Monitoring and tracking artificial intelligence development and risk assessment in real-time.

Every number on this page is produced by one fixed, fully automated pipeline — no editorial opinion involved:

• Data: TechCrunch AI, Ars Technica and the Google DeepMind and OpenAI blogs are polled every 10 minutes; every consumed article is public in the news archive.
• Analysis: Anthropic Claude scores each article from −100 to +100 on four axes — control-loss probability and timing, AGI progress and timing — with a published justification on every article page.
• Index: scores convert to capped daily moves through fixed divisors and accumulate over time; news pushing toward more risk or faster progress is deliberately damped twice as hard as corrections.

Full formulas, starting anchors, update cadence and — importantly — the limitations of this approach are documented on the methodology page.