Change the record, not the standard.

2026-10-15 / Existing-record claim review

Reviewed the cited public sources for all 27 pre-existing records and added dated, source-scoped claim-review notes across the 29-record registry. Recorded 15 additional vendor-reported human-comparison or validation claims, including Minds, under the same standard used for the two new submissions.

Reports were added for Aaru, BluePill, Decisions Lab, Electric Twin, Evidenza, Fairgen, Inqvey, Lakmoos, Minds, PersonaPanels, Replism, Simile, Synthetic Users, Uxia, and Vurvey Labs. No established evidence check was upgraded. Metrics remain attached to their own task and source, not compared as a universal accuracy score. This source-claim review does not re-certify existing artifact or method-review checks.

For the other records, the reviewed sources did not supply a specific product human-reference result. Human training data, general accuracy language, studies of other systems, and a benchmark interface were not promoted into validation claims. This is a scoped search outcome, not proof of global absence.

Replaced Vurvey’s two unavailable citations with its current platform, People Models, and Lab pages, retaining the prior URLs in the review metadata. Its current public description is reflected in the profile, with direct human validation labeled vendor reported and its autonomous model identified as private preview. Minds’ issued v0.1 dossier remains historical and links to the current profile for later source claims.

Earlier same-day expansion / v0.5

Added Ask Your Audience (AYA) in synthetic respondents and Synthetic Consumer Lab in synthetic research workflows after founder submissions and review of the public first-party sources linked on each profile.

Both records document product claims and disclosed limitations. Their public human-comparison claims are labeled vendor reported, with source links, while method review, independent reproduction, and human-reference validation remain not established. No established-validation count increased.

The matrix now says documented for claim attribution and distinguishes vendor reports from established checks and missing reviewed evidence. This clarifies the existing standard without upgrading a vendor claim into validation. Only claims attributable to the reviewed public sources were included.

2026-07-21 / Registry expansion v0.4

Added Simile from its public platform description and a contemporaneous independent Gallup research announcement. The record is classified in population simulation and remains source-reviewed; no new independent-validation, reproduction, or human-reference evidence checks were marked.

Gallup's announced evaluation is still underway, so it is recorded as context rather than as a result. This is a new source record, not an endorsement or comparative result.

Earlier same-day expansion / v0.3

Added Beehive AI Synthetic Personas, BluePill AI, Decisions Lab, Fairgen, GWI Synthetic Audiences, Inqvey, MiroFish, and Uxia under the same source-review standard. MiroFish has an additional recorded artifact check because its source code is publicly available; that is not a validation result.

Earlier same-day expansion / v0.2

Added Evidenza, Electric Twin, Epistemix, PersonaPanels, and Vurvey Labs under the same source-review standard.

Correction rights without editorial control.

1. 01

   Scope

   Anyone may submit a factual correction, a primary source, an accessible research artifact, or a disclosure of a conflict or limitation.

2. 02

   Review

   Editors assess the supplied evidence against the published methodology. A vendor response is evidence to review, not a reason to accept a claim.

3. 03

   Notice

   When practical, an affected organization receives a correction window before a material factual change. Silence is never treated as agreement or validation.

4. 04

   Record

   Accepted material changes receive a dated entry describing the affected record, what changed, why, and the source. Editorial rationale is preserved.

5. 05

   Boundary

   Submitting a correction does not grant editorial control, improve an evidence check, create an endorsement, or change the funding disclosure.

Landscape changes remain auditable.

A primary workflow lane is a navigational classification based on public descriptions. A system can have many methods, but receives one lane so the landscape does not imply that human-participant tools, synthetic respondents, population simulations, and benchmarks are interchangeable.

See the Synthetic Research & Simulation Landscape → and the full classification protocol →.