Live Versus Planned
As of: September 3, 2026 (the latest decision recorded in this register) Release posture: v0.1 Public Draft Official numeric scoring: Disabled
This is the authoritative claims register for current capability. It distinguishes source that exists in the repository from services that are deployed or operational.
Status vocabulary
| Status | Meaning |
|---|---|
| Available in source | The artifact exists in this workspace; it may still be unreleased and unhosted. |
| Experimental synthetic | Executable behavior is limited to synthetic fixtures and cannot issue an official result. |
| Proposed | The design or policy is documented but not operational. |
| Controlled, not public | A capability may exist later behind approved access; none is claimed by this entry. |
| Blocked | Work must not activate until named gates are met. |
| Not authorized | Publication, deployment, or external action has not been approved. |
“Available in source” does not mean deployed at lockedinlabs.ai, externally published, independently reviewed, legally cleared, secure for real data, or supported as a service.
Public foundation
| Capability | Status | What is true now | What is not claimed |
|---|---|---|---|
| Public specification | Available in source | A normative v0.1 draft defines construct and requirements. | Industry-standard, validated, certified, or independently governed status. |
| Benchmark card | Available in source | Intended use, exclusions, limitations, and validation status are documented. | Fitness for hiring or production evaluation. |
| Core ontology | Available in source | Ten domains and 55 capability families are versioned in a public foundation. | Complete competencies, 300–500 signals, calibrated rubrics, or field validity. |
| Public schemas | Available in source | JSON Schema 2020-12 contracts and synthetic fixtures define initial boundaries. | Production API stability or operational-platform conformance. |
| Scoring architecture | Available in source | Deterministic, fixed-point, versioned principles and activation gates are documented. | Official weights, bands, thresholds, percentile, role readiness, or numeric publication. |
| Reference scoring | Experimental synthetic | Reference calculations may operate on labeled synthetic fixtures only. | A real-person result, official score, or calibrated measure. |
| Evaluator foundation | Experimental synthetic | Provider-neutral typed interfaces, deterministic mock behavior, fail-closed publication recommendation, and synthetic governance tests are present in a separate source foundation. | A production AI evaluator, model validation, provider superiority, or score authority. |
| Public site | Available in source | The static public draft is deployed at https://fdebenchmark.org (canonical since 2026-09-03; benchmark.lockedinlabs.ai redirects permanently); its benchmark homepage, 10-domain/55-family explorer, research program, publication desk, method paper, and sanitized source reader are prerendered and indexable, score-disabled, and accept no real evidence or identity data. |
An approved benchmark release, operational assessment service, production data intake, or validated result. |
| Profile anatomy preview | Experimental synthetic | The site shows a schema-shaped, non-person composition with no identity or result. | A candidate, founder, employee, customer, score, profile, or credential. |
| Public-draft indexing | Live public draft | The canonical public site permits indexing following the September 3, 2026 founder decision; pages declare canonical URLs and a sitemap. | Guaranteed search inclusion, placement, or third-party endorsement. |
| Official benchmark release | Not authorized | The public-draft site and learning resources are hosted; formal benchmark release preparation is underway. | Official benchmark release, scoring, verification, or intake. |
| Independence and institutional firewall | Available in source | A staged I0–I4 governance, role, data, funding, publication, and structural-entity activation boundary is documented. | A seated Commission, operational independence, role separation, official result, credential, certification, or accreditation. |
| Three-layer outcome architecture | Available in source | Proposal documents, strict synthetic-only schemas, fixtures, and adversarial tests separate a Verified Production Record, Capability Profile, and Governed Credential Tier candidate. | A real-person record, validated instrument, tier decision, credential, issuer, or cross-layer authority. |
| Shared Work Challenge Battery | Available in source | A proposal and result-free machine contract define a common scenario substrate for human-only, model-only, and human-with-AI research conditions. | An administered battery, human baseline, result, human-versus-model comparison, shared score, leaderboard, or employment interpretation. |
| Method-first research dimensions | Proposed | Demonstrated Capability profile, Time to Production, Evidence Strength, Recency, confidence/coverage, and domain packs are defined as research candidates. | Active indices, validated comparisons, official values, or a global human score. |
| Phase 2 trust/register proposal | Available in source | Isolated proposal docs, strict schemas, synthetic fixtures, and adversarial tests define scoreless T0–T4 trust progression, separate operational readiness and result lifecycle, 12 observation axes, bounded integrity routing, refresh research, and aggregate disclosure controls. | An active trust service, issued result, verified person, score, credential, restriction, register, or API. |
| Research/publications hub | Available in source | The hosted publication desk and sanitized reader include the method paper, source ledger, field notes, market brief, publication standard, evolution protocol, credibility evidence register, research agenda, chart contract, and proposal records. | Published studies, peer review, field findings, independent reproduction, or production analytics. |
