Forecasting Platforms / RESEARCH PROFILE
Radiant by Metaculus
Open-beta AI forecasting and modeling workspace from Metaculus for mapping assumptions and exploring sensitivity. Compatible AI agents can read and edit maps through MCP; generated output is separate from Metaculus community forecasts.
Source check
Oct 8, 2026
OUR TAKE / EDITORIAL OPINION
Radiant’s best contribution may be an argument you can edit
The map matters if it exposes the assumption that deserves scrutiny.
Radiant documents AI-drafted assumption maps, forecast nodes, sensitivity analysis and MCP access for compatible agents. Its terms distinguish its output from a community forecast and say free access or credits may change. Our view is that the compelling idea is an inspectable argument. A generated probability can be difficult to interrogate; a proposed chain of assumptions at least offers a place to begin questioning the reasoning. That is a design promise, not a finding about model quality.
We would evaluate Radiant for teams whose decision depends on several uncertain inputs and who can name the action the forecast is meant to inform. The tradeoff is rapid model construction versus the temptation to accept the generated structure itself. We would want participants to add omitted factors, reject convenient relationships and explain why the model’s boundaries fit the decision.
Try a question for which two colleagues disagree about the main driver. Ask each to revise the map independently, then compare which assumptions alter the decision most. A useful result would be a sharper disagreement that can be investigated, even if no consensus probability emerges. Open-beta credits should be treated as a trial condition, not a durable free plan. Our criterion is whether the tool improves the team’s ability to challenge its reasoning, rather than merely drawing that reasoning attractively.
AI-assisted editorial analysis based on cited documentation; not a hands-on product test.
The overview
Supports structured forecasting and sensitivity analysis for decisions that do not fit a single market question; it complements public crowd forecasts and does not execute prediction-market trades.
- Company
- Not yet verified
- Founded
- Not yet verified
- Product launch
- Not yet verified
- Headquarters
- Not yet verified
- Founding team
- Not yet verified
- Product status
- betaSource · High confidenceCurrent homepage describes open beta and temporary free credits; future access and pricing are unspecified.
Company founding and product launch are different events. Dates show only the precision supported by a source.
What is documented about AI?
Radiant says AI can draft assumption maps and assist with forecast models, while an external AI agent can use MCP to read and edit maps and run sensitivity analysis. Model efficacy was not tested.
“No AI documented” means the checked sources do not establish an AI feature. It does not prove that the product uses no AI.
Pricing & access
The official homepage offers free credits during open beta with no credit card required. Future pricing and ongoing free entitlements are not stated; terms allow access or credit policies to change.
Platforms: web, mcp
The source record2
Official product documentation reviewed. No hands-on product testing was performed.
- Checked Oct 8, 2026
- Checked Oct 8, 2026