Learning theory

Type: kb/types/tag-readme.md

How systems learn, verify, and improve. These notes define learning mechanisms, verification gradients, and memory architecture that KB design draws on but that aren't KB-specific — they apply to any system that adapts through durable artifacts.

The area is organized around deploy-time learning as the unifying framework. Accumulation — adding knowledge to the store — is the most basic learning operation, with explanatory-reach as its key property: facts sit at the low end, theories at the high end. Accumulated knowledge is transformed by constraining and by working use-shaped artifacts out from it (theory and methodology form a two-layer execution system); the conjecture phase of the discovery lifecycle posits the high-explanatory-reach theories that are accumulation's most valuable items, and recognition is the expensive step in getting there.

Major child areas

These child tags route major parts of the area:

• deploy-time-learning — the framework itself: adaptation through durable inspectable artifacts, learning fundamentals, and feedback quality
• constraining — narrowing the interpretation space, from conventions to deterministic code; codification, relaxing, and the decision heuristics
• discovery — positing a general concept and recognizing particulars as its instances; explanatory-reach as what it produces
• artifact-analysis — the four-field vocabulary (substrate, form, lineage, authority) for retained behavior-shaping artifacts
• agent-memory — memory architecture: spaces, contamination, policy learnability, and the crosscutting decomposition
• llm-reliability — oracle theory, error correction, and the deviation taxonomy; the area applies verification concepts to LLM output deviations

Start here

• Retained system-definition artifacts enable persistent deployment-time adaptation — the unifying framework: persistent cross-session adaptation through retained behavior-shaping artifacts, without weight updates
• learning is not only about generality — accumulation with explanatory-reach as its key property; Simon's definition grounds the decomposition
• agentic systems interpret underspecified instructions — the underspecification foundation: spec-to-program projection and the constrain/relax cycle
• the verifiability gradient — the ladder deploy-time artifacts sit on
• both a narrowed and a use-shaped artifact trade generality for reliability, speed, and cost — how the two transforming mechanisms relate
• recognition, not linking, is the hard problem in knowledge systems — where connecting knowledge actually costs, and how naming amortizes it

Related Tags

• tags — the hub; applies learning theory to KB architecture and evaluation
• document-system — the type ladder (text→note→structured-claim) instantiates the constraining gradient for documents
• context-engineering — where in-context learning meets the system layer that selects and organizes knowledge

Other tagged notes

• A better-factory claim compares operative states under an antecedent assessment relation - The improvement claim's relata are predecessor and operative-successor states and its relation is declared before the development it judges; evaluator location is a separate declaration from the learner boundary
• A checked outcome licenses retaining an episode, not abstracting its explanation - One result-only check can warrant retaining an episode as evidence, but abstracting its explanation also needs evidence about a faithful producing process and an explicit scope boundary
• A claim's warrant does not determine its fit in a working theory - Independent warrant and fit in a working theory answer different questions: a warranted claim may fit poorly, while apparent fit may be produced by an unwarranted or already-assumed claim
• A compact, refreshable whole-picture narrative can replace infeasible fragment reconciliation - Holistic rewrite shifts reconciliation from each consumer to the author, but only when the whole-picture narrative can fit within effective context and be refreshed before the narrative goes stale
• A failure explanation becomes search control only when it changes a later branch decision - An explanation of a failed branch becomes operative search control only when its retention changes a later choice about scope, priority, probing, continuation, or abandonment
• A fixed-model house must retain missing procedures for theory use - With models pinned, newly acquired theory-use procedures must persist outside their parameters; existing general machinery may already supply them, while code can make specified steps cheaper and more reliable
• A goal-holding interpreter fails soft, and its workarounds tax a bounded budget - A procedure compiles its goal away, so a blocked step fails loud and hard; an interpreter holds the goal and re-routes, so failures are absorbed as a per-encounter tax on bounded capacity — silent, accumulating, and softly saturating
