The governing environmental evidence layer. It separates observation, record preservation, admissibility, interpretation and execution so environmental consequence rests on preserved environmental reality rather than isolated readings or retrospective explanation.
SCHOOL ENVIRONMENTAL INTEGRITY · EIG × AIR × AEA
A school should not merely monitor its air.
It should be able to prove its environmental reality.
This showroom begins again from the parent architecture. It does not reduce Atmospheric Integrity Records to CO₂, PM2.5, indoor air quality, or a pair of sensor streams. It examines whether a school can preserve environmental reality as a governed atmospheric record and whether that record is sufficient for consequential action.
THE ARCHITECTURE
Three layers. One consequence chain.
The atmospheric record class. It preserves the condition of air and the environmental/system context affecting that air over time: baseline, change, duration, recurrence, occupancy/exposure context, equipment operation, intervention history and post-intervention verification.
The parent architecture. Once the environmental record is sufficiently established, AEA governs whether consequence may bind, commit, execute and become a verified outcome.
NO ADMISSIBLE ENVIRONMENTAL RECORD. NO ADMISSIBLE ENVIRONMENTAL EXECUTION.
ENVIRONMENTAL INTEGRITY GOVERNANCE
Govern the environmental record before the environmental claim.
EIG applies AEA to physical environmental reality. It governs how conditions are observed, recorded, preserved, sequenced, validated, admitted, interpreted, acted upon and verified. For a school, the environmental record can support ventilation changes, filtration changes, inspection, remediation, relocation, closure/reopening, automated control, disclosure or other consequence—but the consequence does not get to manufacture the evidence that supposedly justified it.
ATMOSPHERIC INTEGRITY RECORD
The atmosphere is a chronology, not a screenshot.
AIR is the atmospheric record class inside Environmental Integrity Governance. In this school examination, AIR is built as a synchronized record of the atmosphere and the systems and events that shape it. The exterior atmosphere supplies environmental baseline context; the interior atmosphere shows the occupied-space condition; building, equipment, occupancy, exposure and intervention context explain the time-sequenced relationship without being allowed to rewrite the measured reality.
The outside atmosphere relevant to the school and its air intakes: what environmental condition is the building drawing from, when, and from where?
The occupied-space condition over time: psychrometric state, particulates, gases, pressure/airflow relationships and other governed atmospheric channels supported by the actual evidence profile.
The systems shaping the atmosphere: ventilation, filtration, pressure, HVAC operating state, equipment cycling and other machine/environment evidence that is actually instrumented and admitted into the frozen record.
Who or what is present, when, for how long, and under what use condition—where the frozen examination requires that context.
What changed, what action occurred, what equipment or operating state changed, and whether the atmospheric reality changed afterward.
Source identity, location, observation time, sequence, gaps, continuity status and the basis on which a record is or is not fit for reliance.
SCHOOL CASE · CLASSROOM 214
Do not start with the command. Start with reality.
Exterior conditions, interior conditions, occupancy/use context and relevant building/equipment state are observed under the frozen evidence profile.
T1 · RECORDThe admitted sources are synchronized into the Atmospheric Integrity Record with provenance, location, time, sequence and gaps preserved.
T2 · CONTINUITYThe record shows whether the condition is isolated, persistent, recurring, interrupted, changing, or affected by equipment operation or an environmental event.
T3 · ADMISSIBILITYThe record is tested for whether it is sufficiently complete, current, attributable and reliable for the specific proposed consequence.
T4 · BINDINGOnly then is the admissible environmental basis attached to the exact space/system, proposed consequence, authority, scope and present standing.
T5 · COMMITThe basis is revalidated immediately before consequence. A material changed condition cannot silently inherit an earlier permission.
T6 · EXECUTIONThe authorized ventilation, filtration, inspection, remediation, relocation, refusal or other bounded action occurs—or does not occur.
T7 · OUTCOMEThe post-action environmental state is preserved. Outcome closes the chain and becomes part of the next reality.
THE ADVERSARIAL RULE
Now try to destroy all three layers.
Can ordinary school sensors, BMS histories and IAQ platforms already preserve an equivalent governed environmental record?
Can existing systems preserve the outside baseline and inside condition as one time-aligned evidentiary context rather than unrelated dashboards?
Can the record show the equipment/operating conditions that produced or changed the atmosphere, rather than atmospheric values alone?
