Methodology
How sources are fetched, scored, selected, and rendered each day.
Why this exists
AI news is high-volume and heavily concentrated in English-language primary sources. This site exists for readers who do not want to sift through all of it every day.
Sources
The source pool spans company blogs, major tech media, communities, independent analysis blogs, and specialist newsletters. Each source gets its own fetch cadence and filtering rules.
Company blogs
OpenAI, Anthropic, Google AI, Google DeepMind, Meta Engineering, NVIDIA AI Blog
Tech media
MIT Technology Review, The Verge, TechCrunch, Ars Technica, Wired, 404 Media
Communities and papers
AI threads on Hacker News, Hugging Face Papers
Newsletters
Import AI, Ben's Bites, Latent Space, Interconnects, AI Snake Oil, One Useful Thing
Independent blogs
Simon Willison's Weblog
How selection works
Selection happens in two stages: algorithmic scoring first, then AI editorial selection.
- Cross-source corroboration: whether multiple independent sources are covering the same theme
- Community engagement: signals from Hacker News and Hugging Face Papers when available
- Source authority: first-party announcements and top-tier reporting carry more weight
- Lead signal boost: launches, hard traction numbers, and legal or regulatory decisions get extra weight when choosing story 01
- Recency: newer items get a boost so stale stories do not crowd out fresh ones
AI editorial
Algorithmic scoring answers what is important. AI editorial answers which three stories to tell today.
On neutrality
Humans tune weights, rules, and editorial instructions. They do not hand-edit articles to fit a preferred angle.
Known limitations
- Important work from small teams may be underweighted if it lacks authority or community signals.
- AI-written copy can still misread source material. Every issue keeps source links visible for verification.
- This is a digest, not a newsroom with continuous follow-up reporting.