{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"computability_boundary_mapping__economics_finance","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"cbm-ef-collateral-allocation-totality-003","proposal_index":3,"version":0,"title":"Totality-First Intraday Collateral Allocation","problem":"A financial institution must assign available collateral lots to margin and settlement calls subject to eligibility, haircut, currency, encumbrance, concentration, custody, and timing constraints. Its optimizer is expected to answer whether a feasible assignment exists for every collateral snapshot. When a heuristic or optimization run reaches its operational deadline, downstream systems can treat the missing answer as “no feasible allocation,” even though the declared snapshot contains finitely many lots and calls and the timeout establishes neither infeasibility nor undecidability. The input representation, numeric precision, divisibility of lots, treatment of changing prices, and evidentiary requirement for an infeasibility verdict are not explicit.","actors":["Treasury and intraday-liquidity managers","Collateral optimization operators","Margin and settlement operations staff","Risk-control function","Custodians and settlement agents","Owners of collateral eligibility and haircut rules","Independent algorithm reviewer"],"observable_state":"For one frozen collateral snapshot, the primary optimizer reaches its resource limit and returns no assignment; the workflow records NO_FEASIBLE_ALLOCATION. A later rerun, alternative heuristic, or manual construction finds an assignment satisfying the same encoded constraints. The original result carries neither a feasible-allocation witness nor evidence that all assignments were exhausted.","consequence":"A computational delay can be mistaken for economic infeasibility, causing funding and settlement decisions to rely on a stronger claim than the computation established; genuine infeasibility can likewise remain unsupported by a checkable exhaustive argument.","affected_objective":"Correct and auditable intraday collateral decisions that distinguish proven feasibility, proven infeasibility, resource-bounded uncertainty, invalid input, and changing-state failure.","intervention":"Define a finite collateral-snapshot contract and provide a constructive total decider before optimizing for speed. Each conforming input contains finitely many integer-unit or indivisible lots, finitely many calls, fixed-point values and haircuts, mechanically evaluable eligibility predicates, explicit concentration formulas, a valuation timestamp, and no unmodeled future events or callable rule programs. The reference procedure recursively assigns each lot to one call or to UNUSED, rejecting constraint-violating partial assignments only when a checked pruning rule proves that none of their completions can work. The number of unassigned lots strictly decreases, so every conforming instance terminates. A satisfying complete assignment produces FEASIBLE with a checkable allocation witness; exhausting every assignment produces PROVEN_INFEASIBLE with an enumeration record. Complexity analysis then estimates growth and establishes an operational envelope. A production solver stopped before exhaustive completion may return FEASIBLE if it has a valid witness, but otherwise returns UNKNOWN_AT_CUTOFF, never PROVEN_INFEASIBLE. INVALID_SNAPSHOT, STALE_SNAPSHOT, and SYSTEM_FAILURE remain separate. The model, reference algorithm, pruning rules, certificates, operational envelope, and recheck triggers are versioned together.","structural_mapping":[{"archetype_element":"Timeout confused with semantic falsity or undecidability","domain_realization":"Failure to find a collateral assignment within an operational deadline is recorded as though no assignment exists."},{"archetype_element":"Problem-class and representation contract","domain_realization":"Lots, calls, settlement units, fixed-point values, eligibility rules, concentration constraints, timestamps, and allowed assignments receive a finite enforceable encoding."},{"archetype_element":"Constructive computability witness","domain_realization":"A uniform enumeration procedure assigns every finite lot to a call or UNUSED and supplies termination and correctness arguments."},{"archetype_element":"Class-wide versus instance-level separation","domain_realization":"The reference procedure proves that every conforming finite snapshot is decidable, while a particular production run may still end at its resource cutoff without a verdict."},{"archetype_element":"Computability versus tractability separation","domain_realization":"Exponential assignment growth is assessed only after total decidability has been established."},{"archetype_element":"Explicit weaker fallback","domain_realization":"An interrupted search can report a witnessed feasible allocation or UNKNOWN_AT_CUTOFF, but cannot infer infeasibility."},{"archetype_element":"Model-relative scope and recheck triggers","domain_realization":"Fractional continuous assets, dynamic future prices, recursively callable rules, rehypothecation cycles, or changed settlement semantics trigger reclassification rather than silently inheriting the finite-snapshot guarantee."}],"mechanism_mapping":[{"mechanism_slug":"constructive_algorithm_and_correctness_proof","role":"Provides the reference decider, a well-founded termination measure equal to the number of unassigned lots, and an argument that complete enumeration returns FEASIBLE exactly when a conforming assignment exists.","counterfactual_removal":"Without the constructive witness, the project could continue treating solver behavior as evidence about computability without possessing a uniform terminating procedure for the declared class."