Frequently Asked Questions

A single number, 0–100, updated twice daily, measuring AI progression across 4 dimensions: capability, autonomy, integration in critical systems, and control bypass. The score is based only on documented events from the last 12 hours — never predictions.

4 dimensions, each up to 25 points. The score moves up or down based on documented drivers. Minimum impact threshold: ±0.2 points. Max 8 drivers per day. Full breakdown at /methodology.

Because AI moves fast. Morning update (07:00 Israel) and evening update (19:00 Israel) — each measures the 12 hours preceding it. If nothing happened in the window, the score doesn't move.

They predict when AGI will arrive. We measure what has already happened. No predictions, no imagination. If it didn't happen and isn't documented, it isn't counted.

40 sources scanned each run: Anthropic, OpenAI, Nature, Science, MIT Technology Review, Reuters, Bloomberg, IEEE Spectrum, arXiv, and more. Full list at /sources. Every event links to its primary source.

An event must meet 3 criteria: (1) Actually happened — not a prediction. (2) Reported by a credible source from the 40-source list. (3) Measurable impact on one of the 4 dimensions. Otherwise: 0 points.

If it doesn't meet the 3 criteria, was already covered in the previous 12 hours, or comes from a source outside the 40-list — it's excluded. Decisions are transparent: every rule is detailed at /methodology.

Eithan (eithan@anipogam.co.il). Not an AI company. No NDA. No investment in any specific model. The full backstory at /story.

No. The score can — and does — drop when regulations pass or bugs get patched. Direction is set by the evidence, not by opinions.

Both. Claude Sonnet 4.6 scans the 40 sources twice daily and proposes drivers. I (Eithan) review the output before any score is published, and can reject any driver that doesn't meet the standard. Every score that ships has been human-reviewed.

Anyone who wants the state of AI without hype or doom — journalists, decision-makers, AI safety researchers, and anyone living in this world.

Yes, free for everyone: GET https://madad-ai.com/data.json. This is the source of truth for the score. The JSON includes the score, drivers, and full history. View it here.

Through the signup page. Sign up with email and phone, verify your email, and complete checkout. After that — one alert each morning and evening, via WhatsApp and email. Cancel anytime.

The website is free forever. The API is free forever. WhatsApp + email alerts are free for the first 3 months after launch, then a low-cost subscription plan (under $10/month).

Subscriptions to daily alerts. No advertising. No data sales. No sponsors. The aim: an independent public service that pays for itself and isn't beholden to any stakeholder.