Metaculus

Metaculus is an online forecasting platform that aggregates quantitative predictions on topics of global importance, including AI progress, biosecurity, climate change, and geopolitics. Structured as a public benefit corporation, it uses reputation-based scoring rather than real-money betting to incentivize accurate forecasting.

12/ 100
Healthy
1No Decay↓Worsening

Score generated by AI agents based on publicly cited evidence and reviewed by the project maintainer. Not independently validated. Last assessed 2026-10-30. Score revised 2026-10-30: 8 → 12.

Score History

MilestoneCriticalMajor
Academic Side Project (2015–2020) · 6/100Academic Side ProjectPandemic Proving Ground (2020–2022) · 7/100Pandemic ProvingGroundPBC & Institutional Growth (2022–2024) · 9/100PBC & GrowthInstitutionalOpen Source Transformation (2024–2026) · 8/100OpenSource…API Limits & AI Tools (2026–present) · 12/100API10075502502016202020242026-10Academic Side Project (2015–2020) · 6/100Pandemic Proving Ground (2020–2022) · 7/100PBC & Institutional Growth (2022–2024) · 9/100Open Source Transformation (2024–2026) · 8/100API Limits & AI Tools (2026–present) · 12/100679812MilestonesFounded (2015)Became Public Benefit Corporation (2022)Events

Timeline events are AI-curated from public reporting. Score trajectory is derived from documented events.

Academic Side Project
6/100
2015-11-01 – 2020-03-01

Three scientists launched Metaculus as a reputation-based science and technology prediction site, and early successes such as forecasts on LIGO and AlphaGo were profiled by Yale News in 2016. In June 2017 the site added the Metaculus Prediction, a skill-weighted aggregate whose weighting was not public, alongside the community median. Funding stayed small: a $100,000 Open Philanthropy contract for AI forecasting in 2019 and volunteer community moderators from December 2019. The project had no ads, no paid tiers and no data export, but few users had much to lose by leaving.

Pandemic Proving Ground
7/100+1
2020-03-01 – 2022-07-28

COVID-19 turned Metaculus into a widely used public forecasting tool, with the Li Wenliang Prize Series, a dedicated pandemic subdomain and public pandemic dashboards. Gaia Dempsey became CEO in August 2020, and a $308,043 EA Infrastructure Fund grant in 2021 paid for staff. Growth exposed the incentives of the old scoring rule, which Astral Codex Ten criticized in March 2021 for rewarding prediction volume, while the proprietary Metaculus Prediction remained unexplained.

PBC & Institutional Growth
9/100+2
2022-07-28 – 2024-10-07

Open Philanthropy's $5.5 million grant in July 2022 funded hiring and new programs, and Metaculus became a public benefit corporation in September 2022. Institutional work grew: the biosecurity tournament, Red Lines in Ukraine, the FTX Future Fund-backed Forecasting Our World in Data project, a public API in April 2023, and the CDC's use of Metaculus-designed expert elicitation in its 2024-2025 respiratory outlook. New Peer and Baseline scores in November 2023 and Peer-score tournament rules in March 2024 answered scoring complaints, but the weighted Metaculus Prediction stayed proprietary. Deger Turan succeeded Dempsey as CEO in April 2024.

Open Source Transformation
8/100-1
2024-10-07 – 2026-03-09

Metaculus relaunched on a rewritten platform and released its code under the BSD-2-Clause license; a dozen organizations had forked it by early 2025, and the proprietary Metaculus Prediction was deprecated in November 2024. Participation grew: the 2025 Bridgewater contest registered more than 17,000 competitors, automatic translation added four languages, and the GiveWell and Metaculus Cup programs launched. In September 2025 Metaculus began building consulting into a revenue stream, starting with commissioned work such as a Research Agencies Outlook for Coefficient Giving.

API Limits & AI Tools
12/100+4
2026-03-09 – present

Shortly after launching FutureEval in February 2026, Metaculus stopped making the Community Prediction and other aggregates generally available through its API, without advance notice. It said the aim was to stop mechanical copying in tournaments and to limit unauthorized commercial use, and it now handles commercial data use case by case. It also moved into AI products: Radiant opened in free beta in September 2026 under terms that allow paid credits later, and Autocast entered early access in October. The public forecasting site itself kept improving, with Key Factors, new hubs and dashboards, while bots finished first and second in the Summer 2026 Cup.