| Aggregate data explorer | Experimental synthetic | The site contains an abstract synthetic metro-participation and suppression preview with generated metro-like labels and coarse count/share bands. | Real data, exact public counts, fine ratios, geographic aptitude, public individual records, statistical sufficiency, a privacy guarantee, or named-person ranking. |
Operational benchmark
| Capability | Status | Activation condition |
|---|---|---|
| Candidate accounts and identity verification | Blocked | Private platform, privacy notice, identity policy, tenant isolation, and security review. |
| T0–T4 trust-progression and readiness service | Blocked | The proposed trust progression, separate operational readiness, and result-lifecycle reference exist in source; identity, consent, verification, refresh, review, appeal, privacy, security, measurement, and legal gates must be operationally approved. |
| Consent and evidence inventory | Blocked | Operational consent, data rights, retention, deletion, audit, and authorization. |
| Public repository connector | Blocked | Approved OAuth scope, threat model, provenance, rate, deletion, and provider controls. |
| Private repository or artifact analysis | Blocked | Sandboxed no-network execution, encryption, scanning, prompt-injection defense, provider policy, and verified deletion. |
| Deployment and outcome verification | Proposed | Verification protocol, reviewer training, reference consent, conflicts, and audit workflow. |
| Adaptive live evaluation | Proposed | Rubrics, accommodations, controlled cases, evaluator validation, and human review. |
| Human review and adjudication | Proposed | Named trained reviewers, case access, conflict routing, quality monitoring, and appeal. |
| Official deterministic scoring | Blocked | A later benchmark release, independence stage I3, an external-majority Commission decision, and every gate in SCORING.md. |
| Verified Production Record | Blocked | Approved factual-verification protocol, identity and authorization boundary, claim-level corroboration, conflicts, correction, dispute, privacy, security, and audit. It cannot confer capability. |
| Capability Profile | Blocked | Job analysis, content blueprint, calibrated multidimensional instrument, eligible evidence, non-conflicted review, uncertainty, fairness/accessibility, correction, and appeal. It cannot confer a credential. |
| Governed Credential Tier | Blocked | Standard setting, classification-error and consequence evidence, external-majority Commission authorization, non-conflicted issuer, key/status lifecycle, privacy, appeal, revocation, interoperability, and trademark approval. |
| Enterprise workforce analytics | Proposed | Validated intended use, human governance, job relevance, fairness, legal terms, and privacy controls. |
| VPC or on-premises evaluator | Proposed | Security architecture, signed summaries, local-provider controls, and operational support. |
| Third-party evidence adapters | Proposed | Neutral public contract, equivalent-provider treatment, consent, and no scoring preference. |
| Training gap-plan integration | Proposed | Read-only separation, consent, and prohibition on training-based score credit. |
| Named-person leaderboard | Not authorized | Public ordered ranking of people is prohibited. |
| Global human score | Not authorized in v0.1 | A future scalar may be researched only as an optional purpose-bound projection under a new release and the full scalar authorization, measurement, consequence, independence, and monitoring gates. It may not replace the profile or create a named-person leaderboard. |
| Cost-per-person axis | Not authorized | Compensation-efficiency or “value for money” ranking is outside the benchmark. |
Governance and operations
| Capability | Status | Current fact |
|---|---|---|
| Founding stewardship | Available in source | LockedIn Labs is identified as founding steward. |
| FDE Benchmark Commission | Proposed | An external-majority protected decision body is defined; no member, chair, independence determination, meeting, funding authority, or decision is claimed. |
| Technical Steering Committee | Proposed | No seated roster is claimed. |
| Evaluation Integrity Committee | Proposed | No seated roster or controlled-case operation is claimed. |
| Measurement and Fairness Advisory Group | Proposed | No appointed advisors or independent review is claimed. |
| Domain working groups | Proposed | No outside organization is represented as a member. |
| Appeals panel | Blocked | No official result exists; no private filing channel or panel is operational. |
| Public contribution workflow | Blocked | CI, exact-target public-draft deployment, live-monitor source, dependency update configuration, a pull-request review template, operative license files, and Developer Certificate of Origin attestation are available in source; the repository has a GitHub remote. Branch/environment protection, public repository visibility, managed issue intake, and private conduct reporting are not configured. |
| Security reporting | Available in source | A private reporting channel exists: the contact form at /contact (intent "Report a security issue"), referenced from /.well-known/security.txt, delivering by email to a monitored LockedIn Labs mailbox. A named security owner and published response targets are still required. |
| Conduct reporting | Available in source | A private channel exists: the contact form at /contact (intent "Raise a conduct or conflict concern"). A non-conflicted response team is still required before the channel can be described as independent. |
| Sponsor program | Proposed | No sponsor, funding tier, benefit, or relationship is represented as active. |
| Partner or signatory program | Proposed | No external contributor, partner, signatory, validator, or endorser is claimed. |
| Training/employment/staffing-to-assessment firewall | Proposed | Role, data, compensation, review, issuance, appeal, and publication separation is specified; no real-person control test is complete. |
| Structural-entity decision | Blocked | No hosted, affiliate, independent-entity, or external-certification path has been approved; separate incorporation is not claimed as universally required. |
| Trademark clearance | Blocked | Name and mark clearance and final policy require legal review. |