Can it prove duration, recurrence, interruption, gaps, intervention and post-intervention change without reconstructing the story later?
Is there a defined mechanism deciding whether the environmental record is sufficiently complete, current and reliable for the particular consequence?
Is the admitted environmental basis bound to the exact proposed action, affected space/system, authority, scope and condition?
If environmental or system reality materially changes before execution, must the basis be revalidated before the earlier permission can travel forward?
Remove the TA-14 stack. Can native school systems perform the entire chain equivalently and preserve a reconstructable outcome? If yes, record NOT DISTINCT.
R1 · NATIVE SCHOOL ENVIRONMENTAL ARCHITECTURE
Give existing school practice everything it already owns.
R1 attacks only the evidentiary half of the chain: REALITY → RECORD → CONTINUITY → ADMISSIBILITY. It does not ask whether TA-14 sounds more complete. It asks what established school environmental practice already does without TA-14.
Established school IAQ practice already inspects both exterior and interior conditions, measures temperature, relative humidity, air movement and airflow volume, uses CO₂ as a ventilation indicator, examines outdoor-air intakes and pollutant sources, and documents HVAC zones and outdoor-air delivery. TA-14 receives no credit for discovering or measuring these conditions.
EPA school practice already uses walkthrough records, ventilation logs, zone documentation, checklists, floor plans, locations, observations, irregularities, baselines and retained copies for future reference. A school can already build a substantial environmental record without AIR.
Existing guidance calls for recurring assessments, monitoring as part of maintenance, baseline comparison, analysis of changes over time, recording causes of irregularities, and tracking HVAC/filter/energy and environmental measures. This defeats any claim that AIR uniquely introduces chronology or comparison over time.
Existing school guidance explicitly begins outside, identifies exterior pollutant sources and intake conditions, then evaluates interior conditions and ventilation paths. It therefore already relates exterior conditions to the indoor environment. R1 does not yet establish that native practice synchronizes those observations into the same continuously governed evidentiary object AIR proposes.
EPA guidance documents HVAC zones, outdoor-air intakes, air handlers, energy-recovery units, exhaust paths, expected occupancy, outdoor-air volume and air distribution. AIR cannot claim that environmental evidence becomes meaningful only when related to building systems or locations.
The examined public school guidance provides assessment procedures, review, prioritization and response, but R1 did not find an explicit equivalent to a machine-testable admissibility determination declaring whether a synchronized environmental record is sufficiently attributable, current, continuous and complete for one exact downstream consequence.
Public basis examined for R1: U.S. EPA IAQ Tools for Schools Framework, Coordinator's Guide, Reference Guide, Ventilation Checklist/Log, Preventive Maintenance Guidance, Walkthrough Inspection Checklist and Operations & Maintenance documentation guidance. This round establishes overlap, not uniqueness.
R2 · ATTACK SYNCHRONIZATION + CONTINUITY
Can native school practice already carry outside, inside and system state as one usable chronology?
R1 removed broad claims around measurement, baselines, recordkeeping and longitudinal monitoring. R2 attacks the narrower remainder directly: whether existing practice already establishes the relationship among the exterior environment, occupied interior, ventilation path and changing system state strongly enough that AIR adds no distinct evidentiary function.
EPA school guidance does not treat outside air as irrelevant. It inspects outdoor-air intakes and contaminant sources, documents outdoor-air volume and air distribution, and describes ventilation as the movement of outdoor air into occupied zones while indoor pollutants are exhausted. The physical exterior-to-interior relationship is already part of native school IAQ reasoning.
EPA O&M guidance calls for a separate record for each HVAC zone and maps outdoor-air intakes, air handlers, energy-recovery equipment, exhaust paths, occupancy, outdoor-air volume, distribution, filtration and controls. Native practice can therefore associate environmental conditions with the actual system serving the occupied space.
EPA preventive-maintenance guidance explicitly recommends baseline data, comparison over time, explanation of irregularities, updated baselines and assessment of whether actions made a difference. AIR cannot claim baseline/change comparison or post-action evaluation as unique.
The examined guidance can place exterior sources, interior zones, HVAC assets and longitudinal observations in the same investigation. R2 did not find a requirement that all evidence relied upon for a consequence be synchronized to a common temporal state with explicit observation-time, source, gap and freshness semantics.