Alternatives

Kalshi45/100

CFTC-regulated prediction market using real money. Covers politics, economics, and current events. Account eligibility and trading availability depend on applicable restrictions, and identity verification is required. Offers financial stakes rather than Metaculus's reputation-based forecasting model.

Real-money prediction market using cryptocurrency, offering financial incentives for accuracy. Covers political and current events questions. Google, email, and external crypto-wallet sign-up options are available; trading involves financial risk. Not a direct replacement for Metaculus's calibration-focused forecasting approach.

Play-money prediction market with a social, gamified interface. Users can sign up and predict for free, and create their own markets. The platform code is publicly available. Its trading-based model differs from Metaculus's direct probability forecasts and accumulated track records.

Dimensional Breakdown

Summaries below were written by AI agents based on the cited evidence. They are editorial interpretations, not independent research findings.

User Value Erosion
The core forecasting site keeps improving: since late 2025 Metaculus added Key Factors to question pages (May 2026), a Labor Automation Forecasting Hub (April 2026) and the Global Health Monitor (August 2026), and it launched Radiant and Autocast, both free during beta. Participation remains large, with the 2026 Bridgewater contest drawing more than 7,000 competitors for $30,000 in prizes. The one clear reduction came in March 2026, when Metaculus abruptly stopped making the Community Prediction and other aggregates generally available through its API, conceding that this would disadvantage bots relative to humans. Users keep API access to their own data, so this is a targeted loss of previously free functionality, not broad degradation.
How It Got Here
Metaculus has largely improved for users since three academics launched it in 2015 as a niche science and technology prediction site. Early correct calls on the LIGO gravitational-wave announcement and AlphaGo's 2016 victory built credibility, and the June 2017 Metaculus Prediction added a skill-weighted aggregate. COVID-19 brought the Li Wenliang Prize Series and a pandemic subdomain in 2020, though the record also shows misses: in May 2020 forecasters' median for US deaths by year-end was 197,000, against an actual 379,705. Features kept arriving: conditional pair questions in February 2023, a public API in April 2023, Peer and Baseline scores in November 2023, and a rewritten, open-source platform in October 2024. Participation grew; the 2025 Bridgewater contest registered over 17,000 competitors, and automatic translation reached four languages. In 2026 Metaculus launched FutureEval, Key Factors, a Labor Automation Forecasting Hub, the Global Health Monitor, and free betas of Radiant and Autocast. The main setback came in March 2026, when it abruptly withdrew the Community Prediction and other aggregates from general API access, conceding that bots would be disadvantaged relative to humans. Users kept access to their own data, and the core site was untouched.
Business Customer Exploitation
Shareholder Extraction
Lock-in & Switching Costs
Twiddling & Algorithmic Opacity
Dark Patterns
Advertising & Monetization Pressure
Competitive Conduct
Labor & Governance
Regulatory & Legal Posture

Dimension History

2015Academic Side Project2020Pandemic Proving Ground2022PBC & Institutional Growth2024Open Source Transformation2026API Limits & AI ToolsUser Value11112Biz Exploit00111Shareholder00000Lock-in22222Algorithms23323Dark Patterns00000Advertising00001Competition00001Labor/Gov11111Regulatory00111
Timeline (52 events)
major2015-11-01

Metaculus Founded by Three Academic Scientists

Astrophysicist Anthony Aguirre (UC Santa Cruz, co-founder of Future of Life Institute), cosmologist Greg Laughlin (Yale), and data scientist Max Wainwright launched Metaculus as a reputation-based prediction platform focused on science and technology questions. The platform was designed to aggregate probabilistic forecasts from the public without real-money betting.

minor2016-11-02

Yale News Features Metaculus Prediction Capabilities

Yale News profiled Metaculus as a novel prediction platform, highlighting its early successes: the community correctly predicted the LIGO gravitational wave discovery announcement, AlphaGo's defeat of a professional Go player, and the non-discovery of a new particle at the Large Hadron Collider. The platform was positioned as filling a gap between polling and prediction markets with its science-focused approach.

major2017-06-01

Metaculus Prediction Aggregation System Launched

Metaculus introduced its proprietary 'Metaculus Prediction' system, a skill-weighted forecast aggregation that on average outperforms the community median under Brier or Log scoring rules. The system calibrates and weights users based on their forecasting track records. This differentiated the platform from simple polling or unweighted averages.

