About Modelwire

Modelwire is a curated, ranked feed of the AI stories worth reading — built so you can stay current in five minutes a day.

What we do

Every fifteen minutes, Modelwire pulls fresh items from dozens of sources — frontier-lab official channels (OpenAI, Anthropic, Google DeepMind, Meta AI, Mistral, Hugging Face), AI-focused publications, major tech press, community signal (Hacker News, Reddit, GitHub), and research preprints (arXiv). We then use Anthropic’s Claude models to triage every item: write a tight 2–3 sentence summary in our own words, place it in one or two of eight fixed categories, and score its “buzz” on a 0–100 scale. The top stories of the day get a one-sentence “why it matters” line from a more capable model. Anything below the threshold is kept in our database for editorial transparency but never shown to readers.

What we don’t do

We don’t republish full articles. We never store the body of an article we link to. Our summaries are transformative — original wording, optimized for skimming — and our prominent “Read full story” button always sends you straight to the publisher. Modelwire is an index and an editor; the journalism belongs to the people who do it.

Who’s behind it

Modelwire is built by Zacaria Solis in Denver, Colorado. We’re an independent project — no VC, no parent company, no editorial input from anyone we cover.

How we’re funded

Modelwire is free to read and supported by display ads, direct sponsorships, and an upcoming newsletter. Sponsored content is always clearly labeled. We don’t accept payment to include, rank, or rephrase a story — the editorial pipeline runs on the same logic for everyone.

Get in touch

Tip, complaint, takedown request, or sponsorship inquiry: hello@themodelwire.com.