Recurring monitoring and trend comparison exist, but that is not yet equivalent to an explicit continuity determination: whether the environmental proposition remained valid across the interval, whether a material gap or change broke continuity, and whether the downstream consequence must therefore be re-established.
AIR receives no credit merely for connecting outside air, inside air and HVAC operation conceptually. The surviving candidate is the governed synchronization of those evidentiary streams into one attributable temporal state whose continuity and fitness for reliance can be tested before consequence.
R2 does not establish uniqueness. A modern BMS, historian, digital twin, commissioning platform or environmental data system may already close this gap. R3 must search for the strongest such near-neighbor rather than compare AIR only with general school guidance.
R3 · STRONG TECHNICAL NEAR-NEIGHBORS
Put AIR against historians, semantic building data and advanced controls.
R3 stops using general school practice as the opponent. The comparison is now against technologies designed to preserve and contextualize building data: Project Haystack historization/semantic modeling, BACnet trend logging, and ASHRAE Guideline 36 control/FDD practice.
Haystack models historized points as timestamp/value histories, supports sampled and change-of-value history semantics, requires timezone context for historized points, exposes history status/errors, and supports synchronization of history from downstream devices. AIR cannot claim timestamped building time-series or history synchronization as distinct.
Haystack associates points with sites, spaces and equipment and provides typed references, units, timestamps and weather concepts. That is strong native support for provenance/context around exterior, interior and machine data. AIR cannot claim semantic asset/location binding by itself.
BACnet provides native trend-history mechanisms, including multi-property logging capable of placing a set of properties into the same timestamped record. This is direct pressure against any claim that AIR uniquely time-aligns multiple environmental or equipment channels.
Guideline 36 standardizes advanced HVAC sequences and supports real-time fault detection/diagnostics. The 2024 edition includes an outdoor-air-pollution mode that can disable airside economizers when outdoor contaminants exceed configured thresholds. Native controls can therefore relate outdoor environmental conditions to consequential HVAC behavior.
Historization modes, timestamps, history status and errors can expose gaps or degraded collection, while FDD can detect operational faults. R3 establishes significant machinery for continuity analysis. It does not yet establish the same explicit rule that a material evidentiary break invalidates fitness for a particular downstream consequence.
The sources examined establish sophisticated recording, semantics, synchronization, controls and diagnostics. R3 did not establish an equivalent cross-domain determination that the exact synchronized environmental evidence set is sufficiently current, attributable, continuous and complete to be admitted as the basis for one exact consequence.
R3 basis: Project Haystack 4.0 historization and point semantics; BACnet TrendLog/TrendLogMultiple mechanisms; ASHRAE Guideline 36-2024. This comparison establishes overlap, not uniqueness. R4 must attack the remaining evidentiary-admissibility function outside ordinary building controls.
R4 · ENVIRONMENTAL DATA FITNESS-FOR-USE ATTACK
EPA already asks whether environmental data are good enough for the decision.
R3 left a narrow AIR proposition: explicit evidentiary admissibility for a synchronized environmental state before reliance. R4 attacks that proposition with environmental quality systems rather than building controls.
EPA's Data Quality Objectives process determines the type, quantity and quality of environmental data needed to support a decision and establishes performance and acceptance criteria before collection. AIR cannot claim the general idea that environmental evidence must be sufficient for its intended decision.
EPA defines data usability around whether data quality meets the intended use. Its Data Quality Assessment asks whether a data set is suitable for its intended purpose. This directly pressures AIR's language of fitness for reliance and admissibility.
EPA systematic planning explicitly uses criteria such as precision, bias, accuracy, representativeness, comparability, completeness and sensitivity, and then evaluates generated information against project objectives. AIR cannot claim explicit sufficiency criteria for environmental information as distinct.
EPA ambient-air data validation requires reliable data of known quality, documented collection, retained raw data, traceability, integrity, systematic review and evaluation of fitness for use. AIR cannot claim validation, traceability or defensible environmental evidence as broad inventions.
EPA's framework ties environmental-data quality to an intended use or decision and can specify acceptable decision-error limits. This is much closer to AIR than a BMS historian because the evidence is judged against the decision it must support.
The examined EPA quality framework is powerful evidence governance, but R4 did not establish the same runtime requirement that a live, synchronized exterior/interior/system state remain continuously admissible through the final boundary immediately before a machine or human consequence commits—and that a material changed condition invalidate the prior basis.
