Learning theory
Type: types/tag-readme.md
This tag gathers theory of how systems learn, verify, and improve: accumulation and generalization, continual learning after deployment, verification and error correction, memory architecture, and self-improvement. The anchor is the theory builder: a system that states its theories in localized units, acts on what they say, criticizes them, and lets the result shape the next round, with whether it learns left as the hypothesis under test. Its members span notes and external system analyses, and apply to any system that adapts through retained artifacts, not only to KBs. Deploy-time learning is the phenomenon child: deployment surfaces what design could not, which is the demand for learning after release. Nearby but different: self-improving-systems holds systems that make operative changes to their own organization, while this tag covers learning in general, whether or not the learner revises itself.
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. A few fundamentals carry only the parent tag: Learning is not only about generality (Simon's definition of learning), LLM learning phases fall between human learning modes, and in-context learning presupposes context engineering. Outside the notes collection, trace-learning techniques in related systems compares the external systems that learn from their own traces.
- deploy-time-learning — the phenomenon: deployment meets users, surprises, and forces change after first release; what use reveals that design could not
- 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
- self-improving-systems — systems that make operative, evidence-responsive changes to their own organization, from gradient learners to theory builders; its children cover the theory builder, the improvement loop, reflection, warranted autonomy, and continual learning
Start here
- Theory builder — localized theories that are consumed and criticized for what they say, with the result of criticism shaping the next conjecture; whether a builder learns is tested, not assumed. The research companion states the program's conjectures about builders
- Retained system-definition artifacts enable persistent deployment-time adaptation — 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
- Constraining and extraction can trade generality for reliability, speed, or 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
- evaluation — where learning claims meet oracles, warrant, and experiment design
- 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 without external assessment carries three obligations - Without external assessment a claim needs its own contradiction-and-support rule, a comparison level for objective change, and a performance measure it does not grade itself, plus attribution when it asserts a cause
- 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 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
- Addressable theory - Definition — an addressable theory is a theory formulated in language whose assumptions, scope, and parts can be inspected and revised individually; a graded structural property, separate from tentative status
- 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 addressable theory can coordinate heterogeneous factory development - Tentative natural-language project theory may provide an addressable way to coordinate heterogeneous factory development while search, testing, and backtracking construct and revise it
- 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
- An optimal long-run learning strategy invests in its own machinery - A machinery improvement is paid for once and reused by every later learning episode, so over a long horizon its return can exceed immediate learning; where it does, an optimal strategy diverts effort to the machinery.
- 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.
- Automating KB learning is an open problem - The KB already learns through manual improvement; automating judgment-heavy mutations needs oracles for connections, groupings, and synthesis we cannot yet manufacture
- 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 the symbolic region of constraining: a rule or operation is committed to an artifact with formal semantics or, more generally, fixed rules that determine what behavior is permitted
- 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
- Competing causal theories can guide distinguishing experiments - Why observationally equivalent mechanisms can tell a theory builder what to test next: a noisy binary example separates evidence acquisition from choosing or verifying an explanation.
- 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 tentative theory search - Exploratory map of backtracking, learning, and complexity models that expose budgets relevant to search guided by tentative theories
- 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
- 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
- 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
- LLM debugging separates specification gaps, instruction violations, and run-to-run variation - Choose a debugging move by checking intent against the specification, output against the specification, and variation across repeated runs; failure frequency alone cannot identify the defect
- 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↔code boundaries are natural checkpoints - LLM↔code boundaries expose concrete inputs and outputs for inspection and replay; deterministic execution preserves rather than corrects a wrongly interpreted argument
- 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 - In Commonplace's arrangement, theory learning and software-factory learning require one connected reflective path; proof-governed switching alone does not settle criticism
- 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
- 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.
- Representational form - Definition - representational form classifies how content 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
- Retained theories may improve sample efficiency under structured shifts - Conjecture: retained theories may reduce target observations under structured shifts; a useful theory's reuse benefit is separate from selecting it by estimated explanatory-reach
- 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 of an addressable theory relies on rationale to locate a failed premise, misleading rationale can direct repair to the wrong part; rationale is one optional repair aid
- 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
- 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
- Tentative theory - Definition — a tentative theory is a theory proposed as a solution to a problem, which stays open to criticism and revision however well it has survived; Popper's term with nothing added, a status of every theory
- 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 building has distinct epistemic, structural, and implementation precedents - Conjecture and criticism, causal self-representation, and persistent artifact editing supply different precedents for a theory builder's operations; similarity on one does not establish the others
- 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