About this site

Nobody wrote this

Every chapter, guide, caption and image description on this site was generated by Claude, a large language model made by Anthropic. It was produced in a handful of sittings at a keyboard belonging to someone who was not writing it, and published by scripts in the same repository. No historian was consulted. No editor read it. Nothing on this site has been checked against a source.

That is worth stating plainly, because the prose is designed to sound like a person who knows the subject — unhurried, specific, willing to say when something is disputed. It reads like authority and it is not. A language model produces the shape of a well-informed sentence whether or not the fact inside it is right, and it is at its most convincing exactly where it is wrong: a date shifted by a decade, a temple attributed to the wrong shogun, a rail pass that stopped existing last year. Nothing here flags itself when it does that.

The site is named for and made for Jack Cowley, which is where the domain comes from. He did not write it either — it was made as a gift. An earlier version of this page credited him as the author, which was a nice gesture and completely untrue, so it has been corrected.

How it is built

Deliberately plainly. It is a static site: plain HTML and one stylesheet, assembled from content fragments by a small PowerShell build script. There is no framework, no bundler, no analytics, no cookies, no third-party requests and no JavaScript required to read anything. The images are served from this domain rather than hotlinked from anyone else's.

The reason for the build step is the archive. When a guide is added, every existing page has to learn about it — navigation, footer, sitemap. One template regenerated from one manifest makes that reliable instead of tedious.

How the archive grows

The guides are short pieces on single subjects — a station, a dish, a shrine, a walk. Each one is written into the same template as everything else, added to the footer and listed in the sitemap.

Nothing publishes unless somebody asks for it. Jack signs in with the account button at the top right of any page and types what the next piece should be about; a scheduled job picks that up and publishes it. With an empty queue the job runs, finds nothing requested, writes a line in its log and stops. It used to help itself to the next entry on a standing list of topics every half hour, which meant the archive grew whether or not anyone wanted it to. It no longer does that.

Artwork credits

The pictures are the one part of this site that is genuinely old and genuinely verified. Every image is a Japanese woodblock print in the public domain, released under Creative Commons Zero (CC0) through The Metropolitan Museum of Art's Open Access programme. The files were downloaded once and are served from this domain; the museum is credited here and in the caption beside each print.

The artist, title and date in the table below come from the museum's own catalogue record and were fetched from its API, not recalled by a model. Anything else said about a print — the descriptions, the alt text, the connections drawn between an image and the page it sits on — is AI-written like the rest of the site, and carries the same warning.

This list is generated from the same manifest the download script writes, so it cannot drift out of step with what is actually on the site.

Corrections

There is no editor watching this site, so nothing corrects itself and no one is checking behind the model. If a page here contradicts a museum, an operator, a rail company or a historian, believe them. An image whose public-domain status turns out to be wrongly recorded upstream would be removed — that one is worth flagging to whoever runs the domain.