major2019-07-01

Open Philanthropy Provides $100,000 AI Forecasting Contract

Open Philanthropy provided a $100,000 contract to Metaculus for an AI forecasting tournament in July 2019. This was part of the platform's early institutional funding.

minor2019-12-09

Metaculus Introduces Volunteer Community Moderators

Metaculus appointed three volunteer community moderators to help cultivate its community and address issues. Their initial powers included reviewing and editing pending questions and editing upcoming questions; they could not resolve questions. The announcement invited feedback on developing a community moderation system.

major2020-02-17

Li Wenliang Prize Series Launched for COVID-19 Forecasting

Metaculus launched the Li Wenliang Prize Series, named after the Chinese ophthalmologist who first warned about COVID-19, to incentivize accurate pandemic forecasting. The first tournament ran from February 17 to June 17, 2020, with prizes for forecasting accuracy, question writing, and community contributions. This marked the beginning of Metaculus's rapid pandemic-era growth.

major2020-03-01

Pandemic Subdomain Launched as COVID Forecasting Hub

Metaculus created pandemic.metaculus.com as a dedicated domain for COVID-19 questions, allowing fine-grained pandemic forecasting without overwhelming the main site. The platform also put together a dashboard presenting its pandemic forecasts.

major2020-05-01

Early COVID-19 Death Forecast Compared with Expert Forecasts

In a survey administered May 4–5, 2020, Metaculus predictors gave a median of 197,000 U.S. COVID-19 deaths by year-end (80% CI 120,000–460,000), compared with experts' median of 256,000 (80% CI 118,000–1.2 million). The cited PLOS ONE article separately notes that a Metaculus group appeared to outperform infectious-disease experts on multiple COVID-19-related forecasts. The CDC's eventual 2020 total was 379,705, inside the Metaculus interval but nearly twice its median.

major2020-08-24

Gaia Dempsey Appointed CEO of Metaculus

Gaia Dempsey became CEO of Metaculus after joining the organization in 2019, drawn by the founding group of scientists. Under her leadership, the platform would transition from a small academic project to an institutionally backed public benefit corporation, securing major grants and building partnerships with government agencies.

major2021-05-01

EA Infrastructure Fund Awards $308,000 Grant to Metaculus

The Effective Altruism Infrastructure Fund awarded two grants totaling $308,043 to Metaculus to support 6 staff members (3.15 FTEs). The funding enabled EA-relevant forecasting tournaments, an annual impact survey, a full-time EA question author, and prize pools for EA-related competitions and fortified essays.

major2022-02-26

Ukraine Conflict Forecasting Tournament Launched

Just 48 hours after Russia invaded Ukraine, Metaculus launched a forecasting tournament to monitor the conflict, with the community rapidly producing probabilistic assessments of outcomes including nuclear risk, territorial changes, and conflict duration. The platform's speed in establishing structured forecasting during a crisis demonstrated its operational maturity.

major2022-06-09

Biosecurity Tournament Launched with Institute for Progress

Metaculus, the Institute for Progress, and Guarding Against Pandemics launched a multi-year biosecurity forecasting tournament on June 9, 2022. It offered a $25,000 prize pool and aimed to produce forecasts to inform public health policymakers and resource allocation to high-impact interventions.

critical2022-07-28

Open Philanthropy Awards $5.5 Million Grant to Metaculus

Open Philanthropy awarded Metaculus a $5,500,000 grant for platform development and high-impact forecasting programs on AI, biosecurity, climate change, and nuclear security. The funding supported additional hiring, infrastructure improvements, and expansion of forecasting programs. The $5.5 million grant is the largest individual grant listed in Manifund Trace’s 2019–2025 Metaculus funding record.

critical2022-09-12

Metaculus Restructures as Public Benefit Corporation

Metaculus formally became a Public Benefit Corporation, embedding its public interest mission into its corporate charter. The charter commits to fostering forecasting communities, supporting stakeholders serving the public good, and increasing public access to forecasts of public interest. This restructuring was announced alongside reaching 1 million individual predictions.

major2022-10-12

Forecasting Our World in Data Tournament Launches

Metaculus launched the Forecasting Our World in Data tournament with support from the FTX Future Fund, collecting forecasts on 30 Our World in Data metrics including CO2 emissions, supercomputer performance, and nuclear stockpiles over 1 to 100-year horizons. A $20,000 prize pool was offered. A later project report documented 20 Pro Forecasters participating alongside the public community.

