Sources and method

Feeds, read daily

Every day at 13:15 UTC the site requests the feeds it has read longest ago (each newsletter's, podcast's or YouTube channel's own RSS or Atom feed, once), so every feed comes round every few days, and adds the issues it has not seen. The Batch publishes no feed; for it the site reads the archive page and keeps only cards that show a title, a link and a printed date. No model reads anything and nothing is guessed: an issue without a date is skipped.

What is kept

Each issue's title, date and link. Never its text or images: the author decides what is free to read, on their own site.

Free or paid, and platform

For Substack newsletters, the site reads the public archive list once a week and counts how many of the latest 12 issues are for paying subscribers only: none is Free, all is Paid, anything between is Free and paid. For other platforms it comes from the subscribe page. The platform is what the feed itself reports (Substack, Ghost, Beehiiv, Buttondown) or the newsletter's own site.

Cadence and status

Cadence is measured from the archive: issues a month over the last six months, or over the days the feed covers when it holds less history than that (most Substack feeds list their latest 20). "Most active" sorts by it. Active: an issue in the last 45 days. Paused: last issue 45 to 120 days ago. Stopped: no issue for 120 days or more.

Topics

Fixed keyword rules over each issue's title and the feed's categories; an issue that matches none gets its newsletter's main topic.

History

Nothing is overwritten quietly. When a newsletter starts, pauses or stops, goes paid or free, moves platform, changes cadence, when its feed breaks or recovers, or an issue is retitled, a row is added to the change log. Issues that drop out of a feed stay. Each week a snapshot of every newsletter's numbers is kept.

The newsletters and their feeds

Checked and left out

The Neuron's public feed was found through its site; NLP News (Sebastian Ruder) and The Gradient are left out because their last issues are too old. Console's feed lists tool reviews rather than issues. Suggest one with Suggest a missing newsletter.

Across Glyph Ibis

Company publishers carry their slug in the shared registry at companies.fru.dev, and their pages link to the company's profile and to the same company on sibling fru.dev sites. Independent writers and hosts are people: when one person is named, their photo comes from Wikimedia Commons (matched through Wikidata by name plus the publication, a company its description names, or their occupation on a unique name; never by name alone) or from their company's own leadership or press page, credited with its license on the newsletter's page. Bylines that name several people get no photo; otherwise the logo or initials show. Nothing is copied from LinkedIn, X or image search.

Disclaimer

Titles, dates and links of issues, episodes and videos from each newsletter's, podcast's or YouTube channel's own public feed, read by a daily agent. New ones are found each day from Substack recommendations, the Apple Podcasts technology chart and companies.fru.dev companies' own sites; each find must pass one coded scope rule (tech, data and AI by a real publisher or person) and the rest are kept with the reason, not shown. Every item belongs to its author and links to the original. The list grows. Photos from Wikimedia Commons, the author's own publication or faculty page, and official company pages, with credit. Logos via logo.dev; trademarks belong to their owners. Anonymous usage stats via Google Analytics. Use Suggest a correction on any page.