Self-improving systems

Type: types/tag-readme.md

A self-improving system makes operative, evidence-responsive changes to its own behavior-determining organization, read against a declared frame of boundary, horizon, and improvement objective. Assign this tag to work on whether or how systems make such changes: their definition and classification, mechanisms, limits, or evidence of improvement. It also includes work that meets a child area's inclusion condition, as specified below. Ordinary task execution or automation alone is insufficient. Tagging an analysis does not certify that the analyzed system improves. Deploy-time-learning covers the post-release demand for change and responses to it, including human maintenance; work on self-improvement as a response may carry both tags. Learning-theory covers learning more broadly, whether or not the learner changes its own organization.

The child areas are theory-builder, software-factory, improvement-loop, reflection, warranted-autonomy, and continual-learning. Substantive fit to any of these heads also qualifies for this parent tag. This includes their definitions, boundary cases, and evaluation questions without requiring a separate claim that a system changes itself. The hierarchy groups areas of inquiry; it does not assert that every factory, theory, or autonomous actor is a self-improving system.

Child areas

• theory-builder — the theory-builder definition and its companions: tentative and addressable theory, retained theory guiding search, theory fit and warrant, the Popperian precedents; the research program Commonplace runs
• software-factory — factories and houses: family-specific production machinery, factory development and learning, universality, task versus product families
• improvement-loop — the proposal-selection architecture: search control, what to learn, oracle accumulation, diagnostic richness, false-positive filtering, omitted versus frozen loop functions
• reflection — reflective systems: addressability, second-order lessons, graded coverage, retrieval misses, Gödel machines, and the Commonplace reflective-trace evidence
• warranted-autonomy — which decisions a computational actor is warranted to take over from humans, the oracle domain that bounds it, closure, and measuring autonomy
• continual-learning — how a deployed system keeps learning outside model weights: the readable-artifact loop, form coevolution, governing writes, and the Bitter Lesson defense of that bet

Definition and objective

• Operative change — a change counts when it affects later operation over the declared horizon through a behavioral-authority path
• Evidence bearing on an improvement objective — gradients, rewards, errors, viability signals, tests, judgments: what carries information about the criterion
• The definition classifies its boundary cases without ad hoc exceptions — ten cases run against the criteria; the stress falls on declaring the boundary, not on the definition
• Self-improvement is relative to a declared objective — the objective is a declared parameter beside boundary and horizon, indexed by the analyst and antecedent in the pathway
• Revising an improvement objective is licensed from outside it — objective change is improvement only against a level outside both objectives
• A proximate target is checked for achievement, not for warrant — between objective and oracle sits a target level whose linking claim no check in the loop tests

Accumulation and compounding

• An optimal long-run learning strategy invests in its own machinery — a lasting machinery improvement can repay its cost across later learning episodes, without fixing its representational form
• Accumulation counts dependence through the retained result — cumulativity is later dependence on what was retained, not on the evidence it caused
• Improvements can accumulate without compounding — compounding needs an earlier benefit to counterfactually improve a later episode
• Compounding is tested in later improvement, not by the accepting metric — displaced productivity measures and causal traces, not the metric that accepted the change

Casebooks and records

• Real self-improving systems occupy combinations no single rung captures — thirteen placements from the Homeostat to Commonplace on the profile fields
• Six reported self-improvement paths expose bounded redesign surfaces — operative redesign separated from revision of governing machinery and from declared editability
• The declared Commonplace frame — the boundary under which Commonplace's own attributions are assessed
• Where change candidates come from in Commonplace — problem-noticing and candidate-drafting beyond a maintainer's judgment
• External systems read as whole systems: Exo (a rewritable executor over a protected substrate), Autogenesis (versioned mutation with rollback), Compound Engineering (a compounding claim separated into its parts), Agno AgentOS (a control plane, not a learner), and AI Agents in Depth (book-level convergence and divergence)

Related Tags

• deploy-time-learning — what deployment reveals and the responses, from human maintenance to self-improvement
• learning-theory — the parent area for how systems learn; the six children above are also its routes into self-revision
• computational-model — the execution substrate these systems run on

