{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__pharmacology_toxicology","archetype_slug":"computability_boundary_mapping","domain_slug":"pharmacology_toxicology","title":"Model-Relative Toxicity Reachability Boundary Workflow","opportunity_summary":"Evaluate a claimed universal toxicity-reachability analyzer by checking the proposed computability reduction, mechanically separating decidable model fragments from unrestricted inputs, preserving UNKNOWN, and attaching explicit guarantees to every result. The candidate offers a safe synthetic-model study, but the existence of the claimed operational problem, the reduction, fragment usefulness, and distinctiveness remain unverified.","adopter_authorizer":"A computational-toxicology model owner, with approval from an independent safety-methods reviewer; consequential clinical, regulatory, dosing, release, and exposure decisions remain with their lawful authorities.","scores":{"meaningful_impact":{"score":4,"rationale":"Preventing timeout or failed proof search from becoming a false SAFE verdict could protect consequential toxicity decisions and avoid wasted attempts at an impossible universal analyzer. The packet does not establish how often such claims or errors occur, limiting a top score."},"stakeholder_pull":{"score":2,"rationale":"The candidate identifies model owners, reviewers, developers, and affected populations, but supplies no evidence that an adopter currently experiences this problem, requests the workflow, or will allocate resources to it."},"incremental_advantage":{"score":4,"rationale":"Relative to the stated unrestricted Boolean analyzer and hardware-optimization rival, the proposal adds a materially different decision boundary: checked fragment membership, explicit UNKNOWN and out-of-scope labels, guarantee-bearing outputs, and escalation rather than false safety. Its advantage depends on the alleged universal claim actually existing."},"distinctiveness_plausibility":{"score":2,"rationale":"The composition is coherent, but prior art is explicitly unsearched and the packet supplies no comparison against formal verification, reachability, model-qualification, or toxicology-assurance practices. Distinctiveness therefore has no affirmative closed-book basis."},"technical_implementability":{"score":3,"rationale":"A sandboxed finite-corpus label audit and fragment checker appear bounded and technically approachable, but the decisive reduction is absent, enforceable useful fragments are undefined, and abstraction precision and integration feasibility are unknown."},"adoption_authority_feasibility":{"score":4,"rationale":"The candidate names a model owner and independent safety-methods reviewer with authority over analysis labels and explicitly reserves consequential decisions for lawful authorities. Cross-organizational approval and actual willingness are not demonstrated."},"evidence_readiness":{"score":3,"rationale":"The packet provides falsifiers, a synthetic test setting, a comparator, halt conditions, and auditable labels. However, it contains neither the reduction and theorem assumptions nor pilot data, fragment-coverage results, or evidence that the target overclaim exists."},"safety_net_benefit":{"score":5,"rationale":"Preserving UNKNOWN, prohibiting SAFE from timeout or failed search, isolating synthetic testing, logging outputs, and reverting to manual review directly provide a strong safety net even before the broader computability claim is established."},"scalability":{"score":3,"rationale":"The labeling and routing pattern could be reused across executable model workflows, but each accepted language, endpoint definition, fragment checker, abstraction, and assumption change requires separate formalization and revalidation."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"Formalize one executable model language and endpoint predicate, exhibit and independently review the proposed reduction, define one mechanically enforceable fragment, create a finite synthetic corpus, and compare label correctness with the stated baseline under fixed bounds.","confidence":"MODERATE","assumptions":["Uses synthetic models only and requires no clinical, animal, or proprietary exposure data.","Requires coordinated computational-toxicology and formal-methods labor plus independent review.","Covers one language, one endpoint formulation, and one fragment rather than a general production system."]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build a production-quality fragment checker, guarantee-aware routing and labeling, logging, access controls, model-language adapters, regression tests, documentation, and integration with one existing computational-toxicology workflow.","confidence":"LOW","assumptions":["A usable reduction and fragment have passed the first evidence step.","One organization and a limited number of existing model interfaces are in scope.","Existing compute, identity, audit, and model-management infrastructure can be reused.","No consequential decision authority is transferred to the analyzer."]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Validate the integrated workflow, conduct independent safety review, train users and reviewers, establish escalation and rollback procedures, qualify label presentation, and monitor a controlled non-consequential launch.","confidence":"LOW","assumptions":["Launch remains advisory until governance criteria are met.","Compliance and quality-system work is organization-specific but does not require a new clinical study.","Abstraction alarms and UNKNOWN volume are manageable enough to permit controlled use."]