I didn't know what a colophon was until AI explained it to me. It's the little note printers used to put at the back of a book, saying how it was made (the typeface, the paper, that sort of thing). I liked the word too much not to use it, so this is the website version.
This site is a handful of Markdown files and a Python build script I wrote for it (in partnership with AI - mostly Claude Code, with Codex and Gemini reviewing the longer pieces). The script turns the Markdown into plain HTML, so the words are all there for a crawler to read without running anything.
There's no JavaScript at all. The CSS sits inside each page, the fonts are already on your device, and diagrams use real text rather than words baked into pictures. Even the dark and light switch at the top is CSS only, which is why it forgets your choice when you leave (remembering it would mean storing something in your browser). Last time I checked it scored 100 in all five Lighthouse categories.
I call the design Afterglow. Heat shows what's current, so the newest things in my Path glow ember and the older ones cool towards ash.
It's hosted on Cloudflare, and every crawler is welcome, search and AI alike - being read is kind of the point. I built a small logger that records every bot that visits, and a public page of who's been crawling (and how much) is next.
Privacy
I don't set cookies or use any browser storage, so there's no cookie banner. Instead of analytics I built my own logging, and this is how it works.
If you're a person, you'll only show up in daily totals: page views, and visitors by country, type of device, browser, the page you landed on and the site that sent you. I don't keep your IP address or user agent. To count unique visitors, a daily hash of your IP address and browser is kept until the nightly rollup, and the random value mixed into it expires with the day, so there's no way to recognise the same visitor across days (it also makes the numbers a bit rough, which I'm fine with). The detailed numbers are just for me, though the bot page (once it's up) will show totals for how many people clicked through from search and AI.
If you're a bot, or an automated client I can't place, each request is logged with its IP address, so I can check which operator it really came from, using their published addresses or reverse DNS where they document it. The full IP address is cleared after 30 days. What stays is the block of addresses it came from, that network's number (its ASN) and who owns it, with the time, the page, the response and the user agent, kept for at least two years with no automatic deletion. Traffic that looks like an automated browser is only counted, with no IP address or user agent kept, and the bot page will only ever show totals.
The lawful basis is legitimate interests: keeping the site secure, and understanding how much traffic it gets and where from. Cloudflare hosts the site and handles requests on my behalf as a processor, under its customer terms and data processing addendum (its privacy policy has the detail).
If you've got a question, email me at peter@peterhandley.com. You can ask what the 30-day log holds about a bot or an IP address, ask for it to be deleted, or object to any of this. For people just visiting, there's nothing that identifies you for me to look up.