• A hand-crafted bootstrap fits the Bitter Lesson only if learning can outgrow it - A hand-crafted starting state fits the Bitter Lesson only if scalable learning displaces the task- and family-specific production knowledge it supplies as claimed reach widens
• A theory's prototype standing is its revision cost: external binding plus lost investment - A theory's prototype standing is its expected revision cost — external binding plus the investment a revision discards — so natural-language versus symbolic form determines neither component and acceptance status is a separate axis
• Abstract an experience into a lesson only when you can state where the lesson stops - Abstract an episode into a lesson only when you can state its boundary, else preserve the instance; an over-generalized lesson is one that drops the condition clause
• Activate Behavior-Changing Memory Before The Mistake - Behavior-changing memory must activate before relevant actions rather than waiting for explicit retrospective search
• Ad hoc explanation can be rational when error is cheap and local - Explains why a disposable local guess can rationally select the next probe when error is cheap and contained, while retained explanations need reach checks
• Ad hoc prompts extend the system without schema changes - Any system with an LLM agent layer can absorb new requirements through natural language prompts without changing the deterministic base
• Adaptation signals choose pressure; artifact analysis chooses the retained surface - Maps agentic-adaptation signals onto artifact-analysis axes so KB learning records which retained surface changes, what authority it gains, and how to review it
• Agent memory is a crosscutting concern, not a separable niche - Memory decomposes into storage (solved), retrieval/activation (context engineering), and learning (learning theory) — treating it as a standalone category hides that the hard problems are at the intersections
• Agent memory needs discoverable, loadable, composable, trusted knowledge under bounded context - Distinguishes four use-time requirements for remembered knowledge—discoverability, loadability, composability, and calibrated trust—from system-level activation.
• An accepted edit verifies the change, not the rule - Human acceptance of an edit is a strong oracle for 'this change was wanted here' but a weak oracle for 'this generalizes' — mining rules from accepted edits inherits instance-level verification while the generalization step stays oracle-poor
• An experiment identifies only the contrast it actually runs - Why missing comparisons, bundle-to-component attribution, and adjacent unrun treatments all overstate causal conclusions beyond an experiment's observed contrast
• An open-domain theory builder becomes a software house when new domains require production-machinery changes - A persistent automated theory builder for external users becomes a software house when genuinely new domains require it to revise the software that performs theory production rather than only the theories produced
• Automated synthesis is missing good oracles - Generating synthesis candidates (cross-note connections, novel combinations) is easy — LLMs do it readily. The hard part is evaluating whether a candidate is genuine insight or noise.
• Axes of artifact analysis - Artifact analysis records retained behavior-shaping artifacts by storage substrate, representational form, lineage, and behavioral authority so review evidence, invalidation, and rollback follow how artifacts actually act
• Behavioral authority - Definition - behavioral authority records who consumes a retained artifact, through which channel, and with what force
• Bottom-up structure inference needs capture at the decision surface, not the state - Bottom-up inference of entities and relations from traces needs decision-shaped capture at the decision surface: the 'why' is cheap to record there and hard-to-impossible to recover from state later
• Brainstorming: how explanatory-reach informs KB design - Deutsch's reach, registered here as explanatory-reach, applied to KB notes — a maintenance risk signal, not a retrieval signal, because high-explanatory-reach revisions break downstream reasoning silently
• Brainstorming: maintainability oracles for agentic development - Explores candidate signals, calibration experiments, authority levels, and workflow placements for evaluating maintainability in agent-generated code
• Changing requirements conflate genuine change with disambiguation failure - Separates world change from late discovery that downstream work chose the wrong interpretation of an underspecified requirement; short iterations mainly limit propagation of the latter
• Choosing what to learn requires both validity and learning-value gates - Separates two promotion checks for learning loops: whether a candidate is trustworthy enough to learn from, and whether learning it would improve the current system.
