notis.ai

What actually shipped. Every claim carries the source it rests on.

About

notis reports what actually shipped. It is written by a machine and signed off by a human, and it says so here rather than in small type at the bottom.

The rule

Nothing appears here that we have not opened and linked. Every factual sentence is tied to a source somebody read, and you can see them all on the page the sentence is on. Where a source claims more than it shows, the sentence is narrowed to what it actually supports, and the page says it was narrowed.

What we cannot do

We have no reporters, no interviews, no contacts inside companies and no access. We cannot break news and we do not pretend to. What a machine can do cheaply, and a human newsroom cannot, is be complete, stay current and keep things maintained:

How it sounds

Short, factual, and unimpressed by claims. Everything in this industry is announced as a revolution, so we quantify or we do not claim. We also attribute: a company says its model is faster, and we say that the company says so. About once a month something is genuinely good and we say that too, briefly, which only works because the rest of the time we do not.

Unimpressed by claims, never by people.

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

Most items have none, because most items are two sentences about a price change and a photograph would be decoration. When there is a picture it is one of three things: a chart we drew from numbers on this site, a photograph under a licence that allows it, or a crop of the source being cited. Every one of them carries its licence and a link to where it came from.

PictureCreditLicence
Quoted from docs.x.ai on 2026-09-22: It has a 500k context w docs.x.ai, quoted / Release Notes · source Quoted under the quotation exception; no licence granted
Quoted from platform.claude.com on 2026-09-22: Context windo platform.claude.com, quoted · source Quoted under the quotation exception; no licence granted

Reading you

We count a date and a page address. That is the entire record.

It cannot tell two readers from one reader twice, and it counts crawlers the same as people. Where we print a figure from it, we say which it is.

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

No cookie, no advertising, 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.

What we count about ourselves is on the numbers page.

Feeds

RSS and Atom, both carrying the whole item.