},{"mechanism_slug":"bounded_domain_exhaustive_search","role":"Mechanically checks every lot-to-call assignment inside a declared pilot or operational envelope and supports an exact infeasibility verdict only after complete coverage.","counterfactual_removal":"Without exhaustive coverage, failure to find an assignment would remain an incomplete search result rather than evidence of infeasibility within the stated envelope."},{"mechanism_slug":"computational_complexity_analysis","role":"Measures assignment-space growth, memory use, certificate size, and deadline feasibility after decidability is established, separating practical capacity from in-principle solvability.","counterfactual_removal":"Without the complexity follow-on, a total but rapidly growing reference algorithm could be mistaken for an operationally adequate allocator."},{"mechanism_slug":"proof_checking","role":"Independently verifies the termination and correctness argument, allocation witnesses, complete-enumeration claims, and the soundness of each pruning rule.","counterfactual_removal":"Without checking, an invalid pruning shortcut or arithmetic mismatch could wrongly certify infeasibility while appearing to implement the reference procedure."},{"mechanism_slug":"semi_decision_with_explicit_unknown","role":"Allows a resource-bounded production search to confirm FEASIBLE when it finds a valid allocation and to return UNKNOWN_AT_CUTOFF when it stops without exhaustive coverage.","counterfactual_removal":"Without the explicit unknown state, operational termination pressure would again encourage a cutoff to masquerade as an infeasibility verdict."},{"mechanism_slug":"proof_by_counterexample","role":"A single valid allocation for a snapshot recorded as infeasible refutes the over-broad claim that a timed-out run established nonexistence.","counterfactual_removal":"Without validating such allocation witnesses, conflicting heuristic results could be dismissed as implementation noise rather than decisive counterexamples to the claimed verdict."},{"mechanism_slug":"computability_boundary_decision_record","role":"Records the positive decidability classification, exact input contract, operational cutoff semantics, proof artifacts, complexity handoff, and changes requiring reclassification.","counterfactual_removal":"Without a versioned record, dynamic valuation, fractionalization, or executable constraint features could enter the workflow while the finite-snapshot guarantee remained attached."}],"causal_chain":["Freezing a collateral snapshot and its numeric and settlement semantics makes each instance finite and effectively enumerable.","The constructive algorithm maps every lot to a call or UNUSED while decreasing the number of unassigned lots, establishing termination for every conforming snapshot.","Completeness of enumeration makes a valid assignment sufficient for FEASIBLE and complete exhaustion sufficient for PROVEN_INFEASIBLE.","Independent checking prevents arithmetic discrepancies or unsound pruning from silently weakening that equivalence.","Complexity analysis reveals where the total reference procedure exceeds operational resources without reclassifying the problem as undecidable.","The resource-bounded implementation returns UNKNOWN_AT_CUTOFF whenever it stops without either a valid allocation or complete exhaustion.","Treasury staff can therefore distinguish economic infeasibility from computational delay and choose manual construction, additional compute, alternate funding, or escalation without a fabricated semantic verdict.","The versioned boundary record forces re-evaluation when the snapshot ceases to be finite, static, or mechanically evaluable under the stated model."],"baseline":"Evaluation baseline: a heuristic or mathematical optimizer searches for an allocation until an operational cutoff. A found allocation is reviewed, while a run producing no allocation may flow through the same NO_FEASIBLE_ALLOCATION field used for a proved negative. Manual retries and alternate solver settings provide additional searches but do not establish complete coverage.","nearest_rivals":["Increase solver runtime, memory, or hardware while preserving the existing Boolean interface.","Run an ensemble of allocation heuristics and accept the first satisfying assignment.","Default timed-out calls to cash collateral or an operational buffer.","Have treasury staff construct allocations manually after every unsuccessful solver run.","Use a continuous or linear relaxation and interpret its output as the answer to the discrete settlement problem.","Restrict support to small portfolios without formally distinguishing unsupported snapshots from infeasible ones."],"remaining_contrastive_claim":"The intervention establishes a positive computability result for every conforming finite snapshot and then hands the problem to complexity and operations. Faster solvers, heuristic ensembles, manual allocation, cash defaults, and relaxations may find useful assignments or contain consequences, but they do not by themselves justify a negative verdict. Unlike an impossibility boundary, this proposal makes PROVEN_INFEASIBLE available in principle while preventing a resource cutoff from inheriting that claim.","authority_safety":{"decision_authority":"The institution’s designated treasury or collateral manager retains authority over pledges, substitutions, funding, and settlement responses. The optimization team may issue computational statuses, while risk control must approve the snapshot contract and independent proof review before PROVEN_INFEASIBLE