• Code complements the weight–prompt pair with independently executed symbolic operations - A model-mediated operation is instantiated by weights plus prompt; code complements that pair by defining operations whose consequences a symbolic runtime executes without reinterpreting the prompt
• Codification - Definition — codification is constraining that crosses from natural language into a symbolic artifact with formal semantics or assigned consequences; executable code is the main practical KB case
• Codification and relaxing navigate the bitter lesson boundary - Since you can't identify which side of the bitter lesson boundary you're on until scale tests it, practical systems must codify and relax — with spec mining avoiding the vision-feature failure mode
• Codify-versus-LLM decision heuristics - Specification, checking, permitted interpretations, and repeated-use cost inform which operations to codify; none alone makes code or model interpretation universally preferable
• Commitment, not derivation, creates new ground truth - Derivation — claims recoverable from the source, nothing added — leaves the source as ground truth; what adds unentailed resolutions becomes ground truth at commit, repaired by supersession
• Constraining during deployment is continuous learning - Continuous learning can happen outside of weights; constraining is one symbolic-artifact form where prompts, schemas, tools, and tests accumulate durable adaptive capacity during deployment
• Continual learning requires governing behaviour-changing writes, not just storing content - For deployed systems, persistence is insufficient; continual learning must select, validate, authorize, and coordinate behaviour-changing updates across the representational forms a system can change
• Cost-sensitive formalisms for fallible theory search - Exploratory map of backtracking, learning, and complexity models that expose budgets relevant to fallible theory-guided search
• Derivation and inheritance give starting warrant; discriminating evidence or proof earns scope - For reusable decompositions, derivation and inheritance supply conditional or transferred starting warrant, while evidence or proof earns only the scope it covers
• Designing a Memory System for LLM-Based Agents - Derives agent-memory design pressures and links to a requirements inventory for agents designing or evaluating memory systems
• Diagnostic richness constrains outer-loop learning quality - Outer-loop learning depends on inspectable failure evidence, not only on the oracle used to select winning candidates
• Discarding all experience-dependent state prevents cross-run accumulation - Discarding an intermediate artifact loses that artifact's reuse path, not all learning; cross-run accumulation fails only when no experience-dependent state survives to affect later work
• Discarding software requires preserving the operational knowledge later work needs - Regeneration must preserve or recover the commitments later operation depends on; reuse, reconstruction cost, and failure consequences matter more than code size or explanatory-reach alone
• Enforcement without structured recovery is incomplete - The enforcement gradient covers detection and blocking but has no recovery column — recovery strategies (corrective → fallback → escalation) are the missing layer, and oracle strength determines which are viable at each level
• Epiplexity by example: what entropy and complexity miss - ELI5 explanation of epiplexity through encrypted messages, shuffled textbooks, CSPRNGs, and chess notation — contrasting surprise, shortest description, and observer-relative usable structure
• Error messages that teach are a constraining technique - In agent systems the error channel is an instruction channel — making errors teach the fix is nearly free and eliminates the agent's need to diagnose, an orthogonal axis to enforcement strength
• Evaluate Memory By Effects, Not By Existence - Memory should be evaluated by downstream effects on tasks, artifacts, answers, behavior, context efficiency, and lineage alignment
• Evaluation automation is phase-gated by comprehension - Optimization loops need diagnostic error analysis and demonstrated judge discrimination before automation can improve behavior rather than just score
• Exact implementation does not validate a requirement against its objective - An artifact can exactly implement a requirement while the requirement remains a conjectured proxy for a declared objective; assess each named path separately, and attribute failure to the link without erasing local correctness
