Disclosure

This page explains how AI is used to make the website itself, and how the website counts its visitors. How AI is used for the articles is explained separately on each article.

The website is built and maintained with help from AI tools. At the start, that tool is Claude by Anthropic, used through Claude Code. It helps with the code and with the look of the website. J.H. decides what is built and what is published.

The website also has short texts of its own that are not part of its content: the short sentence that introduces the website, notices such as "This website is still being built," and the words in menus and links. These texts carry no label. AI tools, at the start Claude by Anthropic, may help word them, and J.H. approves every one. This page says so once, in general terms, and on purpose does not credit each sentence: J.H. decided that a precise credit for every short sentence would not help readers.

The articles are different. Every article carries its own label near its title. The label says who wrote the article and how AI was involved, with specific credits: which model, from which provider, and what it did. The editorial policy explains these labels and credits.

When a model drafted or translated a page under one of the website's standard system prompts, that prompt is published on the website, in every version. You can find them on the "Standard system prompt" and "Standard system prompt for translations" pages in the footer.

The website counts page views, per page and per day, to see which pages are read. Views by automated programs (bots) are counted separately, as far as the website can tell them apart. The name a browser gives itself is checked to tell bots apart, and then forgotten. For a bot, the website only counts which known bot it says it is, as a name from a fixed list, such as "google"; a bot that is not on the list counts as "other". That name is not verified: any program can give itself any name. For human visitors, and for bots that present themselves as a browser, no name is stored at all. No IP address and no cookie are stored either. So the counts are page views, not people. When there are requests on the website for addresses that do not exist: the website only keeps a record of the last ten different addresses, each with how often it was asked, and when, and whether a person or which bot last asked. The address is kept exactly as asked, minus anything after a question mark. When a reader arrives from another website, only that website's domain is counted, such as the name of a search engine, never the full address. The website sets no cookies.

As web hosts commonly do, the website's hosting provider keeps server logs of each request, including IP addresses and browser details. J.H. does not use them for these statistics. The counts are not published; only J.H. can open them.