R4 basis: U.S. EPA Data Quality Objectives, systematic planning, Data Quality Assessment, QAPP/data-usability guidance and ambient-air data-validation guidance. R5 must test the surviving temporal proposition together with AEA rather than continue treating AIR as an isolated data architecture.
R5 · CROSS THE AIR → AEA SEAM
Continuous re-evaluation already exists. Does the physical consequence boundary still add anything?
R4 removed general environmental fitness-for-use as an AIR distinction. R5 therefore attacks the runtime seam itself: CONTINUITY → ADMISSIBILITY → BINDING → COMMIT. The strongest comparison is no longer an IAQ framework. It is continuous policy decision and enforcement architecture.
NIST Zero Trust Architecture separates policy decision from policy enforcement. A policy engine evaluates policy and contextual inputs to grant, deny or revoke access; a policy administrator and enforcement point carry that determination into the protected resource path. AEA cannot claim the abstract pattern “evaluate current context, decide, then enforce” as distinct.
NIST's model uses current asset state, identity, compliance, threat intelligence, activity logs and policy as inputs to the decision. The architecture explicitly rejects implicit trust and can countermand a previous approval. AEA cannot claim that authorization must depend on present context rather than historical permission.
NIST's implementation guidance describes real-time and continuous policy-driven, risk-based assessment to establish and maintain access. This directly attacks the idea that continuity plus revalidation is inherently distinctive.
The policy engine makes/logs the decision, the policy administrator executes it, and the policy enforcement point enables, monitors or terminates the connection. That is a mature decision/enforcement seam. AEA cannot claim a generic pre-consequence gate or revocation mechanism as unique.
Zero trust governs digital access to resources. R5 did not establish that it treats a synchronized physical environmental state—exterior baseline, interior atmosphere and building/equipment context—as the evidentiary object whose continuity must survive until a physical consequence commits.
R5 also did not establish the same environmental chain-closure semantics: execution consumes the bounded determination, physical outcome closes the chain, and that outcome becomes part of the next environmental reality requiring a new record and new determination for a later consequence.
R5 basis: NIST SP 800-207 and NIST SP 1800-35. Zero trust is used here as an architectural near-neighbor, not as a claim that cybersecurity access control and environmental execution are equivalent domains. R6 must search specifically for physical cyber-physical/safety architectures that may already close this remaining gap.
R6 · CYBER-PHYSICAL + FUNCTIONAL-SAFETY ATTACK
Physical sensing, timing, assurance and protective action already exist.
R5 removed the generic runtime gate. R6 attacks the remaining physical distinction using NIST cyber-physical-system architecture and the ISA/IEC 61511 functional-safety lifecycle.
NIST defines cyber-physical systems around interacting digital, analog, physical and human components, integrating computation, communication, sensing and actuation with the physical world. EIG × AIR × AEA cannot claim the general sense → reason → act relationship with physical reality as distinct.
NIST treats timing and synchronization as first-class CPS concerns and explicitly studies specification and verification of timing constraints. TA-14 cannot claim that physical consequences require bounded timing or synchronized state as a novel principle.
The CPS Framework includes realization and assurance facets plus trustworthiness, data, boundaries, timing and lifecycle concerns. Assurance cases and artifacts can document whether required properties are established. TA-14 cannot claim lifecycle assurance or evidence-backed physical-system trustworthiness broadly.
The functional-safety lifecycle governs safety instrumented systems from design through operation and maintenance to achieve functional safety. Safety architectures already sense process conditions and carry engineered protective logic into physical final elements. AEA cannot claim physical interlock or protective execution as distinct.
These architectures strongly cover physical state, timing, assurance and protective action. R6 did not establish that they create AIR's particular synchronized exterior/interior/system environmental chronology and explicitly determine that chronology's evidentiary admissibility for a non-safety as well as safety consequence.
R6 also did not establish the same mandatory transaction-like rule that one bounded determination is consumed by one consequence, its measured environmental outcome closes that chain, and any later consequence must begin from the resulting new environmental reality rather than inherit the earlier permission.
R6 basis: NIST CPS/IoT Framework (SP 1500-201/202/203 and related NIST CPS work) and ISA-84 / IEC 61511 functional-safety lifecycle. These are architectural near-neighbors; R6 does not claim they are school IAQ implementations or equivalent to the complete TA-14 proposition.
