About
Next Token records how AI got here, one piece a week. The name comes from how a language model works: it picks the next piece of a sentence — the next token.
Two kinds of writing
An event checked against a source link — exactly one per record. Numbers a company reports about itself are marked as company-reported. If the source can't be checked, it doesn't go in.
A summary an AI (Claude) wrote of one piece of material it read (a talk or an explainer), without a link to the original. Notes carry a yellow tag and can be filtered on the timeline.
How this site is made
Most of what you read here is made by AI. One person runs the site and does not read or edit every item, so here is plainly who does what.
What the AI (Anthropic's Claude) does
- Every week it collects candidate events from a fixed list of places — paper repositories, lab and company blogs, government records and so on.
- For each candidate it opens the original source page, reads it, and writes the record with the date, actor and content matched to it. If there is only a secondary report, the record is labelled “Press” or left out.
- It also writes the notes (summaries of material it read), each week's question and threads, the weekly quiz, the answers in the question box, the Five-Minute AI Lessons and the AI History Module, the AI Words glossary, the people & company profiles and the introductions to each section, and the English translation.
Automated checks before publishing
- Every record has a source link and a date
- Rules that match the date and the actor against the source
- Checks for forbidden words and for format (how dates and numbers are written, style)
- Numbers, dates and names in the translation compared with the original
What the owner (factnotes) does
- Sets the direction and the rules — what gets recorded or left out, and how sources are chosen.
- Sets how mistakes get corrected.
- Does not read every item.
So there can be mistakes. Every record links its source — check anything important there. If you find a mistake, tell us through the contact form (topic “Error in a record”) and it will be checked and fixed.
Primary sources first; secondary reports are labelled “Press”. 88% of the 1,321 records rest on a primary source.
Source labels
Each record shows what kind of source it rests on.
A paper in a repository or journal
An announcement by the company, lab or organisation itself
A record from a government, parliament or international body
A secondary report — the primary source couldn't be found
Records what someone said, not a fact
Records a trend observed in a community
Learn
Five-Minute AI Lessons opened one a day for 60 days and now arrive one a week, for students aged 13–18: one experiment, three quiz questions graded on the spot, and slides as a PDF. The AI History Module retells the same record at three levels, and the Part Reviews sum up each part at three depths.
Follow along
New records and each day's lesson are in the RSS feed. The whole record can be downloaded as a dataset.
Who makes this
factnotes — AI is moving so fast that new models and tools arrive every few weeks. This site was made to check and analyse that progress against its sources, in order to learn it and to teach it. That is why every record carries a source, and the same story is retold as five-minute lessons and five-minute terms that students and teachers can use.
To get in touch, use the contact page. The rules for using the site are in the terms of use.
Using the content
Text and data are CC BY-NC 4.0: free for non-commercial use with attribution. The dataset is available as CSV and JSON.