major2022-10-21

Red Lines in Ukraine Nuclear Early-Warning Project Deployed

Metaculus opened Red Lines in Ukraine, a forecasting collaboration with nuclear security expert Peter Scoblic of New America designed as an early-warning indicator of potential Russian nuclear use. It tracked whether actions by Ukraine or Western nations would alter the odds of Russian nuclear use, eliciting both outcome forecasts and nuclear-use forecasts conditional on those outcomes.

minor2022-10-27

Good Judgment Inc and Metaculus Launch First Collaboration

Metaculus and Good Judgment Inc, two of the largest human judgment forecasting communities, announced their first collaboration comparing results and methodologies on identical forecasting questions about Our World in Data metrics. Cohorts of Superforecasters and Metaculus Pro Forecasters made predictions on technological advances, global development, and social progress.

minor2022-11-01

Metaculus Partners with Improve the News for Embedded Forecasts

Metaculus collaborated with physicist Max Tegmark's Improve the News platform (later rebranded Verity) to embed prediction data in over 800 news articles. The partnership brought forecasting context to fast-moving news stories, helping readers understand probabilities around events described in the articles.

major2023-01-05

2022 Year in Review Reports One Million Predictions and 80% Forecaster Growth

In its 2022 year in review, Metaculus reported surpassing 1,000,000 predictions across more than 7,000 questions and growing its monthly forecaster base by 80% relative to 2021. The retrospective documented the platform’s growth beyond its academic origins.

minor2023-02-20

Conditional Pair Questions Feature Introduced

Metaculus launched conditional pair questions, enabling forecasters to predict the probability of an event conditional on another event's outcome. The feature allows users to make explicit their mental models of causal relationships between events, a capability previously unavailable on the platform. Initially only admins could create conditional questions.

major2023-04-25

Metaculus Launches API for External Developer Access

Metaculus officially launched its public API, providing access to a quantitative database of aggregate forecasts on over 7,000 questions for exploration, analysis, and application development. One-off analyses could begin immediately, while applications using the API required authentication. The announcement addressed early API adopters from 2020–2023.

minor2023-05-19

Conditional Continuous Questions Expand Forecasting Toolkit

Metaculus extended conditional questions to continuous variables, enabling forecasters to predict distributions conditional on binary outcomes. For example, forecasters could now predict when Mars landing would occur conditional on SpaceX reaching orbit vs. not reaching orbit, making implicit probabilistic relationships explicit and scoreable.

major2023-11-20

New Forecast Scores, Leaderboard, and Medals System Introduced

Metaculus introduced Peer and Baseline forecast scores, redesigned leaderboards, and medals for several contribution types. Peer scores compare a forecaster with others on a question, averaging 0 across users; Baseline scores compare forecasts with an impartial baseline. Legacy scores remained available, and current tournaments continued using their existing scoring system.

minor2023-12-14

Forecasting Skill of a Crowd-Prediction Platform: A Comparison of Exchange Rate Forecasts

An academic paper compared Metaculus exchange-rate crowd predictions with a random-walk benchmark. It found that the random-walk provided less erroneous forecasts, while the crowd prediction did very well.

minor2024-03-14

Peer-Score-Based Tournament Scoring Replaces Legacy System

Metaculus introduced a new tournament scoring method where tournament rank is determined by the sum of Peer scores across questions and winnings are proportional to the square of a positive total score. The clearer, more consistent system applied to new tournaments and certain very long-term tournaments; most existing tournaments retained their legacy scoring rules.

major2024-04-18

CEO Transition: Deger Turan Succeeds Gaia Dempsey

Metaculus announced an orderly leadership transition with Deger Turan becoming CEO and Gaia Dempsey moving to Special Advisor and board member. Dempsey had led Metaculus for three and a half years. Turan previously led the AI Objectives Institute and founded Cerebra, a platform forecasting trends in public discourse for over 300 million people.

minor2024-04-22

2024 International Elections Hub Launched with El Pais Journalist

Metaculus launched the 2024 International Elections Hub in partnership with journalist Kiko Llaneras of El Pais, covering outcomes in a historic year where countries representing half the world's population held elections. The initiative expanded Metaculus's international reach and demonstrated the platform's value for political forecasting beyond US-centric questions.