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Maintain fragment definitions and language adapters, review assumption changes, rerun regression and false-SAFE tests, audit labels and logs, support escalations, and update training and boundary records.","confidence":"LOW","assumptions":["Deployment remains limited to one principal workflow and a small set of model languages.","Existing infrastructure absorbs most compute and security overhead.","Major new languages or endpoint classes would be separately funded rather than treated as routine maintenance."]}},"research_burden":"HIGH","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"UNCERTAIN","reason":"The packet describes an observable universal SAFE/TOXIC interface and a consequential failure mode, but marks the situation as a hypothesis and provides no external evidence that such a class-wide analyzer claim or timeout-as-SAFE practice exists."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"The candidate explicitly identifies a computational-toxicology model owner and independent safety-methods reviewer as authorizers for analysis labels, while bounding their authority."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The workflow makes a testable claim that checked fragment routing and explicit SAFE, TOXIC, UNKNOWN, timeout, and out-of-scope labels will outperform the stated unrestricted Boolean baseline without permitting a synthetic false-SAFE result."},"bounded_next_evidence_step":{"status":"YES","reason":"The authorized first step is confined to one formalized language, one enforceable fragment, a finite synthetic corpus, fixed bounds, independent reduction review, and a label audit against the baseline."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"Testing is restricted to synthetic models; consequential dosing, trial, release, exposure, and care decisions are excluded; and explicit halt, rollback, logging, and manual-review conditions are supplied."},"implementation_cost_scope_and_range":{"status":"YES","reason":"The proposed evidence and deployment stages are separable enough to assign broad resource-equivalent bands, provided the stated assumptions limit work to one language, fragment, organization, and workflow. Actual integration and compliance costs remain low-confidence."}},"blocking_evidence":["Evidence that an operational project actually makes the alleged exact, always-terminating class-wide toxicity-reachability claim or converts timeout and failed search into a safety verdict.","A fully instantiated executable language, toxic-endpoint predicate, total computable encoding, answer-preservation argument, and independent review of the proposed halting reduction.","A mechanically enforceable decidable fragment with a valid total procedure and demonstrated coverage of decision-relevant model patterns.","Comparative synthetic-corpus results showing improved label differentiation and no synthetic toxic witness labelled SAFE.","Measurements of abstraction precision, false-alarm burden, UNKNOWN frequency, and user handling of guarantee labels.","Prior-art comparison sufficient to assess whether the workflow composition is distinct."],"next_evidence_step":"Run the specified sandboxed synthetic study for one executable model language: independently review the instantiated reduction, implement one fragment-membership checker, and compare the boundary workflow with the stated unrestricted Boolean baseline on a preregistered finite corpus under fixed bounds. Falsify advancement if the reduction fails, membership cannot be enforced, label differentiation does not improve, or either approach reports any known synthetic toxic witness as SAFE.","research_questions":["Does a class-wide exact terminating toxicity-reachability claim or equivalent Boolean interface exist in the target workflow?","What precise model syntax, execution semantics, exposure inputs, endpoint predicate, horizon, nondeterminism, and external-information assumptions define the claimed analyzer?","Does the proposed reduction provide a total computable encoding and preserve answers under independently reviewed assumptions?","Which model fragments have mechanically decidable membership and valid total reachability procedures?","What share of decision-relevant synthetic patterns falls within those fragments, and how often do remaining inputs yield UNKNOWN or unusable abstraction alarms?","Does the workflow improve correct differentiation of SAFE, TOXIC, UNKNOWN, timeout, and out-of-scope relative to the stated baseline?","Can users and downstream systems preserve guarantee labels without treating UNKNOWN or timeout as SAFE?","How does the composition differ from existing practices, if any, once prior art is examined?","What revalidation triggers are required when model expressiveness, endpoint definitions, or assumptions change?"],"recommendation":"PARTNERED_RESEARCH","uncertainty_constraints":["Closed-book assessment: no external validation of the problem, prior art, prevalence, demand, realized impact, or costs was available.","The candidate supports only a conditional computability-boundary claim because the decisive reduction is not exhibited.","Biological toxicity is not claimed to be undecidable; conclusions apply only to the declared executable language and formal endpoint predicate.","The deployed model class could already be finite or decidable, in which case the main issue would be complexity rather than computability.","Operational value could be low if useful fragments exclude most relevant models, abstractions create excessive alarms, or guarantee labels are stripped downstream.","Cost bands are resource-equivalent planning ranges based on assumed scope, not observed quotations or point estimates."],"closed_book_prior_art_boundary":"Prior art is explicitly unsearched. This assessment makes no claim that the computability analysis, fragment routing, guarantee labels, UNKNOWN preservation, or their composition is novel, uncommon, protectable, or absent from computational toxicology, formal verification, hybrid-systems reachability, pharmacometric qualification, or adjacent assurance practice."}