How it works

Every hour

We check the AI newsletters, feeds, websites and YouTube channels we follow. An AI model reads each new item, picks out the news, and decides whether it is a new story or more on one we already have. For a new story it also reads the original pages the item links to, when they can be opened, and writes the article from that material. New information goes into that story's timeline, so one event keeps one page however many sources cover it.

Written by AI

An AI model (Anthropic's Claude) writes the summaries, and they go online automatically. Nobody reviews each one first. The model is told to use only what the sources say, and it does no reporting of its own, but it can make mistakes, so check the original before you rely on a story. If something is wrong, please tell us at office@ikangai.com.

How to read the labels

Official

From the company, authors or authority themselves, with a link to their own page.

Confirmed

Several independent outlets or named sources.

Reported

One outlet, not independently verified yet.

Rumor

A leak, an anonymous source or speculation. Treat it with care.

The red dots count the separate sources carrying a story. They show how widely it is being reported, which is different from whether it is true: several newsletters can repeat one press release.

For agents

AI agents can read this site through a read-only MCP server at https://ai-news.ikangai.com/mcp, with no sign-in. There is also an llms.txt, a Markdown version of every story, and the RSS feed.

Links and privacy

Links go straight to the destination, with tracking parameters removed. The site has no ads, tracking or third-party scripts.

Sources

AI Engineering, Forward Future (Matthew Berman), Import AI (Jack Clark), Interconnects (Nathan Lambert), Latent Space, OpenAI, Simon Willison, The AI Report, Theo - t3.gg, Turing Post