The mission

Artificial intelligence is the most documented field of its time and one of the hardest to follow. The research is on preprint servers, the announcements on company blogs, the tools in newsletters, the roles on a hundred career sites, the events on ticketing platforms, and the news everywhere, repeated faster than it is checked. The information exists; it is scattered, uneven and rarely verified at the point where a reader meets it.

The result is a paradox: never has so much been published about AI, and never has it been harder to know what actually happened, who said it, and where.

AI.info was created to close that gap. Not another feed, but one place built around the questions people actually ask — what changed this week, whether a claim has a source, who is hiring for this, where the field meets next — where the paper, the lab that wrote it, the roles it is hiring for and the room where it will be presented sit on the same page, and where a story is not published until its quotation has been found and its source has been opened.

What is here today

—Stories—Essays—Papers tracked—Tools—Open roles—Events ahead—Companies—Creators

counted now · every catalog is maintained continuously, never scraped once and left

The standard

The method rests on one rule: no model grades its own work. Between a draft and the reader stand a machine check and a second model, and a third hears any appeal; none of them is the model that wrote it, and a doubt is settled only by evidence found on the page. Every story passes the same five stations, in this order, and goes back to the first whenever something is wrong.

The editorial pipeline: a primary source; a machine check of every quotation and figure against the pages the story cites; an adversarial second reading by another model; a review that holds whatever fails; publication — and, when something is wrong, a correction that goes back to the source.The standard · the pipelineSourceprimary onlyCross-checkby machineSecond readinganother modelReviewheld if it failsPublishedsource linkedCorrectionSource → Cross-check → Second reading → Review → Published5 stations · one return

Primary sources

Every story links to the original — the press release, the paper, the filing — never to an outlet that repeated it. No story is published without its source.

Verification

No draft reaches a reader unchecked. A machine first looks for every quotation on the page it is credited to, and for every sum and percentage on the pages the story cites. A second model — never the one that wrote the draft — then treats each claim as wrong until it finds it confirmed, on those pages or elsewhere on the web. A quotation that cannot be found stops the story.

How it is written

News stories are drafted with AI models and checked by other models, to this standard. A draft that fails a check is held. When the second reading fails it, a third model hears its appeal; a failure that stands sends the draft back to be rewritten and checked again, or drops it. Nothing is published over a failed check. Every day a further model re-reads a sample of what has gone out. The editor designed this method and answers for everything published.

Corrections

When something is wrong, the article is corrected and says so, with the date of the correction. Write to us: every report is looked into the same day.

What you will not find here

  • a paywall, or a story cut in half to sell the rest
  • sponsored placement: no tool, company or event is on the site because it paid to be
  • a story without its primary source linked
  • a tracking pixel in anything sent by email

The project

AI.info is a personal project, born of a practical experience: following the field meant a dozen tabs open every morning, and none of them told you which of the headlines had a source behind it. What was missing was not information but a place that held it together, cross-referenced it and checked it.

Building that place proved more demanding than expected. Sources have to be found at the origin, names and figures reconciled across releases, companies told apart from their subsidiaries, events and roles kept current as they open and close. Every story, essay, company page and event on the site goes through the same process, and every one of them links to where it came from. The process is not perfect, but it is taken seriously, and it is what the site is for.

The aim over time is for AI.info to be a reference for anyone who works with or around artificial intelligence: readers, researchers, engineers, founders, recruiters, students and journalists. Not a feed, but a working record of the field, useful, verified and built one fact at a time.

Something wrong, missing, or yours and misrepresented? Write in. A correction is made in the article and noted there. Every list on the site can be followed without an account: the jobs feed, the events feed and the events calendar take the same filters as their pages.

Created by Gabriele Masetti

The sections