Other tagged notes

• A benchmark that holds the client fixed exports the least-warrantable decisions by design - A fixed-client benchmark measures worker capability; it leaves broader closure untested when the client supplies internal production decisions, while ordinary user requirements and acceptance may remain external
• 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 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 complete theory path does not establish improved capacity - Mediation, empirical contact, response to criticism, and recurrent mediation support different claims; none alone establishes improved capacity, and retained addressable theory is one realization
• A consumption channel delivers force without the history that earned it - A consumption path can promote content into a higher-force role without checking whether an authorization covers that content, version, and use
• 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 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 method's ceiling bounds the method, not the transfer it already made - Separates envelope expansion, where a responsibility leaves the residual human work, from performance gains inside a fixed envelope, so a bounded method reaching its ceiling does not retract the transfer it already made
• A methodology governs its own extension only as far as it settles the meta-decisions it raises - A retained methodology governs the consequential extension decisions it supplies or imports; actor competence can carry the process further without making those choices settled by the method
• A proposal-selection improvement loop requires search, evaluation, and operative retention - A proposal-selection improvement loop — candidates generated, evaluated with possible non-adoption, and accepted changes made operative — requires search, reject-capable evaluation, and operative retention
• A repeatable operative path keeps a redesign class open to revision - Operationalizes repeatable operative revision for a named redesign class as a causal path through representation, evidence-bearing determination, admission, installation, dependence, and continuity
• A retained instruction preserves what testing selected - Explains why an instruction generated from model weights can still add KB value: testing selects a procedure under a criterion and retention makes that choice reusable.
• A retained-theory intervention isolates one surface, not the whole program theory - An intervention on retained theory estimates that surface's causal contribution under matched conditions; influence, explanatory guidance, acquisition, and whole-system theory possession remain different claims
• A retrieval miss is a local reflective-path failure - A missed relevant artifact leaves its represented aspect inert for the affected task and discovery route, while other loading paths and reflective aspects can remain causally connected
• A search controller is tested by what it brings to stronger evaluation - A search controller should be evaluated by the branches and probes it routes into stronger evaluation, not by treating every provisional judgment as an acceptance claim
• A 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
• 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
• A vibe-noting trace shows persistence enables revision, not certification - Evidence from one Commonplace note history: persistence enabled later semantic development while review exposed omitted risks, attribution drift, a link error, and an unresolved authority boundary
• 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
• 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
• 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 agentic substrate becomes a software factory through family-specific production machinery - Maps the bounded-call agentic substrate to Greenfield's software-factory ontology without calling every generic harness or generated program a factory
• An omitted improvement-loop function and a frozen one need different repairs - Five proposal-selection systems expose frozen functions, while a direct-update contrast shows why absence of a gate is not omission; HyperAgents supplies a preliminary partial unfreezing
• 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
• Backtracking keeps lightweight search control provisional - Backtracking preserves the provisional status of a heuristic branch choice by restoring an earlier usable state and redirecting search after contrary evidence
• Broad software demands create pressure for agentic factory development - Broad software demands make exhaustive predefinition of useful family-specific production machinery practically implausible, motivating agentic factory development without ruling out a fixed universal substrate in principle
• Causal and proof obligations are two formal routes to assessing explanatory-reach - Causal and proof obligations demonstrate two ways formal symbolic systems can assess explanatory-reach inside a warranted model
• 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.
• Citing retained theory at the decision point is a mediation trace - A decision record that cites the theory it followed supplies cheap, checkable evidence that the theory was consumed — necessary for a record-based mediation claim, but short of showing correct or load-bearing use
• 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
• Commonplace as a reflective self-improving system - Commonplace witnesses that a human-inclusive KB can be reflectively self-improving on one pathway despite uneven coverage and human-gated design judgment
• Commonplace builds a theory builder and tests whether it learns - Commonplace builds a system meeting the theory-builder conditions and tests whether it learns; criticism of content is compared against trial and error, and fine-grained addressability and high persistence against builders with less of each
• 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.
• Computationally directed self-improvement is a fixed-boundary reallocation ending in contraction - The progress question for self-improving systems is not category membership but which decision-bearing functions humans still supply; the endpoint test is whether the boundary can be contracted to exclude them
