Trust and editorial method
Useful AI guidance should show its work.
AI Amigos separates authorship, historical provenance, current testing, editorial review, independent review, and production proof so a visitor can see exactly what has and has not been verified.
People and review ownership
Vijay Bhoyar directs the project and is the named author of the current working evidence specifications. The AI Amigos Editorial Desk remains an organizational byline for source-preserved legacy material and internal protocol checks. Neither role is described as independent domain review. Outside review requires a named reviewer, qualification, exact version, scope, conflict disclosure, and independently controlled verification.
Publication gate
- Define one distinct job to be done and intended audience.
- Use at least two authoritative external sources; prefer primary evidence.
- Provide actionable steps, a worked example, limitations, and a practical artifact.
- Name the author, reviewer, review date, next review date, and substantive changes.
- Require explicit human approval before editorial publication.
Evidence tiers
- Independent
- Issued and controlled by an outside organization, publisher, reviewer, or adopter.
- Jointly verified
- Prepared with AI Amigos and explicitly verified by an outside participant.
- Author controlled
- Published by AI Amigos; useful for provenance but not independent recognition.
Outcome and benchmark gate
- A private project records the baseline, experiment versions, test cases, human decisions, and follow-up outcome.
- Only deliberately sanitized fields can enter a proof-pack submission.
- A reviewer checks privacy, claim scope, evidence lineage, and limitations.
- Approved public records retain versions and correction history.
- Benchmark cohorts remain suppressed below 10 reviewed reports, below three independent organizations, or when one organization supplies more than half the cohort.
Review intervals
- 90 days: vendors, products, careers, technical guides, and working methods.
- 180 days: evergreen foundations.
- Gated review: healthcare, finance, safety, credentials, legal, or other high-stakes claims.
Status language
- Source preserved
- The historical route and date remain, but the page has not passed the evidence gate. It is noindex and excluded from discovery.
- Editorial protocol
- Internal structure, privacy boundary, sources, limitations, and dates passed editorial review. It is not an outcome claim.
- Reviewed outcome
- A sanitized run report and follow-up measurement passed the stated review gate.
- Benchmark
- A multi-organization cohort passed sample, comparability, privacy, concentration, methodology, and review gates.
Corrections and disclosure
Send corrections, source suggestions, review proposals, or adoption verification through contact actions. Material corrections remain in the correction registry. AI may help prepare drafts and checks; a person must approve publication. Funding, affiliate, vendor, reviewer, or participant relationships must be disclosed on the affected record.
Public change log
- 2026-08-25: Published the defined endeavor, three author-controlled working methods, evidence tiers, public claim ledger, empty independent-evidence registries, named project authorship, multi-organization benchmark gates, evidence schemas, and private counsel-data boundary.
- 2026-08-24: Released the verified practice network, local project dashboard, four editorial reference protocols, benchmark suppression, field challenges, and legacy-guide quarantine.
- 2026-08-23: Added the local-first workbench, seven-stage proof loop, proof-pack boundary, and optional account architecture.
- 2026-08-22: Added four tracks, five local-only tools, twenty templates, and inactive-by-default external integrations.