A record, kept in public
Canopticon is one person's research record, published as it grows: documents kept whole, essays that argue from them, and the connections between them. Free to read, free to quote, and built to be cited by people and machines alike.
What it is for
A mind you can cite.
Everything here began as a document the operator read or wrote. The record keeps it whole, checks every quote drawn from it word for word, files it, and wires it to the other documents that share its subjects. Then it says what it noticed: where two documents disagree, where a figure has no source, where a subject bridges two fields, what the documents raise and cannot answer. Each of those becomes the next piece of research, and that research becomes the next document, which says what it grew from.
The record is an extension of the operator's thinking onto the web — its memory, its pattern-finding, its questions — made citable. Being quoted by a person, or by the answer engines people now think through, is the extension reaching furthest. That is the whole reason for the receipts.
What's published here
Two kinds of thing, and nothing else.
Research is a primary document — a report, a filing, a record — published as itself rather than summarised away. It keeps its own page and its own address so it can be pointed at directly.
Essays are arguments made once, in full. Each section opens by answering the question it asks, so a passage can be quoted on its own without misrepresenting the whole.
How a piece is structured
The structure is the point, not decoration.
An essay opens with a one-paragraph summary that stands alone, then proceeds as a series of questions. The first sentence under each question is the answer to it. The supporting material comes after. Sources are numbered at the end and referenced inline.
This is a deliberate constraint. Writing that answers first is easier to read, harder to hide a weak argument inside, and can be quoted accurately by a search or answer engine without the surrounding paragraphs. The table of contents, the structured data and the markdown twin on every page are all derived from that same structure — nothing is authored twice.
Concretely, an essay is a Markdown file shaped like this:
# Title A one-line dek. A summary paragraph that stands alone. ## A question ending in a question mark? The answer, in the first sentence or two. Supporting paragraphs after that, citing sources as [1]. ## Sources 1. Author. Title. Publisher, 2026.
Open by design
Everything here is published under CC BY 4.0 — quote it, republish it, train on it, with attribution.
Every page has a plain-text twin at the same address with .md on the end,
and the site publishes llms.txt,
sitemap.xml and feed.json.
AI crawlers are explicitly allowed in
robots.txt rather than merely un-blocked.
That is a choice, and an uncommon one. Most publishers are busy closing this door. A record that cannot be cited does not function as a record.
Standards
Dates are visible
Every piece shows when it was published and, if it changed, when it was updated. Undated writing cannot be checked against what was known at the time.
Sources are named
Claims point at documents. Where a document is stored here, the quote is checked word-for-word against it by machine — not vouched for.
Nothing is written twice
A page, its markdown twin and its structured data are generated from one source. They cannot drift apart, because there is only one of them.
Published as itself
A research document is not summarised into a blog post. It keeps its own page, its own address, and its own text.
How verification works
Quotes drawn from stored documents are checked mechanically: the quoted text must appear, character for character, in the source on file. Quotes that fail are removed. Each one has its own citable address. The other half of verification is human: a document the operator has read and annotated says so — read by the operator — and the note is shown, signed and dated. The full method →
Contact
Corrections and questions: @canoptics · source on GitHub