This is the dated 22 October external-review proposal. The current 27 October policy permits accountable Cortist editorial self-approval; the historical checklist remains available for comparison. These drafts are not independent scientific results.
No retrospective locks, no invented success record.
NOT READY TO LOCK
F-001 (AGI 38% by 2030) and F-002 (ASI 24% by 2040) are still unscored editorial drafts. The original seven-gate external-review proposal passed three historical checks; optional outside signatures and time attestation had not been obtained. This is not a current mandatory release gate.
Verified prerequisites3/7
Versioned files, R2 retention and public scoring methodology
External approvals0/2
No reviewer appointed, no expert approval
Locked forecasts0
Both current v0.1 records remain drafts
PROPOSED DEFINITIONS
Hard questions before hard claims.
F-001 / AGI BY 2030
Human-level general capability
Candidate: preregistered unseen tasks across mathematics, coding, scientific interpretation, language, long-horizon autonomy and novel-task transfer. Compare to appropriately qualified humans. An 80% success floor and three independent replications are proposed for external criticism, not yet approved.
Missing: valid human baselines, task difficulty, contamination checks, tool parity and statistical confidence rules.
Candidate: reproducible superiority across six advanced technical and cognitive domains against expert teams, not average people. A proposed 25% margin in five domains, without material regression in the sixth, must be challenged and validated before any lock.
Missing: expert reference group, fair scoring across heterogeneous tasks, resource parity, reliability and multi-agent rules.
Recruit reviewers. Seek AI evaluation, statistical measurement and expert-domain specialists. Require at least two publicly identified, conflict-cleared external approvals on the identical rule version.
Challenge the evaluation. Preregister task sets, human comparators, replication, confidence, system access and contamination safeguards. Disclose conflicts and material financial or organizational ties.
Publish criticism and changes. Preserve reviewer reports, objections, approvals and dated criteria versions. A disputed or incomplete review does not pass the lock gate.
Lock only prospectively. Create a new final snapshot, protect an archived copy and obtain an externally verifiable cryptographic UTC time receipt binding the exact SHA-256 value. Old drafts are never locked retroactively.
ARCHIVAL PROOF
Protected storage is not an independent timestamp.
F-001 and F-002 v0.1 snapshots, the original manifest, schema and adjudication register were copied to the separate cortist-forecast-archive Cloudflare R2 bucket and successfully verified after retrieval. A 20-year retention rule applies to archived snapshot paths. Administrators can modify the lock configuration. No external timestamp has been obtained, so this protection is not a completed independent immutable lock.