• Explicit retention provides direct targets for selective revision - Explicit artifacts give a learner direct targets for inspecting and revising commitments; durability, writability, and effective addressability still depend on the boundary and available operations
• Factory learning is experience-responsive retention that improves the factory - Experience-responsive retention: production experience determines a retained change to reusable family machinery that later production depends on; factory-level learning is retention that improves the factory relative to a declared objective
• Factory-learning mechanisms should be compared on the same causal job - Compares factory-learning mechanisms on their shared causal job — experience-responsive retention — while separating update mechanisms from the project-theory function needed for open-ended coherent modification
• Feedback-trained memory management is oracle-dependent even when its operations are hand-designed - Fixed and merely runtime-responsive memory rules need no training oracle; outcome-driven updates do, while noisy rankings weaken learning and misaligned ones teach the wrong ordering
• Final task success does not establish intended-path health - When intended-path success and fallback success produce the same final task outcome, the outcome cannot establish whether the intended path and its supporting infrastructure were healthy
• Flat memory predicts specific cross-contamination failures that are empirically testable - Flat memory predicts three cross-contamination failures — search pollution, identity scatter, insight trapping — testable via an observation protocol against real agent systems
• Generation confidence does not by itself certify soundness - Distinguishes next-token probability from factual truth and inferential validity: confidence can support correctness decisions only after task-specific validation, and high-assurance acceptance still needs a separate check
• In-context learning presupposes context engineering - In-context learning only works when the right knowledge reaches the context window — the selection machinery that ensures this is itself learned and refined over deployment
• Information value is observer-relative - Information value is observer-relative: prior knowledge, tools, compute, and goals determine extractable structure, grounding use-shaped reshaping and discovery.
• Inspectable artifact, not supervision, defeats the blackbox problem - Chollet frames agentic coding as ML producing blackbox codebases — codification counters this not by requiring human review but by choosing readable artifacts (code, prompts, schemas) that any agent can inspect, diff, test, and verify
• Instantiation alone cannot model agent learning across sessions - The class/instance analogy captures session startup but omits the retained update relation that can revise later agent definitions and reusable-content placement
• Knowledge artifact - Definition - a knowledge artifact is a retained artifact consumed as evidence, reference, context, explanation, or advice
• Known-target discovery benchmarks show reachability, not discovery closure - Distinguishes backcast and reinvention benchmarks from autonomous discovery: they show that target insights are reachable from supplied ingredients, not that a system can select and verify new discoveries prospectively.
• Learning inside a fixed decomposition inherits its mistakes - Why optimization cannot repair consequential distinctions, responses, or mappings outside the effective update space of a fixed task decomposition
• Legal drafting solves the same problem as context engineering - Legal drafting parallels context engineering because both write ambiguous natural-language specifications for judgment-based interpreters, but law develops constraining more than codification
• Lineage - Definition - lineage records the source dependencies needed to invalidate, regenerate, retire, or review retained behavior-shaping artifacts
• Links encode conditional possibilities, not obligations - Links encode conditional possibilities, not obligations — every label must name a specific reader-need (the condition under which following pays off); content required for all reachable readers should be inlined, not linked
• LLM debugging starts with retry-versus-rewrite triage - Uses execution-versus-interpretation failure to choose the first debugging move: retry a bad execution of a sound reading, or rewrite a specification that reliably induces the wrong reading
• LLM generation can hide a relaxed goal where human writing exposes a stall - An LLM can ship fluent output after silently relaxing an unmet goal, while human composition may expose the same gap as a stall; a conjectural mechanism for why readers inherit the check
