Developers & agents

Everything on this site that is meant to be read by a program. No auth, no keys, no rate limits beyond common sense — it is a static profile site, and it exists to be fetched.

Endpoints

OpenAPI 3.1 spec — /openapi.json
Every machine-readable endpoint on this site as a parseable spec — unique operationIds, typed responses, and the 404 problem schema.
llms.txt — /llms.txt
Short LLM-readable profile plus an index of everything else here, with absolute URLs. The right first fetch.
llms-full.txt — /llms-full.txt
The exhaustive citation-friendly profile: story, projects with outcomes, talks, engagement criteria.
about.json — /about.json
schema.org Person JSON with the full works graph — every project, talk and repo cross-linked by @id.
sitemap.xml — /sitemap.xml
All public URLs with change-frequency hints.
robots.txt — /robots.txt
All crawlers welcome, AI agents explicitly so; Content-Signal permits search, ai-input and ai-train.

Markdown twins

Every HTML page has a plain-Markdown twin — append .md to its URL, or send Accept: text/markdown to the HTML URL and content negotiation returns the same thing.

Errors

Unknown paths return a real HTTP 404 — never the homepage wearing a 200. The body follows your Accept header: an RFC 9457 problem document for application/json (with a resources map of every real endpoint), a short Markdown recovery page for text/markdown, HTML only for browsers. A plain Accept: */* — curl, most agents — gets the JSON.

Versioning & rate limits

Every machine-readable response carries X-Api-Version. There are no URL-path versions: this surface is additive, and nothing is removed or repurposed without a new major version announced in llms.txt at least 90 days ahead, after which retired paths return 410 with a Sunset date. Responses also carry RFC 9331 draft RateLimit headers — an advisory 240/60s budget; the site is static and not currently throttled. Full policy lives in the OpenAPI info block.

Formats

One URL, three representations: send Accept: application/json to any page for a PageDocument object (url, title, content as Markdown), Accept: text/markdown for raw Markdown, or just append .md to the URL.