From evidence workflow to trusted adoption.

Evidence needed before the next release stage

November 1, 2026: these are acceptance requirements, not completed results. Phase 2 must remain disabled until verified evidence is recorded.

1. Real outcomes: use consenting researcher projects. Retain raw measurements, methods, exclusions, source versions, outcome amendments and provenance. Verify save, reopen, export and restore plus owner/reviewer isolation. Exclude fixtures, fictional examples and unknown origins. The existing domain threshold is 10 eligible outcomes across 3 investigations; meeting a count alone does not establish validity.
2. One complete public investigation: name the consenting lead and independent reviewer, document qualifications and conflicts, preserve original plan, data, analysis, objection, response and changed next test. Obtain exact-content publication consent. Publish limitations and reproducible materials alongside the result. No lead or reviewer is appointed by this page.
3. First-use verification: observe a consenting first-time researcher completing Question → Outcome without coaching. Record abandonment, confusion, time and help needed; navigation clicks are not completed research.
4. Independent security assessment: agree assessor, scope, target build and authorization; use synthetic accounts. Require findings, remediation and independent retest, plus recovery and retention evidence before sensitive-work claims. Developer tests do not satisfy this requirement.

Prepare the researcher pilot · Independent assessment scope

1. Interoperability

• Implemented: public Zotero user/group Web API metadata preview and explicit sync for collections of up to 60 top-level items. Updates to existing metadata remain pending; researcher notes and review history survive. No API keys, private-library OAuth, PDFs or Zotero writes.
• Existing: paired Zotero Desktop metadata pathway, CSV reference intake, Markdown/JSON handoffs.
• Prepared: private adoption candidate from current project fields, ORCID checksum validation, optional OSF target and checksum-wrapped export. No account verification or deposit occurs.
• Requires configuration: Mendeley/OAuth developer access, authenticated ORCID account linking, OSF deposit authorization and a repository/registrar agreement for DOI issuance. No paid integration is enabled.

2. Living dependencies and review

• Existing: explicit claim/decision dependencies, permission-checked linked decision views, on-demand publisher notice checks and attributed reassessments.
• Existing: isolated owner projects, selected reviewer snapshots, append-only reviewer feedback, revocation, private backup/recovery and exact snapshot exports.
• Not established: unattended worldwide monitoring, email notifications, complete dataset-version coverage, cross-institutional identity/qualification verification, blind-review confidentiality agreements or E2EE. No global coverage claim.

Retractions and corrections request reassessment, not automatic invalidation. Missing notices do not establish validity. Blindness requires review of text and metadata; account separation alone does not establish independence.

3. Real pilot and trust

The Type I society mission should be tested through a bounded energy decision, such as useful heat recovery after 24 hours under matched initial energy and ambient conditions. No consenting team or actual measurements have been established here.

1. Select a lab with a real recurring decision, an accountable lead, accessible sources, permissions, manageable confidentiality and availability for follow-up.
2. Freeze an evaluation protocol and reference set before grading. Compare total work and evidence quality with the team’s usual tools.
3. Preserve original plan, inspected sources, substantive challenge, response, revision, unresolved issues and actual outcomes.
4. Record failed attempts, support effort, repeat use and actual purchasing decisions separately from interest.
5. Publish only consented artifacts, with review scope and limitations. Fictional examples remain tutorials, not proof.

Ownership, publication and source protection

The ASTRA application source remains private. The portable decision schema is separately published under a schema-only MIT license. No public application-source mirror or researcher-data release is authorized; application code and record contents are outside the schema license. A separate public data-model repository is an optional future decision requiring explicit rights review and permission.

Realistic delivery gates

These are planning windows, not promised completion dates. Do not mark a gate passed merely because time elapsed or a JSON package exists.

Connect references → · Prepare the adoption handoff → · Deliverables and support →