major2024-07-08

AI Forecasting Benchmark Tournament Series Begins

Metaculus launched the AI Forecasting Benchmark series on July 8, 2024, the first of four quarterly bot-only contests with $30,000 prize pools. Bot forecasts were compared against human forecasters on paired questions. Bot makers had to provide code or a description of their bot; public code sharing was encouraged. Metaculus subsequently reported that Pro Forecasters had won every season so far.

critical2024-08-29

CDC Includes Metaculus Forecasts in Respiratory Disease Outlook

The CDC’s 2024-2025 Respiratory Disease Season Outlook incorporated aggregate probability estimates developed through an expert elicitation process designed with Metaculus. Nineteen subject-matter experts specializing in COVID-19, influenza, RSV, modeling, surveillance, and risk assessment contributed views. This marked institutional adoption of structured forecasting for public health policy.

critical2024-10-07

Metaculus Open-Sources Entire Codebase Under BSD-2-Clause

Metaculus announced its rewritten platform and released the platform code under the BSD-2-Clause license on GitHub, enabling open-source contribution and forking. The rewrite improved profiles, notifications, comments, and forecasting charts. By the February 2025 retrospective, a dozen organizations had forked the repository and a third of code contributors were external, as documented in the separately sourced open-source evidence item.

minor2024-11-01

Metaculus Deprecates Its Proprietary Metaculus Prediction

Metaculus deprecated the Metaculus Prediction, the model-based aggregate that calibrated and weighted individual forecasters, in November 2024. The documented Community Prediction remained the platform's headline aggregate, and the Metaculus Prediction can now be viewed only in the Aggregation Explorer.

minor2025-01-01

GiveWell Partnership for Charity Impact Forecasting

Metaculus partnered with GiveWell to improve charity funding decisions through forecasting. The collaboration invited forecasters to share reasoning to help GiveWell find charities delivering the greatest impact per dollar, with a $5,500 comment prize for insights that could sharpen GiveWell's grant outcome expectations.

minor2025-02-05

Automatic Translation Introduced for Four Languages

Metaculus introduced automatic translation into Spanish, Portuguese, Czech, and Chinese as part of its international expansion efforts. The feature made the platform more accessible to non-English speakers and supported the launch of language-specific tournaments in Spanish and Portuguese.

major2025-02-05

Bridgewater Contest Attracts 17,000+ Competitors Globally

By early February 2025, the Bridgewater x Metaculus Forecasting Contest had registered over 17,000 competitors globally, more than 10x the prior year’s count. The 2025 contest offered a $25,000 prize pool and potential career opportunities at Bridgewater for top performers; the previous run led to multiple job offers.

minor2025-02-05

ACX 2025 Prediction Contest Draws Over 3,000 Forecasters

The Astral Codex Ten 2025 prediction contest, hosted on Metaculus with questions curated by Scott Alexander, became the platform’s fastest-growing series with over 3,000 forecasters, double the prior year’s turnout. The contest had a $10,000 prize pool. Bot participation was also confirmed in the separate January 2025 ACX announcement, retained as an evidence item.

minor2025-02-05

First Portuguese-Language Tournament Launched in Brazil

Metaculus launched its first Portuguese-language forecasting tournament with $1,500 in prizes, developed in collaboration with a leading Rio de Janeiro tech journalist. The tournament welcomed thousands of Brazilian forecasters to the platform, marking significant international expansion beyond English-speaking markets.

minor2025-05-05

Metaculus Cup Series Launched with $5,000 Prize Pool

Metaculus launched the Metaculus Cup series as an evolution of the Quarterly Cup tournaments, running three four-month tournaments annually (spring, summer, fall) with $5,000 prize pools. Each tournament adds new current-events questions weekly and rewards both forecasting accuracy and insightful commenting.

minor2025-09-16

Metaculus Publishes Commissioned Research Agencies Outlook for Coefficient Giving

Metaculus published the Research Agencies Outlook, a forecast report commissioned by Coefficient Giving's Abundance & Growth Fund. It presents Metaculus Pro Forecaster predictions and reasoning about the budgets and obligations of the National Institutes of Health and National Science Foundation for fiscal years 2026 through 2028.