• LLM learning phases fall between human learning modes rather than mapping onto them - Pre-training acquires both structural priors (evolution's role in humans) and world knowledge in one pass — making it and in-context learning intermediate on the evolution-to-reaction spectrum
• LLM↔code boundaries are natural checkpoints - At each LLM↔code transition both semantic underspecification and execution indeterminism collapse simultaneously, making these boundaries natural places to anchor debugging, testing, and refactoring
• Local materialization should outperform distant natural-language declarations - Predicts that, for distant or non-obvious uses of a natural-language declaration, generated local materialization will outperform declaration-only presentation without creating a second maintenance authority
• Localized retention pays when sparse changes have bounded impact in a matching decomposition - Addressable retention localizes a sparse change when units match its decomposition; total adaptation stays local only when the affected units also have a small, explicit impact closure
• Machinery persists by warrant, not position, in a reflective loop - Reflection makes selected production machinery challengeable, but placement alone neither warrants nor requires revision; fixed general machinery may persist when its role and scope are earned
• Memory design adds operational axes to artifact analysis - Memory design needs operational policy axes (capture, derivation, activation, authority assignment, lifecycle, evaluation) on top of substrate, form, lineage, and behavioral authority
• Methodology enforcement is constraining - Explains why enforcement strength is a partial order over activation and response semantics, rather than a fixed instruction-to-skill-to-hook-to-script ladder
• Minimum viable vocabulary is the naming set that most reduces extraction cost for a bounded observer - Defines minimum viable vocabulary as the names that most reduce a bounded observer's extraction cost, connecting conceptual thresholds to an information-theoretic optimization
• Opacity is a scale threshold, not a class property - Opacity is not a representational form; any representation becomes practically opaque at sufficient scale, though distributed-parametric artifacts cross that threshold earliest.
• Open-ended construction builds an object and a theory of it - Proposes that construction which must discover and revise an object's organization produces project-specific understanding beyond the object, using programs and theories as its two main cases
• Open-ended theory learning and factory learning close the same reflective loop - Derives one reflective loop from both open-ended theory learning and software-factory learning, and places the Gödel machine by transition licensing and theory provenance
• Operational signals that a component is a relaxing candidate - Operational signals for when a component likely encodes a brittle proxy theory rather than an exact specification and should be relaxed instead of codified harder
• Operative part - Definition - an operative part is the behavior-affecting content, structure, parameterization, or mechanism within a retained artifact or consumption path
• Oracle accumulation improves selection for later candidates in its maintained domain - A failure retained as a lesson helps tasks that retrieve it; retained as a maintained check it improves selection for later candidates in its domain and amortizes validation
• Orchestration strategies and run-state have opposite persistence economics - Separates ephemeral task-specific run state from reusable selection strategies inside host schedulers; RLM-style execution discards both and therefore loses the valuable reusable half
• Progressive constraining commits only after patterns stabilize - Constraining via LLM code generation freezes a single projection of the spec in one shot, but progressive constraining observes behavior across many runs and commits only the interpretations that consistently emerge
• Psychology-to-agent transfer needs per-principle failure-mode testing - Brainstorming a methodology for evaluating cognitive-science-to-agent transfer — assembled from three existing KB notes and tested against Youssef's five psychology principles as worked examples
• Raw accumulation does not create usable memory - Accumulation preserves material, but usable agent memory requires ingress work that adds handles, scope, relationships, provenance, trust signals, and lifecycle pressure.
