How Neural Dispatch is made
Neural Dispatch publishes two briefings. Pulse Check is the world’s news in six stories, free, Monday, Wednesday and Friday, with Thursday and Saturday optional. The AI Model Risk Brief arrives Monday, Tuesday, Thursday and Friday, for model risk, compliance and governance teams, and anyone accountable for how AI is used where they work. This page is how both are made, and what neither claims to do.
Neural Dispatch is written and run by Zachery Link, who builds the system that writes each edition and reviews the editions it delivers to decide what changes next.
Where the stories come from
We read publications and public bodies on a schedule through the day. For Pulse Check that is thirty-two newsrooms, from world desks to science journals. For the AI Model Risk Brief it is forty-three sources: regulators and enforcement agencies, security researchers, academic preprints, trade press and the AI labs’ own announcements, plus a daily web search for the beats that publish least.
Each new source is tried before it is added, and one that goes quiet is dropped. Everything is collected and ranked before a word is written, so each story is chosen from what was published, not from what a model remembers.
How Pulse Check chooses its stories
Stories rank by how many separate newsrooms covered them, so an edition leads with what most of the press agreed mattered. A story many newsrooms keep covering day after day, such as an election campaign, stays in view even when no single day of it leads. Opinion pieces, analysis, video and galleries are left out before the selection sees them. Each outlet gets one story wherever the edition allows, and never more than two. The aim every edition is the six stories that matter most. Most mornings the sixth is the Wildcard, away from the main headlines but worth knowing: anything from a Roman coin hoard to a World Cup final.
How the AI Model Risk Brief chooses its stories
Six beats compete each morning: AI Regulatory Watch, Enforcement & Failures, Frontier AI, Model Validation, AI Security and Industry Intelligence. The three with the strongest news win a section, so a quiet week in regulation gives its place to a loud week in security. There is no fixed table of contents.
How an edition is written
Claude, Anthropic’s language model, writes both from the articles chosen. A Pulse Check summary may state only what its source reported, and names and links that source.
Every AI Model Risk Brief section is written from a published article and names and links it. If that link cannot be made, the section is marked BACKGROUND instead, with no link, rather than pointing you at the wrong page.
Every section ends on the so-what: not just what happened, but what it means for your work, and what you would change because of it in a validation plan, a model inventory or a vendor review.
Citations to regulations
The AI Model Risk Brief never cites a regulation from memory. Beyond the article a section is written from, it links only to documents on a list our team maintains, recording each one’s official link and whether it is still current. A new rule can be reported as soon as it is news, and is linked once our team has added it. If a citation is not on the list, the Brief leaves it out rather than guess.
What we check
Every edition is checked as it is written: its links, its sources, its length, its categories and repeated outlets, and for the AI Model Risk Brief whether each regulation it names is current. Weekday editions are then read again as they arrived in an inbox, against the articles they were written from, and what that finds goes back into the rules, so each fix carries into every edition after it.
What this does not do
It does not confirm a claim against a second source. A story is reported as its source reported it, so if a publication got something wrong, the edition will carry that error with a link to it.
Pulse Check leaves opinion out by how a publisher labels a piece, so an analysis published without that label can still get through.
Neither is legal or compliance advice, and neither tells you what your obligations are. They tell you what changed and what a practitioner would look at because of it.
Corrections
If an edition gets something wrong, reply to it or write to support@readneuraldispatch.com. It goes to our team for review.
Reading it
Pulse Check is free by email, Monday, Wednesday and Friday, with Thursday and Saturday optional, and the latest edition is on the web in full. The AI Model Risk Brief’s Tuesday edition is free by email and is the same edition subscribers read that morning, not a preview of one; all four mornings are $19/month, Pulse Check included. The latest Tuesday edition is free to read on the web, with no account.
Read the latest Pulse CheckRead the latest AI Model Risk Brief