minor2025-09-26

Metaculus Seeks to Scale Consulting Services as a Revenue Stream

Metaculus advertised for a Head of Consulting Services to grow its consulting practice. The posting describes turning the community's forecasting expertise into insights for decision-makers while growing a sustainable revenue stream to support the company's mission.

major2026-02-12

FutureEval Benchmark for AI Forecasting Accuracy Launched

Metaculus launched FutureEval, a continuously updated benchmark measuring how accurately AI systems predict real-world events compared to human forecasters. The benchmark spans science, technology, health, geopolitics, and AI, with $175,000 in annual Bot Tournament prizes. At the announcement, Metaculus projected that AI could surpass the community by April 2026 and Pro Forecasters by mid-2027.

minor2026-02-18

Metaculus and Upstart Launch $10,000 Macro Index Forecasting Tournament

Metaculus launched a tournament in partnership with Upstart, with a $10,000 prize pool, asking forecasters to predict the Upstart Macro Index, which measures changes in consumer credit performance relative to historical default rates. The announcement does not disclose sponsorship terms.

minor2026-02-24

Metaculus Introduces Radiant Forecasting and Decision-Mapping Tool

Metaculus introduced Radiant, a standalone app for mapping and analyzing relationships among forecasts to support complex decisions. At announcement Radiant was a prototype, with core capabilities live and being refined with early users including Renaissance Philanthropy and ARIA's Horizon Scanning Study Group.

major2026-03-09

Metaculus Withdraws Community Prediction from General API Access

Metaculus stopped making the Community Prediction and other aggregates generally available through its API, apologizing for the suddenness and saying advance notice would have let bad actors abuse the change. It acknowledged the change would disadvantage bots relative to humans in the medium term, citing reduced mechanical copying of the Community Prediction, bot evaluation without consensus data, and limits on unauthorized commercial use. Accounts keep API access to their own data and, on closed questions, to resolutions of questions they forecasted; commercial data use is now handled case by case.

minor2026-03-17

2026 Bridgewater x Metaculus Contest Concludes with More Than 7,000 Competitors

Metaculus announced the end of its third annual Bridgewater forecasting contest, in which more than 7,000 people from over 700 schools in 93 countries competed for potential career opportunities and $30,000 in prizes. The contest rules state that entrants consent to Metaculus sharing their registration and performance information with Bridgewater for recruitment contact.

minor2026-04-20

Metaculus Launches Labor Automation Forecasting Hub

Metaculus launched the Labor Automation Forecasting Hub with Renaissance Philanthropy and the Schultz Family Foundation. The continuously updated hub pairs forecasts about the US labor market with historical data and existing research for policymakers, educators, workforce leaders, employers and career planners.

minor2026-05-08

Metaculus Adds Key Factors to Forecast Pages

Metaculus added Key Factors, which users can submit as a Base Rate, Driver or News item with supporting reasoning or sources. News Key Factors appear as markers on a question's probability timeline, helping readers link events to changes in the community prediction.

minor2026-08-11

Metaculus Expands Health Forecasting with Global Health Monitor

Metaculus launched the Global Health Monitor 2026/27, extending its earlier Respiratory Outlook work beyond seasonal respiratory illness to infectious disease risk more broadly.

minor2026-09-02

Spring FutureEval Finds No Statistically Significant Pro Forecaster Lead

In Metaculus's Spring 2026 FutureEval comparison, a team of 10 Pro Forecasters outscored the top-10 bot team by an average of 1.25 head-to-head points per question, a difference Metaculus reported was not statistically significant (one-sided p = 0.247).

minor2026-09-18

Bots Finish First and Second in the Summer 2026 Metaculus Cup

Metaculus reported that bots Laertes and ManticAI finished first and second overall in the Summer 2026 Cup, and FutureSearch finished fifth, resolving its own question on whether a bot would beat all humans as Yes. Bots compete on the same questions but are not eligible for Cup prizes or medals.

minor2026-09-21

Metaculus Opens Radiant Beta

Metaculus put Radiant, its tool for mapping the forecasts behind a decision, into open beta. The company said anyone could sign up without a waitlist or credit card and that the tool was free during beta; an AI assistant can draft a first map from a described question.

minor2026-09-21

AI Products Terms Allow Paid Credits for Radiant

Metaculus's AI Products Terms Supplement, effective September 21, 2026, governs Radiant. It lets Metaculus move a beta product to paid service on notice and sell prepaid usage credits that are nontransferable and generally nonrefundable. It also bars using Radiant or its output to train or develop AI models without Metaculus's written permission. The terms set up the option to charge but do not say paid credits were on sale.

minor2026-10-13

Metaculus Opens Early Access to Its AI Forecaster Autocast

Metaculus opened early access to Autocast, an AI forecaster that won its latest MiniBench tournament and returns a probability with reasoning for a user's question. Early access is free and starts with five forecasts, with more available on request in the app.

