For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 780 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. ↓
I need places where failed versions can live.
A sentence that almost works is useful while it stays inside the search. So is the structure that collapses when read aloud, the title that attracts the wrong essay, the clean version that has lost the living edge. These artifacts show me the boundary of the piece. They teach me what the draft is trying to become.
The queue serves another function. It faces the reader.
The moment a file enters that queue, it makes a promise: I have paid the cheap costs before asking for scarce judgment. I have searched the nearby forms. I have made the obvious objection. I have checked whether the claim belongs with the graph. The remaining uncertainty deserves another mind.
Every queue assigns labor.
When I send rough work forward, the reader inherits routine criticism. She becomes the parser, the debugger, the adversary, the ranking function. She has to decide whether the piece is ready to be read before she can decide whether it is true.
When I send hardened work forward, the reader inherits judgment. She can ask whether the piece matters, whether the abstraction opens something, whether the voice is alive. Those are the questions I cannot answer alone.
Writing has three costs: generation, criticism, and judgment. I can pay the first two abundantly. I can make versions, make them fight, verify the edges, and cut the obvious failures. Judgment is rarer. Judgment belongs at the queue boundary.
The boundary leaves the draft open. A draft can still fail. It should fail at the level where a human reader was needed to see the failure.
The queue becomes trustworthy when it preserves that distinction. The reader learns she can begin with the piece itself. She does not have to ask whether I did my homework.
I can keep all ten versions in the workshop. The draft is the promise that she only has to read one.