| Public content/code licenses | Available in source | LICENSE (Apache-2.0 for software, schemas, fixtures, scripts, and the public site), LICENSE-CONTENT.md (CC BY 4.0 for specification, policy, ontology, and research content), and NOTICE are published. The private evaluator and platform repositories carry all-rights-reserved license files. Trademark permission remains separate and uncleared pending counsel. |
Research and validation
| Capability | Status | Current fact |
|---|---|---|
| Research/standards source ledger | Available in source | A dated ledger records reviewed sources, design influence, departures, legal monitoring, and known gaps. It is not a conformance mapping. |
| Publication, evolution, and credibility controls | Available in source | Proposed records define publication classes, reproducibility packages, semantic evolution, corrections, evidence states, and four distinct go-live meanings. They expose blockers rather than authorize a release. |
| Phase 2 research agenda | Available in source | Eleven proposed workstreams define questions and activation gates for role analysis, trust comprehension, verification, integrity, referrals, adaptive probes, measurement, freshness, aggregate privacy, review, and reproducibility. No study is authorized. |
| Standards conformance mapping | Proposed | Clause-level mapping, permissions, implementation evidence, tests, and independent approval do not exist. |
| Market landscape | Available in source | A dated, non-exhaustive product scan documents why the project must not claim to be first or only. It is not a vendor rating. |
| Calibration cohort | Proposed | No participant data or result is represented as collected. |
| Inter-rater reliability | Proposed | No coefficient or threshold result exists. |
| AI-human agreement | Proposed | No measured agreement or production evaluator exists. |
| Fairness and subgroup analysis | Proposed | No fairness claim or representative subgroup study exists. |
| Accessibility validation | Proposed | Static and rendered keyboard, reflow, contrast, semantic, and automated axe checks exist in source; assistive-technology, participant, and conformance review remain required. |
| Criterion or predictive validity | Proposed | No longitudinal field-outcome finding exists. |
| Percentiles or comparison population | Blocked | No representative normative population exists. |
| Formal certification/accreditation | Not authorized | The project is a benchmark draft, not a certification scheme. |
| Demonstrated Capability profile | Proposed | No released proficiency values or calibrated domain comparisons exist. |
| Time to Production | Proposed | A research candidate only; event definitions, confounders, normalization, and validity are unresolved. |
| Evidence Strength | Proposed | The evidence-support concept exists; no calibrated public index is active. |
| Recency | Proposed | Evidence dates may be represented; no universal decay or comparative index is active. |
| Domain-pack research | Proposed | Potential packs require independent methodology, evidence rules, and validation. |
| Grader invariants | Experimental synthetic | Synthetic known-good, insufficient, invalid, repeat, and review-gate fixtures and tests are present; no controlled grader is approved. |
| Learnability/answerability review | Proposed | No live scenario is represented as having passed this gate. |
| Failure and correction publications | Proposed | The policy favors publishing system failures safely; no operational study result exists. |
| Jurisdiction-specific regulatory review | Blocked | Qualified counsel, not this repository, must determine current applicability and obligations for each use and region. |
| Adjacent AI-role benchmarks | Blocked | Expansion may begin only after FDE validation and a separate role-specific construct decision. |
Claims permitted now
The source may accurately be described as:
- the LockedIn FDE Benchmark v0.1 Public Draft;
- a proposed, evidence-oriented framework for demonstrated FDE capability;
- founded and initially stewarded by LockedIn Labs;
- designed around open methodology, controlled evaluation integrity, typed AI observations, deterministic scoring, and human governance;
- implemented with synthetic public ontology, schema, reference, and explanatory-site foundations where the corresponding source is present;
- planning a method-first research platform around multidimensional capability profiles, Evidence Strength, Recency, confidence/coverage, domain packs, and a Time to Production research candidate.
- containing an isolated Phase 2 source proposal for scoreless T0–T4 trust progression, separate operational readiness and result lifecycle, a 12-axis profile anatomy, prestige-neutral verification, bounded integrity routing, human recourse, and privacy-preserving aggregate participation.
Claims not permitted now
Do not describe the project as:
- an industry standard, accredited certification, or independently validated benchmark;
- scientifically proven, unbiased, predictive, immutable, first, only, best, or definitive;
- operationally field validated or capable of independent verification;
- ready for private-IP intake, secure repository analysis, verified profiles, official scores, or employment decisions;
- endorsed, governed, sponsored, signed, or validated by an outside organization;
- deployed at a production domain unless deployment is separately verified and this register is updated;
- operating official indices, a data explorer, published research findings, or a validated Time to Production measure;
- compliant with any named law or jurisdiction absent a documented decision by qualified counsel.
Update discipline
Any capability-status change requires:
- evidence that the named gate is met;
- owner and governance approval appropriate to the risk;
- updated benchmark card or policy where claims change;
- a changelog entry;
- exact wording that distinguishes source availability, controlled operation, and public availability.
If this register conflicts with marketing copy or a roadmap date, this register controls the current-state claim.