Governance and scientific boundary

Uncertainty is preserved, not compressed into confidence

Evidence labels describe what a claim rests on. They are never averaged, scored, or converted into a confidence percentage, because doing so would hide exactly the information a reviewer needs.

Institutional governance

Authority is separated from review

Program leadership

Defines remit, approves inputs, and owns delivery. Program leadership cannot independently elevate its own findings.

Scientific advisory review

Domain-qualified reviewers challenge methods, interpretations, and evidence labels before release or gate decisions.

Governance authority

Controls conflicts, disclosure, dual-use handling, publication status, and the recorded basis for institutional decisions.

Research standards

Methods before conclusions

Required practice

Traceable sources and preserved raw observations
Explicit separation of observation, inference, and hypothesis
Controls, nulls, deviations, and contradictory evidence reported
Named review ownership and reproducible decision criteria

Conflict-of-interest policy

Financial, professional, and intellectual interests declared before review
Conflicted reviewers recused from evidence elevation and publication approval
Commissioning relationships disclosed with published work
Unresolved conflicts recorded as a limitation, never omitted

Publication methodology

Every public output states what kind of work it represents

Work class	Meaning	Permitted claim
Internal research	Institute-directed method or research development.	Reports methods and findings at their documented evidence state.
Commissioned analysis	Work scoped and funded for an external organization.	Decision support for the stated scope; not independent validation.
Independently validated science	Findings reproduced by qualified parties independent of the originating team and commission.	Validation only within the reproduced protocol, controls, and stated limits.

Evidence model

Eleven descriptive labels

Evidence labels

Document-Supported

Meaning: A statement traceable to a supplied document, paper, patent, or notebook entry.
Minimum evidence: At least one identified source with locatable passage and provenance record.
Permitted use: Describing prior art, stated methods, and claimed conditions.
Prohibited interpretation: Treating a documented claim as a measured or reproduced result.
Required reviewer: Principal Investigator or delegated research analyst.
Upgrade path: Attach primary data and calibration records to reach Data-Supported.
Downgrade / quarantine trigger: Source withdrawn, unverifiable provenance, or contradicting primary record.

Operating rules

01Never convert uncertainty into false certainty.
02Never claim extraordinary breakthroughs without replication-grade evidence.
03Preserve accountable human review at every gate.
04Separate scientific proof from commercialization potential.
05Separate observed data from interpretation.
06Preserve null and contradictory results in the record.
07Maintain provenance and audit history for every claim.
08Keep dual-use analysis at governance-safe levels of detail.
09Quarantine unsupported or compromised claims rather than deleting them.
10State limitations in every executive output.

Decision outcomes

Where an evidence state can legitimately lead

Recommendations are subject to authorized human judgement. The map shows routes, not conclusions.

Decision branch

Proceed

Operational rationale: Evidence supports continued investigation at the current scope.
Minimum evidence condition: Data-Supported with an accepted residual-risk statement.
Primary decision owner: Lab Director
Required next artifact: Next-best-experiment queue
Reversal condition: A new artifact class is found capable of reproducing the effect.
Governance note: Proceeding is not endorsement of any interpretation of the effect.

Accountability

Human responsibility is not delegable

Always required of a human

Approving which materials enter the engagement
Signing every phase gate and evidence-state elevation
Deciding to proceed, pause, replicate, or stop
Accepting or rejecting a red-team verdict
Authorizing external publication or transfer

Never asserted by the framework

That a phenomenon is real, novel, or mechanistically explained
That a result is validated in the absence of independent replication
That a modelled result is an observation
That regulatory, export-control, or legal requirements are satisfied
That an absence of detected artifacts proves an effect

Responsible use

Dual-use and disclosure posture

Artifacts are screened for misuse exposure at a level of detail that does not itself provide uplift.
Sensitive technical specifics are summarized by effect, never by reproducible procedure, in public-facing material.
Contradictions and null results are retained in the register and reported, never deleted.
Quarantined claims remain visible in audit history with the reason and the reviewer recorded.
This site requests no classified, export-controlled, proprietary, or personally identifiable material.

Frontier Sciences governance

Controls for governed lab-incubator engagements

Scientific evidence labels

Every consequential claim carries a descriptive evidence state; observations, interpretations, models, hypotheses, speculation, contradictions, and quarantine remain distinct.

Human review requirements

Qualified reviewers approve inputs, methods, phase gates, evidence-state changes, replication claims, releases, and consequential decisions.

Approved-input access

Only authorized, provenance-bearing materials enter scope. Access is non-intrusive by default and production write access requires explicit authorization.

Audit and provenance

Source identity, custody, transformations, assumptions, reviewer actions, dissent, and disposition remain traceable.

Research confidentiality

Handling requirements are established before material transfer; the public intake accepts non-sensitive scoping information only.

Simulation boundaries

Every model declares assumptions, excluded conditions, validity limits, sensitivity, and the evidence required for physical validation.

Anomaly and artifact handling

Unexplained observations remain unresolved until conventional mechanisms, artifacts, calibration, software, environment, and bias are attacked.

Replication requirements

Advancement claims require a reviewable packet containing methods, data lineage, controls, nulls, uncertainty, code or logic, and independent reproduction criteria.

Biological safety

Analytical support never authorizes biological operations; biosafety, ethics, clinical, and regulatory authorities retain control.

Dual-use awareness

Inputs and outputs are screened for misuse, disclosure, export-control, safety, and security exposure where relevant.

Commercialization separation

Funding interest, intellectual-property potential, market demand, and commercialization readiness do not elevate scientific evidence.

Decision authority

KRYOS v6 supplies structured decision support. Authorized human leaders retain every binding scientific, safety, clinical, regulatory, operational, and investment decision.

Visible policy register

Controls that apply across every Institute program

Scientific integrity

Observations, inferences, hypotheses, contradictions, nulls, and deviations remain distinguishable and traceable.

Evidence classification

Every consequential claim states what it rests on; labels are descriptive and never converted into a confidence score.

Research independence

Commissioning, delivery, challenge, and evidence-elevation authority are separated and disclosed.

Publication corrections

Material corrections are dated, explained, linked to the prior wording, and never silently overwritten.

Responsible disclosure

Release scope is reviewed against safety, privacy, security, contractual, and dual-use obligations.

Data governance

Only approved material enters an engagement; provenance, access, retention, and disposition remain recorded.

Framework attribution

Original named frameworks, KRYOS integration, established methods, Institute workflows, and client implementations remain distinct.

AI-assisted analysis

Any material AI assistance is disclosed with model limits; it cannot approve evidence or make binding decisions.

Scientific claims

No output independently establishes validity, novelty, mechanism, safety, readiness, or successful replication.

Source provenance

Claims retain source locators, acquisition context, ownership, review history, and unresolved gaps.

Commercialization separation

Scientific evidence state is never elevated by funding interest, IP value, market demand, or partnership attention.

Human review

Every evidence elevation, release, phase gate, and consequential decision retains a named accountable owner.