• 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
• 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
• Disconnected witnesses do not establish a full causal path through theory - Evidence of recurrent learning through theory must identify the joins of one causal path; connected use and criticism still need separate evidence of improved capacity
• Distinct residue classes require distinct functions in a self-improving architecture - Different reasons for an untransferred decision identify different missing functions; a single process can supply several, and the current carrier split is not a permanent requirement
• 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
• 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 construction is not evidence of production-knowledge acquisition - Recursive software-factory construction is prior art, but the demonstrated constructors receive the family definitions, metamodels, mappings, and expertise that determine the produced factory
• Factory development - Definition — factory development constructs or revises reusable family-level production machinery rather than one product's lifecycle state
• 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
• False-positive generation is filtered; false-positive acceptance becomes operative - False-positive generation faces evaluation before retention, while false-positive acceptance becomes operative and can compound
• Gödel machines are a proof-governed case of reflective self-modification - A Gödel machine admits self-rewrites through proof under its current formalization; this restricts admission without establishing how many useful changes are reachable or how reliably they are found
• Holding a program theory means sustaining coherent search under delayed feedback - Holding a program's theory is tested by whether a partial, tentative account of what the program is for keeps modification search, backtracking, and recovery coherent until delayed evidence arrives, not by whether the first change is right
• Improvements outside the admitted formal language need a pre-formal stage somewhere - An improvement whose concepts have no expression in a loop's admitted formal language is reached only through a pre-formal stage, inside the loop or fixed at design time in the choice of language; translation relocates that stage
• 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
• Increasing computational autonomy relocates human effort to the frontier instead of reducing it - In an open-ended system, increasing computational autonomy need not cut total human hours — attention moves to the frontier — so measure improvements per human judgment, not human time
• 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
• 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
• Lightweight search control allocates further search without licensing adoption - A search judgment is lightweight when its authority stops at allocating further investigation, probing, continuation, suspension, or abandonment rather than licensing an operative change
• LLM-executed methodologies are metacircular interpreters, not compilers - Self-hosting LLM methodologies are closer to metacircular interpreters than compilers: agents re-interpret natural-language rules each session, while stable paths codify into validators and commands
• 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
• Make Authority Explicit - Memory architecture must state who can read, write, promote, activate, enforce, revise, and retire memory across risk levels
• Measuring autonomy well enough to see it improve is an open problem - Autonomy is reported per function rather than scored as a percentage, but that profile does not yet support comparison across systems or time
• Mechanistic constraints make Popperian KB recommendations actionable - Bounded context and underspecification don't just permit conjecture-and-refutation — they require it; derives three concrete practices (falsifier blocks, contradiction-first connection, rejected-interpretation capture) from KB mechanics.
• Methodological and computational closure track different changes - Methodological closure tracks what a retained method settles; computational closure tracks the absence of human decisions during the assessed operation, without requiring every judgment to have explicit criteria
• Methodology with incomplete coverage and its live theory fallback form a two-layer execution system - In open or incompletely covered domains, the theory-derived fast path and live theory fallback co-execute while methodology-native content follows a separate maintenance regime
• Missing rationale does not exclude a theory builder; weight-only retention excludes one across runs - The August 2026 Prime Agent, Recuris, and Apodex reports: retained rules without a recorded rationale leave theory-builder membership open, while Apodex carries only weights across runs, where no unit says anything, so no builder spans its runs
• Moving the interpretation–enforcement boundary requires cross-form coverage - Moving responsibility between model-interpreted rules and formal enforcement crosses natural-language and symbolic forms, so governing the transfer requires coverage of both and their mapping
• Natural-language project state may specialize weight-resident search heuristics - The natural-language part of project state may specialize general search heuristics already represented in an LLM's weights by supplying current intent, theory, branch history, and constraints
• Naur's human-only conclusion needs more than the absence of explicit criteria - Naur's human-only conclusion needs a further premise connecting unformulated judgment to computational inability; this reading preserves his functional tests without claiming that learned criteria are inexpressible
• 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 improvement must allocate search before decisive evaluation is available - Open-ended improvement must choose which questions, candidates, experiments, or proof paths to develop before decisive evidence about them is available; even a Gödel machine's proof gate retains this prior search problem
• 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
• 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
