What this site runs, what it records, and the choice you get to make about it — in plain sentences.
Umamicounts pageviews and copy events — no cookies, no fingerprinting, no personal identifiers, no cross-site anything. It is how we know whether anyone uses this reference at all. It runs for every visitor, and both its script and the counts it sends go through this domain rather than the vendor's, so blocklists that cut the vendor's domain don't skew the only counts we keep. Fetches of the plain-text files (the .md mirrors, llms.txt) and any non-browser request for a page are counted on the server the same way: a URL and a user-agent class. For requests that aren't from a browser, a truncated copy of the user-agent string the agent sends is kept too — that is how a new crawler gets recognized. Nothing else.
Mixpanel(event analytics) andMicrosoft Clarity(session replay) record how the site is used in enough detail to count as personal data, so they are optional:
One cookie and four values in your browser's local storage. None of them is an identifier, none is sent to anyone else, and nothing here follows you to other sites.
No accounts, no names, no ads, no selling or sharing of data. The waitlist stores exactly the email you type into it, for exactly one send. Our success metric is copy events — whether the markup was useful enough to take — not how long you stayed.
Both controls apply immediately and reload the page.
Reset clears your stored choice; in regions with the banner it will ask again, elsewhere the defaults return.
Optional analytics (Mixpanel, Microsoft Clarity) help improve this reference. Cookieless page counts run either way.Details