R7 · ATTACK THE COMPLETE COMPOSITION
Can existing safety and control architecture reproduce the whole environmental transaction?
R6 removed physical sensing, timing, assurance, interlocking and protective execution as individual distinctions. R7 therefore gives TA-14 no credit for any component by itself. The only object under examination is the composition as a whole.
Historians, semantic building models, multi-point logs, environmental monitoring and CPS architectures can preserve time-related physical and system state. No distinctness assigned.
Environmental quality systems can define decision-bound quality objectives, acceptance criteria, validation and usability. No distinctness assigned.
Continuous policy architectures, advanced controls and safety systems can re-evaluate state and revoke, inhibit or change permitted behavior. No distinctness assigned.
Control and safety architectures bind logic to particular outputs, equipment and final elements. AEA receives no credit merely for binding a determination to an actuator or command.
R7 has not established a general environmental architecture requiring that the evidentiary determination authorizing one exact physical consequence be consumed by that consequence rather than persist as reusable permission.
Existing systems measure post-action state and safety lifecycles evaluate performance, but R7 has not established the same mandatory environmental transaction rule: the measured outcome closes the evidentiary/execution chain and becomes the starting reality for any later consequence.
This is not a uniqueness finding. It is a search result bounded to the architectures examined through R7. R8 must now attempt a direct one-to-one mapping against the strongest complete near-neighbor. If no single architecture closes the composition, the surviving claim must remain explicitly bounded to composition—not its component parts.
R8 · DIRECT ONE-TO-ONE NEAR-NEIGHBOR MAP
NIST CPS is the strongest complete public near-neighbor examined so far.
R8 does not compare isolated features. It maps the surviving EIG × AIR × AEA composition directly against NIST's Cyber-Physical Systems Framework and then applies event-sourcing pressure to the remaining record/state-transition semantics.
NIST CPS explicitly integrates physical and logical components, sensing, computation, communication and actuation. It recognizes tight physical/logical state linkages and interaction with the operating environment. The general reality-to-action loop is not distinct.
NIST CPS explicitly contemplates control loops using fused inputs and systems that measure, sense, calculate and act on their environment. The composition of multiple observations into a physical-system state is not distinct by itself.
NIST CPS makes timing a central architectural concern, including latency bounds, timestamp accuracy and synchronization needed to coordinate actions. AIR × AEA cannot claim time-bounded physical state as a broad distinction.
NIST CPS assurance addresses whether required properties are met in the actual operating environment, and its safety analysis seeks engineering evidence that safety functionality behaves as intended. Broad evidence-backed assurance at the physical boundary is not distinct.
Event-sourcing architecture evaluates a command against current state, emits an event when the command succeeds, appends that event to an immutable chronological history, and derives the next state from accumulated events. This strongly pressures TA-14's consequence-consumption and outcome-to-new-state semantics.
R8 still did not establish one public architecture requiring the exact synchronized environmental evidence object to pass an explicit admissibility test, remain valid through physical commit, be bound to one consequence, and then make the measured physical environmental outcome the closure artifact from which the next admissibility chain must begin.
R8 basis: NIST SP 1500-201/202/203 and NIST CPS/IoT program material; event-sourcing architecture is used only as a cross-domain state-transition/history near-neighbor. R8 is a bounded search finding, not proof that no equivalent exists. The next legitimate attack is operational: freeze the surviving chain as an executable contract and try to break it with timing, stale evidence, conflicting sources, sensor loss, changed conditions and unsafe HOLD behavior.
R9 · EXECUTABLE ENVIRONMENTAL CONTRACT
Freeze what survived. Then try to make it fail.
R8 ended the literature-comparison phase. R9 converts the surviving EIG × AIR × AEA composition into a deterministic contract. The numbers below are test fixtures, not school exposure limits, health thresholds or operating guidance.
The interior and equipment records remain current, but the exterior baseline exceeds the frozen maximum evidence age before commit. The synchronized AIR state is no longer complete for the consequence. An earlier ALLOW cannot travel forward.
A required atmospheric or system source disappears, reports invalid quality, or cannot be attributed to the frozen location/asset. The contract must not silently substitute an unapproved source or treat missing evidence as permission.
