Newsdesk
Engineering
Data & AI
Industries
Enterprise Systems
Go-to-Market
Longform
DesignIndiaAll stories

Where this comes from

Every figure on this page is read out of the archive when the site is built, including the unflattering ones.

How much the model decides

The distinction that matters is not which model is used but what supplies the facts. A title, a link and a date taken from a feed or an API cannot be invented; only the choosing and the phrasing are left to a model.

Facts come fromThe modelChannels
1GitHub Releases API, endoflife.datenot used at allRelease Radar
3CISA's catalog, the arXiv paper recordpicks and phrases, never writes a linkSecurity Weekly Digest, Supply Chain & ERP, arXiv Picks
14the feed record — title, link, datepicks and phrases, never writes a linkAI Tech Weekly Digest, Chips & Silicon, Climate & Energy, Craft & Leadership, Databases & Data, Design & UX, Engineering Reading, Fintech & Payments, HR & Payroll Tech, Marketing, Mobile, Open Source, Sales & Revenue, Web & Frontend
2search results it actually retrievedalso finds the storiesDevOps Weekly Digest, India Tech Weekly

Sources

136 feeds are configured across 20 channels, and 350 distinct publications have supplied at least one published story. No source is trusted by name: a link is kept only if its domain is on the channel's list and the page actually resolves.

Most publishedStories
CISA KEV94
arXiv90
SAP HCM Community59
Planet PostgreSQL51
SaaStr45
Frontend Masters43
LWN42
Android Developers39
Search Engine Journal31
Simon Willison30
Nielsen Norman30
UX Collective30

What is withheld

2229 stories are archived; 2166 appear on the site. The 63 difference is not editing. It is stories whose link went to a search page, to an expired redirect, to a page that no longer exists, or that repeated another story in the same issue.

The archive keeps all of them. It is the record of what ran, and it is what makes it possible to tell whether any of this is improving.

Links that no longer work

Every published link was checked on 2026-11-15. Of 2065, 2027 still resolve and 38 answer 404. Those 38 are withheld from the site until they come back; the archive keeps them, because what was published is a matter of record.

Most were never real: a model writing a plausible path on a genuine publication's domain. The check that would have caught them now runs before anything is published, so this number should fall rather than grow.