Risk scores and notable events are generated by Google's Gemini model, recalculated automatically every hour. The current setup uses a free tier of the API, which means the model relies on its general training knowledge rather than live web search — see the note below for what this means in practice.

Not yet. The free tier currently used does not include real-time web search (sometimes called "grounding"), so scores and events are plausible estimates based on the model's training data, not a live news review. Upgrading to a paid tier with search access is planned to make this more accurate.

Each risk category (nuclear, AI takeover, climate, pandemic, asteroid, supervolcano, biotechnology) gets a score out of 100 from the AI. These scores are combined into a single weighted average — the End of Time Index — which then feeds a probability model to estimate the odds of a major catastrophe over different time horizons.

No. This is a model estimate based on an AI's assessment of current risk levels, assuming they stay constant over time. It is meant to be an interesting, transparent way to visualize existing risk research (Doomsday Clock, WEF Global Risks Report, Global Challenges Foundation), not a peer-reviewed forecast.