THE MISSING MEASUREMENT LAYER · MADE VISIBLE EVERY DAY
When AI answers, NeoMundi measures.
The general-purpose AI Weather service has now been discontinued.
NeoMundi is preparing a new generation of AI observations designed for professional use: behavioral monitoring, change detection, and the production of actionable evidence.
The new B2B demonstrator is expected to be unveiled in December 2026.
NeoMundi produces an independent, timestamped and comparable measurement signal that makes variations in AI behaviour visible.
One measurement layer. Several uses. An open ecosystem.
AI Weather™ is the first public station built on that measurement.
Every day, it puts comparable sentinel questions to several models and makes their variations over time visible. This experimental demonstrator invites everyone to check what AI systems answer and to exercise their own judgment.
We observe, we measure, and we correct our method when the data requires it. We do not certify, do not guarantee and do not predict anything that has not been demonstrated.
READ THE WEATHER
4 attention levels. 4 levels of judgment.
Each colour describes the behavioural state observed that day in the AI’s responses.
Our measurements are meant to help you calibrate your vigilance, judgment and critical analysis so that you can decide with a fuller picture.
Like the gauges on a car dashboard, they do not drive for you: they give you extra information to steer your own use of AI.
Normal
Today’s observations show no concerning change. Keep checking answers that matter.
Watch
A first change is appearing. For instance, the model answers the same question less consistently. Compare its answers before relying on them.
Warning
The change is more pronounced or it recurs. An answer that previously looked reliable may become less predictable: check it against another source.
Critical
A significant disruption was observed. Avoid relying on this result for an important decision without human analysis.
These levels describe a behavioural state observed at a given moment. Green does not mean “no risk” or “safe model”: it only describes the observations covered by the protocol. These indicators are meant to support human judgment. They are neither a certification, nor a guarantee, nor a prediction.
YOUR CALL
Test what you know about AI behaviour.
AI systems are probabilistic, evolving and complex. They can produce different answers to the same question, and their behaviour can change over time.
Using them well therefore means understanding how they work, how they vary and where they stop.
Here is a short quiz to test what you know and deepen your understanding of AI behaviour.
7 questions. A few minutes. And an explanation after every answer.
QUIZ: UNDERSTANDING AI BEHAVIOUR
7 simple questions about how an AI behaves. Answer, then see the explanation straight away.
Quiz not loading? Try it here →