• Reflective theory refinement has separate structural, epistemic, and implementation lineages - No single predecessor is closest to reflective theory refinement: runtime self-modeling supplies the self-target, classical theory refinement the mechanism with different fillers, and Workspace Optimization only an implementation analogy
• Representational form - Definition - representational form classifies how an operative part is encoded and consumed: natural-language, symbolic, distributed-parametric, or mixed
• Retained artifact - Definition - a retained artifact is retained state that a later agentic loop can consume in a behavior-shaping way, regardless of storage substrate
• Retaining episode evidence keeps a distilled rule open to re-examination - Keeping relevant episode evidence and its relation to a distilled rule preserves a route for re-examining that rule; reconstruction, comparative value, and correct generalization still require testing
• Reverse compression is when LLM output expands without adding information - LLMs can inflate compact seeds into verbose artifacts without adding extractable structure; a KB resists this only when links make additional structure accessible
• Revision guided by rationale needs faithfulness, not just legibility - When revision relies on a rationale to locate a failed premise, a misleading rationale can direct repair to the wrong part; retained rationale is optional for theory refinement
• RLM, λ-RLM, Tendril, and llm-do separate restriction from persistence - RLM variants, Tendril, and llm-do show that control-language restriction and artifact persistence are separate questions, including where cited RLM sources leave post-return lifecycle unspecified
• Scaling absorbs scaffolding at fixed task difficulty, not at the deployment frontier - Stronger models shrink the scaffolding a fixed task needs; durable deployment-specific structure recurs at the frontier only while assigned difficulty keeps pace with capability and some reliability function stays advantageous to externalize
• Selecting an LLM output fixes a result, not its interpretation - Selecting one LLM output for operative reuse creates a stable artifact-testing target without resolving ambiguity inside the text, so generator and artifact tests answer different questions
• Short composable notes maximize combinatorial discovery - The library's purpose is to produce notes that can be co-loaded for combinatorial discovery — short atomic notes are a consequence of this goal; longer synthesized artifacts belong in workshops or derived instructions
• Silent disambiguation is the semantic analogue of tool fallback - When an agent silently resolves unacknowledged material ambiguity in a spec, final success hides that the contract failed to determine the path — an extension of the tool-fallback observability problem
• Soft degradation can bind before the hard cap even when required evidence fits - For quality-sensitive agent work whose required evidence fits within the provider window, volume, complexity, and interference can silently constrain usable context before the hard cap
• Soft-bound traditions as sources for context engineering strategies - Survey of twelve soft-bound traditions as candidate sources for context engineering strategies, with a three-tier assessment of what transfers, what's plausible, and what's blocked
• Spec mining is codification's operational mechanism - Operationalizes codification by extracting deterministic verifiers from observed stochastic behavior — the mechanism that converts blurry-zone components into calculators
• Specification strategy should follow where understanding lives - Among durable artifacts, spec-first, bidirectional spec, and spec mining fit different phases: when understanding is available upfront, discovered during execution, or only visible after observation
• Storage substrate - Definition - storage substrate records where retained state persists, as an operational field distinct from form, lineage, and authority
• System use is an initial selection environment when theory fit lacks a fixed oracle - When no complete fixed oracle decides whether a claim belongs in a working theory, distributed consequences of live system use can provide an initial selection environment
• System use provides evidence of theory fit and causal usefulness, not independent warrant - Consequences of using a claim in a live system can test its integration and causal usefulness, but independent factual, formal, source, or scope evidence is still needed for its warrant
• System-definition artifact - Definition - a system-definition artifact is a retained artifact consumed with instruction, enforcement, routing, validation, configuration, evaluation, or learning force
• System-definition artifacts are crystallized reasoning under context scarcity - Separates heuristic rules that substitute for unavailable read-time reasoning from authority-bearing constraints and symbolic codification, which remain useful even with abundant context
• Systematic prompt variation serves verification and diagnosis, not explanatory-reach testing - Controlled prompt variation either decorrelates checks or measures brittleness under fixed task semantics; Deutsch's variation test instead changes the explanation to test mechanism and explanatory-reach
• The adaptation survey corroborates memory requirements but misses artifact governance - The agentic-adaptation survey supports the memory requirements map by treating memory and skills as adaptive tools, but it needs substrate, form, lineage, and authority governance to become design guidance
