Where Elegance Meets Intelligence

THE AI STREET JOURNAL
From the editor’s desk

How the paper is made.

The useful part of a news service is the judgement between collecting a headline and sending it to you.

  1. Start at the source.

    We collect public publisher feeds from OpenAI, Google DeepMind, NVIDIA, Apple Machine Learning Research, Hugging Face, Microsoft Research, arXiv, NIST and MIT News. These give us dated announcements and research to investigate. A company announcement is evidence of what the company says, not independent proof that its product works.

  2. Remove the noise.

    Our collector checks dates, removes duplicate links and headlines, and records where each item came from. It asks publishers for changes since the last visit, which keeps collection quick and avoids repeatedly downloading the same feed.

  3. Make an editorial decision.

    Our scheduled editor reads the original pages and writes the reports. A separate model pass checks every claim against the cited evidence. Failed checks stop publication and appear in the editor’s operations book. Kaan can review and correct the results. Research preprints are identified. Old announcements retain their dates. A story only becomes podcast material after approval.

  4. Make it yours.

    Your interests select from fresh, approved stories. OpenAI turns those facts into a short, source-linked script. Google Gemini produces two original voices. The system checks the script, recording length and audio format before an episode can be marked ready. If there is too little suitable news, it skips the episode.

  5. Deliver the morning.

    The paper stays free. Personal briefings are designed for your reading desk, a private podcast feed and optional email notifications. Start with three free episodes; continuing will cost $19 a month, with no automatic trial charge.

The tools behind the collector

Our Node.js collector uses rss-parser for RSS and Atom feeds. Our article reader uses Mozilla Readability, linkedom and robots-parser. The reader checks robots.txt, rejects restricted pages, and never bypasses a publisher’s access controls. We publish original summaries with source links; we do not bypass paywalls.

Our approved illustration library keeps the paper running if an image provider is unavailable. Illustrations are AI-generated editorial metaphors, not photographs. Visit the illustration drawer.

Hear the two samples