CortistSuggest a forecast
CORTIST / THE PUBLIC FORECAST REGISTER / OCT 2026

Thirteen AI questions.
Different evidence stages.

When might AI reach AGI or superintelligence, change our jobs, threaten human control, master long tasks, design approved medicines or increase energy demand? Explore thirteen documented AI futures, including chips, science, the economy, governance, robotics and cybersecurity.

Explore the records ↓Read our protocol ↗

Thirteen documented research questions, not validated statistical predictions: eight provisional numerical drafts, three illustrative internal models and two questions without official percentages. None of those 13 topic drafts is locked or scored; the two separately registered 2027 milestone forecasts are editorially locked.

Evidence Pulse 2.0 checks the published Cortist News index when JavaScript is available. The source-linked, dated research articles remain accessible through Cortist News.

EVIDENCE PULSE · OCT 27 2026Dated editorial snapshot

New reporting is not a new probability.

All eight provisional numerical forecasts remain on their first recorded draft. Contextual news can be reviewed without silently changing published estimates.

Topic-by-topic evidence index ↗ · Full version history ↗
NEW / VERSIONED FORECAST RECORDS

When the evidence changes, the record stays.

Cortist now tracks the newest published estimate and every documented version for all thirteen topics. AGI, superintelligence, robotics, cybersecurity, chips, science, economy and governance each have separate, provisional numerical draft snapshots. Five other questions have no official numerical probability. Nothing is independently locked or scored yet.

Explore all thirteen version histories ↗Download versioned register ↗Read the evidence and scoring rules ↗
The historic v0.1 records remain unchanged. New assessments require separate timestamped versions and an explicit source-based rationale, not silent overwrites.
How to read this register: F-001 to F-004 are provisional 38%, 24%, 35% and 45% editorial probability drafts; A-001, M-001 and E-001 have illustrative 86%, 46% and 54% internal model calculations, but no official assigned probability; X-001 and J-001 have no numerical estimate. Zero independently locked or scored forecasts. Editorial standards ↗
SHORT HORIZON / 2027

Which AI milestones can we verify next year?

Five primary-source research questions with deadlines, prospective outcome rules and unresolved quality gates. Candidate definitions, not scored forecasts.

Explore five Milestones →
01 / THIRTEEN CONNECTED QUESTIONS

The questions that matter.

One evidence standard. Thirteen distinct subjects, time horizons and degrees of uncertainty.

F-005 / CHIPSPROVISIONAL
AI CHIPS & COMPUTE / FUTURE MILESTONE

How fast will AI computing improve 2030?

A fivefold gain in accelerator performance per dollar by 2030?

EDITORIAL DRAFTv0.1

Subjective probability, no independent calibration.

UNSCORED / 2030Review F-005 ↗
F-006 / SCIENCEPROVISIONAL
AI & SCIENCE / FUTURE MILESTONE

When will AI make a real scientific discovery 2035?

A novel, independently reproduced autonomous discovery by 2035?

EDITORIAL DRAFTv0.1

Subjective probability, no independent calibration.

UNSCORED / 2035Review F-006 ↗
F-007 / ECONOMYPROVISIONAL
AI & ECONOMY / FUTURE MILESTONE

When will AI measurably change economic growth 2035?

A sustained one-point US productivity boost attributable to AI by 2035?

EDITORIAL DRAFTv0.1

Subjective probability, no independent calibration.

UNSCORED / 2035Review F-007 ↗
F-008 / GOVERNANCEPROVISIONAL
AI & GOVERNANCE / FUTURE MILESTONE

Will governments agree on frontier AI rules 2032?

Three enforceable G20 frontier-AI safety regimes by 2032?

EDITORIAL DRAFTv0.1

Subjective probability, no independent calibration.

UNSCORED / 2032Review F-008 ↗
F-004 / CYBERSECURITYPROVISIONAL
CYBER RISK / CONTROLLED SECURITY TESTING

Autonomous cyber
agents by 2030?

Can an AI agent repeatedly pass independently evaluated, actively defended sandbox cyber ranges?

NEW EDITORIAL DRAFTModel v0.1

Active-defender criteria under review. Not locked or scored.

DEC 31, 2030 / UNSCOREDExplore F-004 ↗
F-003 / ROBOTICSPROVISIONAL
PHYSICAL AI / INDUSTRIAL AUTONOMY

Full-shift robot autonomy
by 2035?

Can a humanoid complete independently verified mixed eight-hour industrial shifts without teleoperation?

NEW EDITORIAL DRAFTModel v0.1

Criteria proposed; not calibrated, locked or scored.

DEC 31, 2035 / UNSCOREDExplore F-003 ↗
F-001 / AGIPROVISIONAL
CAPABILITY / GENERAL INTELLIGENCE

Will AGI arrive
by 2030?

Could AI satisfy an independently reviewed general intelligence test before 2031?

PUBLIC EDITORIAL ESTIMATEModel v0.1

Draft criteria. No independent lock or Brier score.

DEC 31, 2030 / UNSCOREDExplore F-001 ↗
F-002 / SUPERINTELLIGENCEPROVISIONAL
CAPABILITY / BEYOND HUMAN EXPERTS

Superintelligence
by 2040?

Could AI exceed expert human abilities across a broad range of domains?

