About

ai notis is daily AI news in plain English. An AI newsroom with a human editor. Every fact links to its source. Our editor, Karl Magnus Horpestad, is responsible for what we publish. He approves each edition before it is published.

Follow ai notis on Instagram and Threads

How we check facts

We publish only what we have checked against a source, and we link that source. Every factual sentence is tied to a source that was read, and the sources are listed on the same page. When a source claims more than it can show, we state only what it supports, and the page says so.

How a story is made

Every story is written in our AI newsroom, so the byline is ai notis, not a person. Each factual sentence is checked against its source, linked on the story's page under "Check our sources".

Each story shows when it was published and, if it changed afterwards, when it was updated. It changes only by a correction, which is listed under it, or when its picture is replaced.

Responsible editor

Karl Magnus Horpestad is the responsible editor of ai notis. He answers for what it publishes, including its corrections and its posts on social networks. Write to hello@ainotis.com. More about the editor.

More about the editorEvery correction

Who publishes ai notis

Karl Horpestad, a person in Norway, runs notis and is responsible for members' data.

For questions about a member's data: hello@ainotis.com

What we can and cannot do

We have no reporters, do no interviews and have no contacts or special access inside companies, so we cannot break news. Because the writing is automated, we can do three things well:

• We cover everything. Every story we checked that day is listed, not only the three top stories.
• We keep pages up to date. Each reference page shows the date of its last check, and says so when a check is overdue.
• We keep a full record. Our topics page shows what we followed, what became real and what came to nothing, including topics we were wrong to follow.

How it sounds

Short and factual. Companies often describe each release as a breakthrough, so we give the numbers or we leave the claim out. We also say who is making a claim: if a company says its model is faster, we write that the company says so. When something is genuinely good, we say that too, briefly.

Each of the three top stories ends with "What it means for you", marked as our view. It adds no new facts: any number in it is already checked in the story.

Corrections

When we get something wrong we correct it on the page, dated, and we keep the original wording so you can see what changed. We do not quietly edit. Every correction is here.

Pictures

Each of the three top stories has one picture; the short items have none. Real pictures come first: the company's own image of what we report on, credited and shown as published; a still from its video, which links to YouTube (your browser does not contact YouTube unless you click); or a licensed photograph of the event, the people in the news or the place.

When there is no real picture, a top story may carry an illustration made with AI: a drawing of a device that stands for the story. It is always labelled “Illustration made with AI for ai notis”, and its labels are written by us from the story's checked facts, never by the AI. An illustration is never a picture of what happened. If none can be made, the story carries a card naming its source instead. Every illustration is in Explore, with its story.

Each picture is credited and links to its source, and we take down any image when its owner asks. We also draw charts from numbers on this site, a card showing one checked number and a card quoting the source's own words; a number on a card is always in the story. On Instagram and Threads, some covers are AI illustrations, always labelled “AI illustration”.

What we count when you read

We count a date, a page address, and where the reader came from, as one of a few labels (search, an AI assistant, social, email, another page of ours, our app on a home screen, or no source given). For a link we posted on a social network, we also count which of our posts it was. That is all we record.

Visits from robots (search engines, AI crawlers, feed readers) are counted apart from readers'. We tell them apart by what the browser says it is, and by whether the request carries the language and request-type lines every browser sends; we look at these while counting and never keep them. To try a new robot test before we use it, we also count, per day, which of the robot tests we are trying a reader's visit would have failed, by kind of page and as one of three words for where it came from; never the page's address, the browser's name or version, or what any line it sent said. A browser that asks not to be tracked is left out. Our own checks of the site are not counted at all: our tools name themselves, and we look at the address of each visit only to tell whether it comes from our own server, and never keep it.

Under each story you can react with Like, Love, Insightful, Surprising, Funny or Not useful. We use the counts to decide what goes in the monthly issue.

We count reactions per story and per day, never who reacted.

The reaction pictures are Microsoft's Fluent Emoji, used under the MIT licence and served from this site.

Following a case is counted the same way, a number per case and a day. Unless you join ai notis, the cases you follow are kept in your own browser, not here.

This count cannot tell two readers from one reader who came twice, and a robot that copies a browser closely enough is counted as a reader. When we publish a number from it, we say what it counts.

So the number we show beside a story is a different count: read on screen. The story's page sends one message, with its address and where it was opened from (kept only as one word from a fixed list, such as "Instagram"), once it has been on a reader's screen for three seconds or the reader scrolls it. Our link-in-bio page and the front page send the same message, so that we can see whether readers who come from Instagram reach a story. A robot that never shows the page never sends it. It is counted from 15 October 2026, as a number per story and day. If your browser says Do Not Track or Global Privacy Control, it is not sent.

We keep that daily record for about 13 months. After that only the monthly totals remain.

If you have not joined ai notis: no cookie, no advertising tracker, no third party, nothing that could be tied back to a person. That is also why there is no consent banner: there is nothing to consent to.

If you join ai notis

Joining is how your saved stories and the cases you follow reach every device you use. All you need is an email address. There is no password: we email you a link, or you use a passkey.

Then we keep your email, what you save and follow, your reactions, and the day you were last here. One cookie keeps you signed in. It is needed for the service you asked for, so it does not need a separate consent. Your reading days are counted only if you switch them on, only you see them, and switching them off deletes them.

Being a member does not change how we count reading: the count described above still records a day, a page and where the visit came from, and does not know which visits were yours. You can download everything we keep about you, or delete it, from your own page at any time. Everything we keep.

The morning email

On the mornings we publish, the morning email brings the three top stories, each with three key points and what it means for you, and up to four short news items, with links to the stories here. It is free, and nothing is sent until you confirm it from your own inbox.

We keep your email address, when you asked and when you confirmed, and nothing else. The list is held by Brevo, the service that sends our email. Our emails count opens and clicks, not who made them; we see totals only. Every issue has an unsubscribe link that works in one click, and leaving deletes your address from our list and from Brevo's. Everything we keep.

Advertising

An advertisement on ai notis is a box labelled "Sponsored · Advertisement", placed by us. It runs no script, sets no cookie, and tells the advertiser nothing about you. We count how many times a page carrying it was sent and how many times its link was followed, a number per day.

In the morning email an advertisement carries the same label, and there is at most one in an issue. We count the issues it went out in and the clicks on its link, the same way.

What we count about ourselves is on the numbers page.

Feeds

RSS and Atom, both with the full text of each story.