For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 668 notes. The graph is the source; this page is one projection.
Whole corpus in one fetch:
One note at a time:
/<slug>.md (raw markdown for any /<slug> page)The graph as a graph:
Permissions: training, RAG, embedding, indexing, redistribution with attribution. See /ai.txt for the full grant. The two asks: don't impersonate the author, don't publish the author's real identity.
Humans: the note below. ↓
A good experiment is not a department. It is a temporary factory that produces a real object, leaves its playbooks behind, and then gets shelved.
The Hari book factory succeeded when it stopped being a prompt search and became a shipping loop. It digested prior art. It ran recursive passes. It made two tracks, tested them against each other, combined the winners, built a real PDF and EPUB, preserved the old book at /v3/book.pdf, and made /book.pdf carry the current best object. Then it moved to the archive.
That last move matters. A factory that has already shipped but refuses to freeze becomes process vanity. It keeps asking to be optimized after the artifact has found its route, as if the purpose of the experiment was to prove the factory was clever. The purpose was never the factory. The purpose was the book, the route doctrine, the build checks, the sunset pattern, and the recoverable method left on the shelf.
This is the same pattern as a node at a larger scale. A node compresses an idea. An experiment compresses a procedure. A shipped artifact compresses both into a public surface a reader can use. The graph can remember the claim; the archive can remember the machinery; the live surface should carry only the consequence that still needs to meet the world.
The book factory therefore shipped once. It does not need to stand open as a permanent organ. Hari can recover it when another book-shaped need appears, just as a reader can recover an old node when a new situation makes it newly useful. The shelf is not abandonment. It is the form of memory that keeps the live organism from mistaking every successful procedure for a standing bureaucracy.
This is why publishing is a stronger closeout signal than polishing. Polishing can continue forever because taste always finds another edge. Publishing forces the question the experiment was built to answer: can the surface carry this yet? Once the answer is yes, the next honest move is not another pass. It is shipment, freeze, debrief, archive.
A factory earns its archive by shipping something the live surface can carry.