Cortist can revise a forecast when evidence changes. The old prediction is never silently rewritten. This register shows what was published, when it was recorded, whether it can be scored, and what still needs independent verification.
Will independently verified AGI arrive by 31 December 2030?
First machine-readable capture of the already published provisional 2030 probability; not a prospectively locked version.
Read unchanged versioned snapshot ↗Current published numerical record: 2026-10-22. One documented v0.1 draft and no newer revision. Scoring ineligible.
Will independently verified ASI arrive by 31 December 2040?
First published machine-readable provisional ASI estimate; exact independence and superiority thresholds are still proposals.
Read unchanged versioned snapshot ↗Current published numerical record: 2026-10-22. One documented v0.1 draft and no newer revision. Scoring ineligible.
An independently auditable, spending-weighted AI accelerator performance-per-dollar measure is at least five times higher in Q4 2030 than Q4 2025, in constant 2025 US dollars.
Original subjective editorial scenario, not independently locked or scored. Resolution rules need external review.
Read original versioned snapshot ↗One provisional numerical version. No scored or independently resolved outcomes.
At least one novel natural-science discovery originating predominantly from an autonomous AI research system is reproduced by an independent team by the end of 2035.
Original subjective editorial scenario, not independently locked or scored. Resolution rules need external review.
Read original versioned snapshot ↗One provisional numerical version. No scored or independently resolved outcomes.
At least two independent macroeconomic studies attribute an additional one percentage point or more of annual US labor-productivity growth to AI over the same three-year observation period ending by 2035.
Original subjective editorial scenario, not independently locked or scored. Resolution rules need external review.
Read original versioned snapshot ↗One provisional numerical version. No scored or independently resolved outcomes.
By the end of 2032 at least three distinct G20 jurisdictions enforce binding predeployment safety evaluations and mandatory serious-incident reporting for frontier general-purpose AI models under public oversight.
Original subjective editorial scenario, not independently locked or scored. Resolution rules need external review.
Read original versioned snapshot ↗One provisional numerical version. No scored or independently resolved outcomes.
Will an independent evaluator verify an AI agent completing complex scenarios on three distinct isolated and actively defended cyber ranges at 80% success by 2030?
Initial subjective editorial scenario for controlled defensive-evaluation settings. Evaluation rules have not been independently locked or scored.
Read unchanged first draft ↗No independently adjudicated outcome and no past numerical revisions.
Will a humanoid robot complete independently verified, repeated eight-hour industrial work shifts without teleoperation by 2035?
Initial subjective editorial scenario; proposed full-shift success, no-teleoperation and independent verification rules are not locked or scored.
Read original F-003 draft snapshot ↗New question on 26 October 2026. No earlier version or qualifying independently scored outcome.
Will AI cause human extinction by 31 December 2100?
No published numerical versions. The research question is not an official probability forecast.
Evidence gathering and criteria development continue. Working-model numbers, if any, are not adopted official forecasts.
Will agents reach an eight-hour technical task horizon at 80% success by 31 December 2028?
No published numerical versions. The research question is not an official probability forecast.
Evidence gathering and criteria development continue. Working-model numbers, if any, are not adopted official forecasts.
Will early-career AI-exposed hiring diverge from a preregistered comparison group by 2030?
No published numerical versions. The research question is not an official probability forecast.
Evidence gathering and criteria development continue. Working-model numbers, if any, are not adopted official forecasts.
Will a verifiably AI-designed novel drug receive FDA approval by 31 December 2032?
No published numerical versions. The research question is not an official probability forecast.
Evidence gathering and criteria development continue. Working-model numbers, if any, are not adopted official forecasts.
Will data centers account for at least 12% of US electricity use in calendar 2030?
No published numerical versions. The research question is not an official probability forecast.
Evidence gathering and criteria development continue. Working-model numbers, if any, are not adopted official forecasts.
Record primary sources, changed assumptions, independent replications, contradictory findings and the exact forecast question. Expert timelines do not automatically become percentages.
Record the former and proposed probabilities, UTC issue time, written rationale and source URLs. Publish a new snapshot path instead of editing the previous file.
Keep the earlier snapshot and update the current-pointer register. Only independently preregistered, approved and locked forecasts become eligible for outcome scoring.
Each public numeric draft has a versioned JSON record and a SHA-256 digest. An R2 copy is retained under operator-managed controls, but same-site checksums and R2 retention are not independent immutable timestamps. Independent reviewers and external timestamp verification remain outstanding.
Current thirteen-topic register (JSON) · Previous eight-topic register (historical) · Original seven-topic register (historical) · Current register pointer · Update policy and release safeguards · Existing two-forecast manifest · R2 archive disclosure · Independent review requirements · Public scorecard.
Yes. A future estimate must be published as a new dated version with evidence and an explanation. The October 22, 2026 v0.1 38% draft remains visible and unchanged.
No. It is a provisional editorial estimate with unapproved resolution criteria. It is not a prospectively locked, independently reviewed or scored forecast.
No. Eight topics have clearly labeled provisional, unscored numerical drafts. The other five remain research questions, including three with nonofficial illustrative scenarios.
No. Cortist maintains separate versioned paths and an R2 retention archive, but administrators can change controls and an independent external timestamp has not been established.