PUBLIC EDITORIAL ESTIMATEModel v0.1

Draft criteria. No independent lock or Brier score.

DEC 31, 2040 / UNSCOREDExplore F-002 ↗
X-001 / AI EXISTENTIAL RISKRESEARCH DRAFT
IMPACT / HUMANITY'S FUTURE

Could AI cause
human extinction?

Understand extinction and loss-of-control scenarios, expert disagreement and missing evidence.

AI EXTINCTION PROBABILITYNot assigned

The 2100 research horizon is not a predicted extinction date.

2100 / OPEN QUESTION / UNSCOREDExplore risk & years ↗
J-001 / AI & JOBSRESEARCH DRAFT
EMPLOYMENT / THE FUTURE OF WORK

Will AI replace
human jobs?

Which tasks, careers and early-career opportunities could change by 2030, and what can labor data verify?

AI JOB DISPLACEMENT PROBABILITYNot assigned

J-001 is a research candidate, not an independently reviewed numerical forecast.

2030 / RESEARCH QUESTION / UNSCOREDExplore jobs & years ↗
02 / THREE NEW MEASURABLE MILESTONES

Capability. Medicine.
Infrastructure.

Three source-backed research candidates with precise deadlines, transparent working models and open review requirements. No official Cortist probabilities have been assigned.

A-001 / AI AGENTS / CAPABILITYRESEARCH DRAFT

Can agents finish
8-hour tasks?

80% reliability on independently evaluated long-form technical tasks - not an entire working day.

8h80% success / by 2028
Cortist probability: not assigned
UNSCOREDRead the forecast analysis ↗
M-001 / AI IN MEDICINE / DRUG APPROVALRESEARCH DRAFT

An AI-designed drug
approved by the FDA?

One therapeutically new drug approval with independently verified AI design of the active compound.

01FDA approval / by 2032
Cortist probability: not assigned
UNSCOREDRead the forecast analysis ↗
E-001 / AI & ENERGY / POWERRESEARCH DRAFT

Will data centers use
12% of US electricity?

Measured US electricity used by all data centers in 2030; not AI-only demand.

12%US electricity share / 2030
Cortist probability: not assigned
UNSCOREDRead the forecast analysis ↗
New / inspect the probability working models: Cortist Calibration Lab ↗ shows version 0.2 assumption-model outputs of 86% for agents, 46% for medicine and 54% for energy - not independently validated forecasts. The earlier 85% and 45% displays were corrected for ordinary whole-percent rounding; version 0.1 remains separately archived. These are not official Cortist forecasts, not independently calibrated probabilities, and not scorable.
Methodology review: 3 HOLD / 0 independent signoffs. Read all review blockers ↗ The public register remains two numerical editorial drafts and five exploratory research questions.
How to read all seven: AGI and superintelligence carry provisional numerical editorial estimates, but remain unlocked and unscored. The other five are exploratory questions. The new candidate pages explain editorial working models, not locked or independently validated Cortist predictions.

How to read this board: 38% and 24% are provisional editorial estimates for their respective deadlines, not independently validated results. The circular graphics display the published estimates, not observed progress. X-001, J-001, A-001, M-001 and E-001 have no assigned probabilities and cannot currently be scored. The year 2100 is a proposed research horizon, not a predicted extinction date.

02 / ACCOUNTABILITY BEFORE ACCURACY

What makes a forecast count?

A strong record needs more than a confident number. It needs defensible definitions, independent review and a fair resolution.

01 DEFINE

Write the question.

Define the event, a deadline and measurable evidence thresholds before evaluation.

AGI / ASI criteria: draft
02 REVIEW & LOCK

Commit in advance.

Independently review the terms and anchor the prospective record before an outcome.

Independent lock: pending
03 RESOLVE & SCORE

Return to evidence.

Compare the outcome with original rules and score only valid resolved forecasts.

Resolved and scored: 0
Forecast Protocol v1.0 ↗Independent review dossier ↗Public scorecard ↗JSON forecast manifest ↗
03 / INTEGRITY OF THE RECORD

No retroactive certainty.

Dated F-001 and F-002 v0.1 records and SHA-256 checksums are publicly available. Independent reviews, external time anchoring and prospective locks are not yet complete. We will not present provisional drafts as scored achievements.

The X-001 risk question has unresolved issues around attributing human extinction to AI and verifying the outcome. No numerical probability is assigned.

CONTRIBUTE / YOUR QUESTION

Help us refine
the evidence.

Cortist now concentrates on seven questions. Help us test their assumptions, definitions and source quality.

Suggest a forecast ↗
SEVEN TOPIC EXPLAINERS

Explore the evidence
behind each question.

Forecasts are precise events. Topics explain the research, history and related news.

TOPIC GUIDE

AGI

When could artificial general intelligence be demonstrated?

TOPIC GUIDE

Superintelligence

What would exceed broad human expert capabilities?

TOPIC GUIDE

AI Existential Risk

Understand extinction and loss-of-control risks.

TOPIC GUIDE

AI & Jobs

Evidence on automation, hiring and future skills.

TOPIC GUIDE

AI Agents

Autonomy, long tasks, reliable tools and oversight.

TOPIC GUIDE

AI in Medicine

Drug discovery, clinical evidence and FDA approvals.

TOPIC GUIDE

AI & Energy

Data centers, electricity and grid constraints.