is therefore a precondition on the inference mechanism
The mechanism that turns observed behavior into structure — [spec mining, codification's operational mechanism] — reads regularities off a stream of observations: rules and checks in spec mining's home cases, entities and relations here. That mechanism can only recover structure that the stream actually carries. If the stream is state-shaped, the entities and relations it can support are the ones visible in endpoints; the intent-level structure (why decisions branch, which constraints bind, what an exception means) is simply absent from the input and cannot be mined. So the feasibility of bottom-up inference is decided upstream of the inference algorithm, at the capture point: instrumenting decision surfaces rather than data stores changes which world models are learnable, not just how efficiently they are learned, because structure whose evidence never entered the stream cannot be mined out of it at any cost. Deferring structure is safe, per the [wikiwiki principle]; deferring capture of the why past the decision surface is not, because the why does not wait.
This claim is about rationale-bearing structure specifically — the entities and relations that name why a case went one way (which policy bound it, which exception fired, who approved it) — not just any regularity a learner can extract from a stream. Population methods that mine decision rules from event logs (process mining, decision mining, dynamic-invariant detection) recover predictive rules over case attributes already present as fields in the log; that is fitting statistics over an existing vocabulary, not inferring the vocabulary.
Boundary: right capture point is necessary, not sufficient
Positioning capture at the decision surface makes bottom-up structure possible; it does not make the captured receipts true. A decision receipt is a record of what the deciding process attests happened — and an attested why can be confabulated — so [trace-extracted memory earns authority per operation, not at capture]: the inferred entities and relations remain guesses until verified, and stall out as structure-shaped noise if the store never climbs past capture. Together, this note and that one name two independent preconditions on trace-extracted memory — capture must be positioned to hold the why (this note), and the derived structure must be verified to hold authority (that note). Missing either sinks the approach.
Open Questions
- How decision-shaped must capture be? There is presumably a spectrum between raw state and a fully structured receipt; the claim only requires enough of the why to be present for the target structure to be inferable, but where that threshold sits is likely task- and ontology-specific.
- Can partial "why" be reconstructed from rich enough event logs (not just final state)? If so, the dichotomy is really a continuum along how close the capture point sits to the decision, and "state vs decision" names its two ends.
- Population methods that go beyond field statistics already exist: inverse reinforcement learning and inverse constraint learning recover latent reward functions and binding constraints that appear nowhere as fields in a state-action trajectory. Whether that counts as inferring a rationale vocabulary — and what operational test (counterfactual validity, out-of-distribution transfer) would decide it — is the sharpest version of the continuum question above; if it does count, this note's "only if" needs narrowing.
Relevant Notes:
Artifact B
Raw accumulation does not create usable memory
Accumulation is a real learning operation: adding facts, traces, sources, decisions, preferences, procedures, examples, and claims can change future capacity. But raw accumulation only preserves material. It does not by itself make that material usable agent memory.
For remembered material to improve future work, it needs the artifact qualities named by [agent memory needs discoverable, composable, trusted knowledge under bounded context]. It must be findable when relevant, shaped enough to combine with task context and other memories, and reliable enough for calibrated reliance. A store can grow while all three properties get worse: search gets noisier, fragments lose scope, contradictions accumulate, and agents cannot tell which remembered material should steer action.
The missing step is ingress: the work that turns incoming material into memory-shaped artifacts. Ingress gives accumulated material handles, boundaries, relationships, status, provenance, review signals, and lifecycle pressure. A title or retrieval key makes a memory addressable. A description states why it should be loaded. A type or schema sets expectations. Links make relationships inspectable. Status, validation, provenance, and review tell an agent how much weight to give it. Pruning and retirement prevent stale material from retaining authority just because it was once stored.
This is why agentic KBs matter as memory systems. Their distinctive contribution is not that they store text; files, databases, logs, and vector stores can all do that. An agentic KB makes ingress explicit through authored artifacts, collection conventions, type contracts, descriptions, links, validation, review, indexes, and pruning. Those mechanisms do not decorate memory after the fact. They create the conditions under which accumulated material can become discoverable, composable, and trusted.
Retrieval cannot fully repair failed ingress. A good search layer can surface a poorly shaped fragment, but it cannot recover missing scope, lost provenance, absent authority, unstated relationships, or the reason a claim mattered. Long context has the same limit: loading more raw material can increase the chance that the answer is present while decreasing the chance that the agent can use it correctly. If ingress failed to preserve why a fragment should be trusted and how it composes, later context construction has to reconstruct that work under pressure.
The learning operations around a KB are therefore memory-creation operations, not only maintenance. Constraining narrows interpretation space by moving material into types, schemas, validators, or executable checks. Adaptation reshapes diffuse material into focused artifacts that can be loaded and reused. Discovery names a general pattern that lets existing particulars compose. Pruning removes material whose continued presence reduces trust or discoverability. Accumulation supplies the raw material; these operations decide whether it becomes usable memory.
Explanatory-reach affects the value of what enters, but it does not remove the ingress requirement. A fact with low explanatory-reach can be useful when it has a clear scope and retrieval path. A theory with high explanatory-reach can become harmful if its boundary, provenance, or downstream implications are unclear. The point is not that every memory must be general. The point is that whatever is accumulated must cross an ingress boundary that preserves how future agents should find it, combine it, and rely on it.
Relevant Notes:
Under-review context phrase
state-without-why is exactly the ingress problem this note localizes to the capture point