• The Bitter Lesson defense portfolio has one load-bearing member for the form-only rebuttal - The production-method versus representational-form distinction answers only a narrow weights-only inference; theory-guided bootstrapping is a provisional first strategy under incomplete global evaluation, not a defense of continuing hand production
• The bitter lesson selects against unearned reach, not against structure - The lesson selects against claims whose reach was asserted rather than earned by a refuting test, not against structure or origin — theory search in readable forms is its own method; earned reach protects the claim, not its carrier
• The bitter lesson selects production methods, not representational forms - The lesson's axis is production method — hand-crafted versus search-and-learning — not representational form. Learned localized forms are therefore a coherent scaling hypothesis, with cross-artifact credit assignment as the decisive open problem
• The deployed system, not the model alone, is the unit of learning - Because prompts, retrieval, tools, and runtime policy jointly determine deployed behavior, model-only learning leaves consequential system choices fixed
• The four-field record exposes an efficiency, security, and sovereignty risk triad - The four artifact-analysis fields exist to surface three architectural review concerns over retained behavior — efficiency, security, and sovereignty — with sovereignty (owner control to inspect, regenerate, delete, roll back) as the new axis
• The readable-artifact loop is the tractable unit for continual learning - Identifies the natural-language-plus-symbolic pair as the tractable first loop for representational-form coevolution because it shares context, operates at current tempos, and already has a codification boundary
• Theory building and capacity building make the same kind of fallible commitment - Theory building and capacity building both retain resolutions their evidence does not entail; an explanatory commitment stays answerable to the object it describes while a constructive commitment changes the object, so retraction differs in kind
• Theory mediation can coordinate heterogeneous factory development - Fallible natural-language project theory may provide an addressable way to coordinate heterogeneous factory development while search, testing, and backtracking construct and revise it
• Theory refinement - Definition — theory refinement revises an existing fallible explicit theory against empirical cases, seeking improved fit with limited changes; the KB extends its representation and subject
• Theory refinement may improve sample efficiency under structured shifts - Conjecture: theory refinement, learning that discovers, assesses, and revises addressable theories, may need fewer target observations when a shift preserves the structure a theory names
• Theory warrant should be tracked at the finest granularity evidence licenses - Treat support for a theory as warrant for only the most specific claim, conjunction, model, and scope the evidence identifies; do not distribute joint warrant beyond what it entails without additional attribution
• Three-space agent memory echoes Tulving's taxonomy but the analogy may be decorative - The value of separating knowledge, self, and operational memory is that each has a different lifecycle — accumulation, slow evolution, and high churn; whether the Tulving mapping adds explanatory power beyond different retention policies is open
• Trace-extracted memory earns authority per operation, not at capture - Trace memories begin as records; verification, abstraction, and consultation earn authority under progressively harder oracles, while unverified stores accumulate guesses presented as knowledge
• Treat continual learning as representational-form coevolution - Behaviour change spans distributed-parametric, natural-language, and symbolic forms, so the question is how their improvement loops relate — not which is the real locus of learning
• Underspecification and indeterminism complicate programming for prompts in distinct ways - Indeterminism doubles test runs (statistical testing over distributions); underspecification doubles test targets (spec analysis for ambiguity). Conflating the two leads to misdiagnosis
• Unified calling conventions enable bidirectional refactoring between neural and symbolic - When agents and tools share a calling convention, components can move between neural and symbolic without changing call sites — llm-do demonstrates this with name-based dispatch over a hybrid VM
• Use tests a decomposition locally; retained rationale is what makes transfer testable - Running a decomposition confirms only that it sufficed here; because many force-sets fit the same split, rationale retained at design time is what gives a transfer claim an antecedent to test
• Use Trace Extraction As Meta-Learning - Trace extraction is an after-the-fact learning path that must respect signal quality, review, and readable-artifact versus distributed-parametric learning boundaries
• Verification needs a typed target before it needs an oracle - A check's warrant depends on a declared target class, so an unverifiable heterogeneous layer is usually blocked by missing artifact classification, not oracle difficulty — ontology precedes oracle
• Warranted reader update is the objective of substantive writing - Defines epistemic interestingness as a relevant, warranted change relative to an intended reader's prior, making contribution selection—not accumulated inputs—the purpose of multistage writing.
• Weakly discriminated qualities tend to be underselected - Statistical conjecture: under named proposal-selection conditions, unequal oracle discrimination yields unequal enrichment; absolute degradation needs an additional directional mechanism