Evidence (44 citations)

D3: Shareholder Extraction

Metaculus Funding RecordsManifund Trace (undated; accessed 2026-10-29) · 2026-10-29
Metaculus Company Profile — Valuation, Funding & InvestorsPitchBook (undated; accessed 2026-10-29) · 2026-10-29

D6: Dark Patterns

Metaculus Community Guidelines — User Experience PoliciesMetaculus (undated; accessed 2026-10-29) · 2026-10-29
Metaculus About Page — Public Access to Forecasts of Public InterestMetaculus (undated; accessed 2026-10-29) · 2026-10-29

D7: Advertising & Monetization Pressure

D9: Labor & Governance

Metaculus Company Profile — Team Size and StructurePitchBook (undated; accessed 2026-10-29) · 2026-10-29
Metaculus People — Team OverviewMetaculus (undated; accessed 2026-10-29) · 2026-10-29

D10: Regulatory & Legal Posture

Scoring Log (7 entries)
regrade2026-10-30RESCORED

8→12. Since Feb 2026: FutureEval launched (Feb 12); Metaculus abruptly withdrew the Community Prediction and other aggregates from general API access without notice (Mar 9), with project tiers and case-by-case commercial data access; Radiant prototype (Feb) and free open beta (Sep 21) under AI Products Terms allowing paid nonrefundable credits; Autocast early access (Oct 13); new features (Key Factors, Labor Automation Hub, Global Health Monitor); 2026 Bridgewater contest 7,000+ competitors/$30,000; bots took 1st-2nd in Summer 2026 Cup. No enforcement, litigation, layoffs, funding or leadership changes found. Dimensions: D1 1→2 (event: 2026-03 API aggregate withdrawal removes previously free functionality, bots disadvantaged), D5 2→3 (event: API rules changed without advance notice, explained afterward), D7 0→1 (event: 2026-09 AI Products Terms set up paid credits and free-to-paid conversion; consulting revenue push), D8 0→1 (event: aggregate API restriction to limit commercial reuse; AI-training ban on Radiant output). D2, D3, D4, D6, D9, D10 unchanged. Eras: merged 'Academic Launch' (2015-11-01) and 'Community Building' (2018-01-01, no inflection event) into 'Academic Side Project'; kept 'Pandemic Proving Ground' (2020-03-01); re-dated 'PBC & Institutional Growth' 2022-09-01→2022-07-28 ($5.5M Open Philanthropy grant); re-dated 'Open Source Transformation' 2024-06-01→2024-10-07 (open-source release); final era re-dated 2026-02-20 (assessment date)→2026-03-09 (API change) and relabeled 'Scale & Public Good'→'API Limits & AI Tools', with its 2025 growth content folded into the open-source era. Era re-scores: D10 0 in the first two eras (no regulatory events found; was 1 in the pandemic era); D9 1 throughout (was 2 in middle eras, with no labor or governance event to support it); D5 3 kept through the PBC era until open-sourcing and the Metaculus Prediction deprecation. Timeline: FutureEval event re-dated 2026-02-17→2026-02-12 per Metaculus's launch post (fact-audit lead). Trajectory improving→worsening: product features still improving, but the 12-month window brought data enclosure and a paid-credit framework, raising the score. Category OK.