is used operationally.","authorized_first_step":"The head of collateral operations may authorize an offline, read-only exercise on synthetic frozen snapshots with no connection to custody instructions, live calls, trading systems, payment rails, or production eligibility data.","excluded_actions":["Pledge, release, substitute, transfer, reserve, or encumber live collateral","Send settlement, payment, borrowing, or trading instructions","Change haircuts, eligibility policies, concentration limits, or margin calls","Convert UNKNOWN_AT_CUTOFF, STALE_SNAPSHOT, INVALID_SNAPSHOT, or SYSTEM_FAILURE into PROVEN_INFEASIBLE","Apply a result after its valuation timestamp or encoded obligations have changed","Use an unchecked pruning rule in a run capable of issuing PROVEN_INFEASIBLE","Extend the guarantee to continuous fractional assets, future market paths, or executable external rules without reclassification"],"halt_rollback":"Halt the exercise if a FEASIBLE witness fails independent constraint checking, a purportedly exhaustive negative is contradicted by a valid in-scope allocation, the reference enumerator omits or duplicates assignments in a way that affects coverage, arithmetic differs across solver and checker, or status labels are collapsed downstream. Withdraw all affected computational verdicts and return the snapshots to unassessed manual review; the offline exercise makes no pledge or transfer requiring financial reversal."},"negative_tests":{"strongest_counterevidence":"The strongest counterevidence to the proposed positive classification would show that operationally necessary constraints are not finite snapshot predicates but invoke arbitrary external programs, continuously evolving prices, unbounded future obligations, or recursively generated eligibility rules. The finite enumerator would then decide only a semantic surrogate, and the real problem class would require a new computability analysis.","problem_falsifier":"The problem is falsified if the current workflow already requires a valid allocation witness for FEASIBLE, a checked complete-search certificate for PROVEN_INFEASIBLE, preserves all cutoffs as UNKNOWN, and prevents stale or out-of-scope snapshots from receiving semantic verdicts.","intervention_falsifier":"The intervention fails if the reference procedure omits any well-formed assignment, can fail to terminate on a conforming snapshot, accepts an allocation violating an encoded constraint, or certifies infeasibility when an independent checker can exhibit a valid assignment. It also fails semantic fidelity if the finite snapshot contract excludes constraints that operators must apply to execute the resulting pledge.","risks":["The frozen snapshot may become stale while computation is running.","Fixed-point valuation or haircut arithmetic may differ from custody or margin systems.","The indivisible-lot representation may misstate assets that can be split or pooled.","An unsound pruning rule may remove the only feasible branch.","Integer overflow, rounding, or currency conversion errors may invalidate witnesses.","Assignment-space growth may make exact exhaustion unusable before an operational deadline.","Enumeration records may become too large to store or independently check directly.","Operators may continue treating UNKNOWN_AT_CUTOFF as practical infeasibility despite the interface distinction.","A feasible encoded assignment may be operationally unavailable because of an omitted custody, timing, or legal constraint.","Later addition of executable eligibility rules may invalidate the finite-predicate classification."]},"next_evidence_step":"Run one offline exercise on ten synthetic frozen snapshots containing no more than eight collateral lots and three calls, so each lot has four destinations including UNUSED and complete enumeration requires at most 4^8 assignments per snapshot. Include planted feasible cases, planted infeasible cases, binding concentration and haircut constraints, an arithmetic edge case, a deliberately stale snapshot, and a forced early cutoff. Produce the input contract, reference algorithm, termination and correctness argument, allocation-witness checker, exhaustive-coverage record, complexity worksheet, and boundary decision record. An independent reviewer must confirm that every planted feasible case returns a valid witness, every PROVEN_INFEASIBLE case receives complete coverage, and every forced interruption returns UNKNOWN_AT_CUTOFF. No live asset, margin call, eligibility rule, or settlement instruction enters the exercise.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 addressed universal solvency analysis for open-ended programmable derivatives and used a proposed impossibility reduction, an enforceable finite-state language fragment, and sound abstraction to govern contract listing. Proposal 2 addressed termination and strategic truthfulness of executable auction mechanisms and used proof-carrying admission with certificate checking and one-sided proof-versus-counterexample search. Proposal 3 instead addresses finite intraday collateral-assignment snapshots and follows the positive side of the computability map: it exhibits a uniform total decider, proves termination by finite descent, reserves negative verdicts for exhaustive coverage, and then transfers the remaining difficulty to complexity and operational cutoff management. It concerns treasury execution rather than contract or mechanism admission, requires neither the contract fragment from proposal 1 nor the proof-carrying submission regime from proposal 2, and can be adopted independently by collateral operations.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}