Loading…
Updated 24 October 2026
Nanofuze teaches AI, and its AI news page shows what changes in it: models, prices, rankings, releases, research and rules. Each item on that page is called a signal, and these are the rules every signal follows.
A signal is only shown with the numbers it was detected from and a link to the public data they came from. The link is on the signal's page, and the Data sources page lists every source, so a reader can always check the original.
Signals are automatic. Every hour a program reads public data (model price lists, the Arena leaderboard, Hugging Face, the release feeds of AI software and the United States Federal Register), compares it with the previous reading and saves what changed.
No language model writes a signal or decides how important it is. A signal's sentence is a template filled with numbers from the source, and its importance, from 1 to 5, comes from written rules that are published on the Data sources page.
A signal says what the data shows, not why. We do not repeat a company's claims about itself as facts. Leaderboard scores are shown with their margin of error, and neighbouring ranks are close to ties.
Where a signal rests on a limit of its source, for example a listing date that may not be the official release date, the page says so.
We do not read, copy, summarise or rewrite articles from other websites. Government documents are shown as published, because United States government text is in the public domain; everything else is data.
If we get something wrong, we fix it against the source and record the change on the Corrections page. To report an error, use the form there. A signal we can no longer stand behind is removed; its address then shows a notice instead of disappearing.
Nanofuze does not accept payment for signals, rankings or placement. What is shown is what the data shows.