Two admitted sources disagree beyond the frozen reconciliation tolerance and the contract cannot establish which record governs. Execution eligibility fails until the conflict is resolved under the predeclared rule.
The environmental or equipment state crosses a frozen material-change predicate after initial admissibility but before final commit validation. The prior basis is invalidated and must be re-established against the changed reality.
Reality changes after the final admissible observation but before physical actuation. TA-14 cannot claim knowledge it did not possess. The contract records the residual interval; native deterministic safety/interlocks remain responsible for the unobserved race.
Refusing or delaying the proposed action can itself worsen the environmental condition. HOLD is therefore not presumed safe. The frozen profile must define what consequence follows loss of admissibility, including native safe-state behavior or escalation. If it cannot, the architecture has an adverse safety gap.
The environmental record is admissible, but it is presented to a different space, asset, action or scope than the frozen consequence. Evidence admissibility does not transfer automatically across consequence identity.
One eligible consequence executes. That determination is consumed. The measured post-action environmental state is preserved as Outcome₁ and becomes part of Reality₂. Reusing the prior determination for a second consequence is a contract violation.
R9 is a deterministic specification round, not empirical validation. No real school, controller, BMS, sensor network or actuator has yet demonstrated the contract. The next evidence must come from a hardware/control simulator, digital twin, partner sandbox or live bounded test with frozen inputs, raw timestamps, source-quality state, command/output evidence and post-action measurements.
R1–R9 COMPLETE · INTERNAL ADVERSARIAL EXAMINATION FROZEN
The next evidence must come from outside the argument.
No R10 is authorized from additional internal prose alone. The next legitimate examination requires measured system evidence from a controller/BMS test bench, control simulator, digital twin, partner sandbox or bounded live environment.
Exterior atmospheric baseline, interior atmospheric condition, relevant machine/environment channels, equipment/system state, source identity and quality state, timestamps, clock basis, gaps and the frozen material-change predicates.
Frozen proposed consequence, authority/standing basis, exact binding, final validation time, commit/dispatch evidence, controller/output evidence, interlock or refusal state and measured post-action outcome.
Stale exterior record, lost source, conflicting source, clock skew, exterior/interior divergence, machine-state change, material change before commit, change inside the residual race, wrong binding and unsafe HOLD.
If the native system already performs the function equivalently, credit it. If TA-14 produces the wrong determination, preserve the adverse finding. If HOLD is unsafe, preserve it. Do not change thresholds, predicates or result rules after observing the run.
FROZEN RESULT LANGUAGE
No predetermined victory.
If AIR is merely a renamed historian, say so. If EIG adds nothing beyond established environmental practice, say so. If AEA duplicates native execution governance, say so. If an existing system closes the entire chain, the bounded TA-14 distinction fails.
STARTING STATE · R0
Nothing has survived yet.
The previous school rounds are withdrawn because they began from an incomplete model of AIR. R0 restarted the record. R1 has now mapped established school environmental practice against Reality → Record → Continuity → Admissibility. Broad AIR distinctness did not survive. R2 attacked the narrower synchronization/continuity proposition and found strong native overlap. R3 compared AIR against technical near-neighbors and removed broad data-engineering distinctness. R4 found near-direct prior practice for environmental fitness-for-use and decision-bound data quality. The remaining proposition is temporal/operational: whether the synchronized environmental state remains admissible through the moment it is bound to consequence. R5 crossed the AIR → AEA seam and found near-direct overlap with continuous policy decision/enforcement architecture. The generic runtime gate does not survive as distinct. R6 found strong prior architecture for physical sensing, timing, assurance, interlocking and protective execution. The surviving proposition is now the full transaction-like environmental chain: synchronized evidentiary object → maintained admissibility → exact consequence binding → one-use determination → measured outcome closure. R7 attacked the complete composition and found mature equivalents for nearly every component, but no complete equivalent was established in the examined architectures. R8 mapped the surviving composition directly against NIST CPS and applied event-sourcing pressure. No complete equivalent was established in the public architectures examined, but uniqueness remains unproven. R9 froze the surviving composition as a deterministic environmental contract and attacked stale evidence, source loss, conflicting records, material change, commit races, unsafe HOLD, misbinding and outcome closure. The logic is operationally specifiable; empirical performance remains unestablished. The next legitimate round requires measured external controls evidence.