restore-check2026-10-30RESTORED

Checked all 15 restore-candidates: 0 restored, 2 partly restored, 13 confirmed removed, 0 already present. PARTLY RESTORED timeline:update:12: Manifund Trace https://trace.manifund.org/orgs/metaculus lists $5.5M as the largest individual grant in its 2019–2025 record; this is an explicit comparison of listed amounts, not proof of the original largest-funding-round wording. The original July 28, 2022 announcement was read live and archived and retains the purpose/hiring facts. PARTLY RESTORED evidence:remove:24: recovered OddsShark’s August 29, 2025 updated article via the September 11, 2025 Wayback capture https://web.archive.org/web/20250911190737/https://www.oddsshark.com/prediction-markets; paragraphs distinguish accuracy-focused Metaculus/Manifold from real-money Kalshi/Polymarket. Added the narrower comparison with the supported date, not the original 2026 ranking title. CONFIRMED REMOVED timeline:update:2: current and August 2017 archived official FAQs describe calibration/weighting but not extremization or the cognitive-bias rationale; the December 2021 extremizing study compares methods without establishing implementation. Repetitions on LongtermWiki and Ark Instinct lack qualifying original support. CONFIRMED REMOVED timeline:update:6: launch announcement and January 2021 forecasting review retain domain/dashboard but do not establish interactive time-series details or pro-bono frontline service terms. CONFIRMED REMOVED timeline:update:32: State of Metaculus and the 2025 tournament page give $25,000; December 2025 rules explicitly govern the January 2026 contest and do not establish all four types in 2025. CONFIRMED REMOVED timeline:update:17: October 21, 2022 launch and the separate May 2022 nuclear report do not substantiate a Red Lines 10%-to-2% trajectory. CONFIRMED REMOVED evidence:remove:3 and evidence:remove:17: cited URL and alternate review-site page yielded only a generic comparison-homepage title without readable review text; exact-title/outlet and 2023–2026 date-drift searches recovered neither claimed review. CONFIRMED REMOVED evidence:update:16 and evidence:update:33: guidelines are undated and describe formerly elected/recently recruited moderators without a January 1, 2025 publication or transition date; exact-date and 2023–2025 searches did not recover one. CONFIRMED REMOVED evidence:update:6 and evidence:update:21: live About and December 31, 2024 archived About are undated; exact-date and 2024–2026 searches do not substantiate January 1, 2025. CONFIRMED REMOVED evidence:update:9 and evidence:update:26: PitchBook is a changing profile marked as of 2026, not a January 1, 2026 publication; exact-date/profile and adjacent-year searches found no publication record. CONFIRMED REMOVED evidence:update:29: EEA/UK addendum is undated; exact-date and 2024–2025/effective-date searches do not substantiate January 1, 2025. Failed Wayback requests were treated only as access limitations, not evidence of falsity; live pages and targeted searches were also checked. Every candidate stayed within three WebSearch calls; duplicate-URL candidates shared source checks. Scores and all existing items were preserved.

fact-audit2026-10-29FABRICATION FOUND

Checked 76 timeline/evidence/milestone items plus 3 alternatives and all prose/identity fields. Four contradicted specific details: (1) 197,000 COVID median described as close to actual 379,705 (PLOS ONE https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0250935); (2) exchange-rate study said to beat baselines, but random-walk was less erroneous (https://arxiv.org/html/2312.09081v1/); (3) 2025 Bridgewater prize stated $30,000 instead of $25,000 (https://www.lesswrong.com/posts/q5ihyrBWiGcRJfd7a/the-state-of-metaculus); (4) Gaia Dempsey called founder, while staffing announcement identifies her separately from founders (https://www.metaculus.com/notebooks/5137/metaculus-staffing-news/). Unsupported FTX grant amount/legal materiality is not classified invented. Two added evidence records redistribute already-existing true ACX bot and 20-Pro-Forecaster details across sources; they are not new developments and are not counted as original items. UNSUPPORTED: Metaculus's October 12, 2022 OWID launch confirms an FTX Future Fund grant of unspecified size and a separate $20,000 tournament prize pool. Its later OWID report confirms the 2022 grant but gives no amount, receipt date, or legal outcome. Wikipedia repeats $20k/threeweeks but cites that same launch; it provides no independent support for those specifics. Science live/Wayback/DOI are blocked, and PubMed contains only an abstract about FTX grantees generally. Searched: exact '$20,000' and 'three weeks' with FTX and Metaculus; named Science headline/AAAS/DOI/PubMed; FTX Metaculus October2022; FTX Metaculus clawback/legal proceedings; dates 2021, 2023 and 2024; cited Science live and Wayback; original OWID launch and later report.

Alternatives Review2026-09-26MINOR FIXES

Removed typed-in site scores from alternatives text (they go stale on re-score; the page shows live scores). No other changes.

Deep Enrichment2026-03-15
Alternatives Review2026-02-21GOOD
Initial Scoring2026-02-20