• Preferential codification concentrates less predictable work at the agent boundary - Explains the negative-selection mechanism by which preferential codification changes the composition of work retained at an agent boundary
• Project-theory possession requires comparing new demands with existing organization - For open-ended modification, project-theory possession includes relating a new demand to existing responsibilities before parallel structure becomes the default; an explicit assimilation branch may counter additive coding-agent patches
• Promote Only When Future Value Exceeds Maintenance Cost - Candidate memory should become durable only when future retrieval or activation value exceeds review and maintenance cost
• Reach-assessment - Definition — judging whether a commitment's claimed explanatory-reach is genuine across natural-language, symbolic, and distributed-parametric forms
• Reflection buys addressability - Self-improvement can accumulate without reflection — parametric learners do — but non-reflective retention gives only indirect handles; reflective retention makes the changed object addressable
• Reflection makes retained lessons second-order: a lesson can reject or rescope a prior commitment - Reflection lets a retained lesson target a prior commitment explicitly — rejecting, revising, or rescoping it — while non-reflective correction acts indirectly through the substrate
• Reflective coverage is graded across representational forms - Reflective coverage is stated per represented form and operation profile; control of an external dependency does not make that dependency part of the system's reflective coverage
• Reflective system - Definition — a system is reflective relative to selected aspects when an internal process uses a causally connected self-representation of them in its operation
• Retained system-definition artifacts enable persistent deployment-time adaptation - Retaining evaluated changes to behavior-shaping prompts, rules, tools, and tests gives deployed systems a persistent adaptation path outside model-weight updates
• 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
• 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
• Six Commonplace paths establish broad addressability, not completeness - A six-path Commonplace audit establishes broad path-relative addressability without establishing completeness, while exposing separate admission and model-realization gaps in the broader revision affordance
• Software factory - Definition — in the Greenfield lineage, a software factory is a configured family-specific software-production environment
• Software house - Definition — a software house is the complete persistent system responsible for developing and evolving software for external users
• 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
• Stale self-description conceals its own staleness - What artifact drift adds when it is reflexive: the process that would detect it consults the artifact that drifted, the trigger has no edit event to hook, and synchronization load scales with autonomy
• 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
• Task families and product families classify different things - Task families group obligations or evaluations; software product families group products through declared commonality, variability, and reusable production scope
• 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 2026-08-30 Commonplace revision used retained theory to guide computational search - A 2026-08-30 Commonplace revision shows retained project theory guiding computational search while the operator supplied decisive global-fit selection
• The augmentation-automation boundary is discrimination not accuracy - Crossing from augmentation to automation requires per-instance discrimination, not aggregate accuracy — discrimination is empirically stagnant, so scaling capability alone cannot cross the boundary
• 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 boundary of automation is the boundary of verification - Synthesis — oracle theory, labor economics, frontier-lab capability predictions, and supply-chain integrity evidence converge on verification cost as the primary structural determinant of automation
• 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 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
• The tag-readme change as an observed causal-connection trace - One bounded Commonplace trace establishes that operative self-representation can exert causal force in both directions; it does not establish system-wide reflective coverage
• Theory builder - Definition — a theory builder applies conjecture and refutation to stated theories: it acts on them, criticizes what they say, and lets the result shape the next round; addressability and persistence are graded
• 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
• Tool usefulness, computational autonomy, warrant, and system power are separate dimensions - Tool usefulness, computational autonomy, warrant, and system power move independently in a human-agent system, so a progress claim has to say which one moved and autonomy gains do not license power claims
• 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
• Universal software factory needs a declared universality axis - Universal software factory is ambiguous unless the universality axis, covered class, supplied inputs, adequacy relation, and resource bounds are declared
• 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
• Warranted autonomy is bounded by oracle domain - Bare autonomy is free, but warranted evaluation autonomy extends only to the candidates an oracle can assess with the required confidence
• Warranted transfer out of the human cut leaves people the hardest-to-warrant decisions - When a system preferentially transfers decisions whose premises, criteria, and checks are available, the remaining human decisions become harder to warrant per decision; this predicts a residue composition, not structural computational openness
• 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
• World models assess explanatory-reach through action-conditioned prediction - Learned world models can assess explanatory-reach when action-conditioned predictions are tested across the interventions or shifts a commitment claims