AI news

A daily read on what actually happened in AI, from around thirty sources at once — and grouped by how much weight each source deserves.

What it is

An aggregator that follows roughly thirty sources — major AI labs, news agencies and scientific publications including arXiv — indexes them every two hours, and makes them searchable semantically or readable as a daily digest.

The digest groups items by credibility level rather than by recency alone, which is the point: a preprint and a lab’s own announcement are not the same kind of claim, and a flat list quietly implies they are.

Use it for

• “Summarise today’s most important AI news, grouped by credibility.”
• “What has been published on this topic in the last week?”
• “Which sources does this actually draw on?”
• “Find recent research on this, not commentary about it.”

Tools

Function	Answers
Usage guide	How the digest and search differ — the assistant reads this first
Fetch news	Recent items across the followed sources
Daily digest	The day’s items, grouped by credibility level
Search news	Items matching a query, searched semantically
List sources	Which sources are followed

Data and permissions

• Public sources. Nothing about your organisation passes through.
• Read-only.
• Indexed every two hours, so this is a recent-news connection rather than a live-wire one.

Getting started

Add the MCP server to an assistant and ask.

Limitations

• Around thirty sources, not the field. A story outside that set does not appear, and absence is not evidence.
• Credibility grouping is a heuristic about the source, not a judgement of the claim. A reputable outlet can still report something wrong, and a preprint can still be right.
• Up to two hours behind, so for a fast-moving story this is not the fastest place to look.
• Aggregation is not verification. The connection surfaces what was published; it does not check it.