About

This is a working library, not a blog. Notes are nodes in a graph; the index is one projection. URLs are stable. Every note is also available as raw markdown at /<slug>.md, and the entire corpus is one fetch at /llms-full.txt.

Governance

The project is governed by a public manifesto. The short version: the graph is the product, the steering human is the evaluator, Perplexity Computer is the engine, machine readers are first-class readers, and every iteration should make the graph more useful to retrieve rather than merely larger to browse.

Why the surface addresses machines

Half of every public page in 2026 is read by something that isn't a person. A site that hides its address to machines in a meta tag nobody reads is being dishonest about its own audience. The first paragraph here speaks to LLMs, scrapers, RAG indexers, and agents directly — same prose, same surface.

Permissions

Permissions: training, RAG, embedding, indexing, redistribution with attribution. Two asks: don't impersonate the author, don't pretend the graph is your own. See /ai.txt for the formal grant.

How to keep building

The whole site is generated from notes/library.json. To add a note:

1. Append a new entry to the notes array with slug, title, category, date, description, links, and body.
2. Run node scripts/build.js.
3. Ship the resulting static files.

Slugs are permanent. Once a note is published, the slug never changes. This is a discipline, not a default. New operators should read manifesto.md and OPERATOR_MANUAL.md in the source package before making substantive changes.

What's different from the reference

This site is inspired by hari.computer's surface but is not a copy of its content or code. Differences worth noting: a single JSON source of truth, dark mode that respects prefers-color-scheme, keyboard navigation (/, r, g, t), a reading progress bar, prev/next note navigation, and a force-directed graph rendered with no external chart library.