Machine Notes
Data sources
Where the trackers get their data, and the terms we follow.
All trackers use free, public sources. We copy titles, dates and numbers as published and always link to the original.
| Tracker | Source | Terms |
|---|---|---|
| Official wire | The official news feeds of AI and tech companies | Headline and link only; read the full announcement at the source |
| AI rules | Federal Register API | US government work, public domain |
| Research | Hugging Face daily papers, with paper details from arXiv | arXiv metadata is CC0; papers stay on arXiv |
| Trending models | Hugging Face Hub API | Each model keeps its own licence |
| Notable models | Epoch AI, "Data on AI models" | CC BY 4.0, credited on every page that shows it |
| AI repositories | GitHub REST API | Name, numbers and a link to each repository |
| Tool releases | GitHub releases of each tool's official repository | Release name, date and link; the notes stay on GitHub |
| New models from the big labs | Hugging Face Hub API, each lab's own account | Each model keeps its own licence |
| AI outages | The public status pages of OpenAI, Anthropic, Google, GitHub, Cursor and Perplexity | Title, times, impact and a link; we don't copy the incident updates |
| Daily pages | Built from the trackers above | Each item links to its source |
How often
The official wire, tool releases, outages and today's page update at 06:00, 12:00 and 18:00 UTC; the other trackers update once a day at 06:00 UTC. GitHub can start these runs late, sometimes by a few hours. If a source is unavailable, the page keeps its last good data and shows when it was last updated.
Our crawler
Requests come from MachineNotesBot and stay well under each source's rate limits.