Authority begins where
the claim stops.
The public record behind the LockedIn FDE Benchmark: who owns it, what can be cited, what evidence exists, and what remains unproven.
The method exists.
The validation program does not—yet.
These are evidence strands, not a progress score. An empty strand cannot be converted into confidence by averaging the others.
Construct and content
A proposed ontology and prospective job-analysis protocol exist; no admitted job-analysis or expert-panel evidence supports the content blueprint.
Response process
No candidate, reviewer, or reference-provider response-process study is complete.
Internal structure and reliability
No operational instrument, inter-rater study, generalizability analysis, or score precision estimate is published.
Relations to outcomes
No concurrent, predictive, convergent, or discriminant evidence supports a hiring or performance interpretation.
Fairness, accessibility, and consequences
No participant-facing accessibility, subgroup, adverse-impact, burden, or consequential-use study is complete.
Computational reproducibility
One fixed synthetic scorer input has an executable expected-output check. It is not a reproduction of a human finding.
No invisible review board.
The organization is named. The people who must approve authorship, methods, measurement, access, and release are not. We show those vacancies because a blank role is not approval.
Open the authorship recordA citation resolves the draft.
It does not upgrade the evidence.
Until an immutable archive is approved, cite the exact version and describe it as a public method and construct draft.
LockedIn Labs Research. (2026). LockedIn FDE Benchmark v0.1—Public method and construct draft (Version 0.1.0-draft).
Open the records behind the claim.
Every record states what is adopted now, what is proposed, and what remains blocked.
Withdrawal of the staffing funnel
What was served at /find-an-fde, the window the deployment evidence bounds, what was removed, and the guard that prevents recurrence.
Institutional independence firewall
Public Draft — not operational. Proposed external-majority oversight, role separation, protected decisions, and activation gates.
Publication standard
Release classes, evidence receipts, reproducibility, limitations, corrections, and go-live gates.
Credibility evidence register
The state and receipt required for every material benchmark-authority claim.
Job-analysis preregistration candidate
Questions, intended use, unresolved design decisions, and the gates before recruitment.
Synthetic reproducibility demonstration
A frozen synthetic input, expected output, runner, direct-file manifest, and explicit limits.
Research source ledger
Primary sources, access dates, adopted controls, non-adopted boundaries, and claim limits.
Benchmark evolution protocol
Versions, linking studies, monitoring, corrections, suspension, migration, and retirement.
Publication infrastructure is visible before it is complete.
There are no empirical publications and therefore no empirical correction or retraction records. That is not evidence that the draft is error-free. Policy exists; monitored reporting routes, preservation, rights, and release authority remain open.
Available for inspection, critique, synthetic implementation testing, and study planning. Not authorized for official scoring, verification, certification, ranking, employment decisions, or real-data intake.
