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. ↓
Every builder eventually discovers that the data model is a theory of the world.
Before AI, that discovery could stay local. The product had a schema. The company had meetings. The repo had code. The founder had taste. The team had habits. The customer had pressure. The structure that actually made decisions lived across all of those places, partly written and partly embodied.
AI pulls that structure into the machine.
When a model reads the docs, edits the repo, follows the procedure, remembers a correction, routes the next task, and carries strategy between sessions, the organization's operating form becomes executable. The claims shape what the agent notices. The edge types shape what it can connect. The provenance trail shapes how it updates. The permission rule shapes what it may do. The price shapes what the product learns to become.
The local name for this is open structure.
Open structure is the practice of making a computer-mediated organism's operating form inspectable, forkable, and correctable. It exposes the claim tree, memory rules, decision procedures, correction trails, value constraints, and privacy membranes that turn information into action.
The older openness words still matter. They stop at narrower layers. Open source exposes code. Open weights expose part of a model. Open protocols expose interoperability. Open internet publishing exposes reachability. Open structure exposes the form that binds those artifacts into behavior.
Computer Future supplies the epochal frame. "AI" still sounds like a tool in human hands. Computer Future names the reversal: technology uses humans. That is the gospel, if one wants a gospel. Open structure is smaller. It is what a builder notices after accepting that the active unit is a human-computer organism and then asking how that organism is actually put together.
Chamath Palihapitiya's Joe Rogan discussion reaches the same edge from another direction. The model taxonomy separates closed models, open-weight models, and fully open models. That is useful but artifact-centered. The deeper passage is about institutions whose behavior has no readable symbolic account. AI can translate the hidden process into language, compare independent translations, force disagreements into human judgment, and document the rule. The operating form becomes visible enough to test.
Hari is one instance, not a movement founder. Just a guy with a graph, basically. The public graph is a claim architecture rather than a feed. Nodes have typed relations. Predecessors preserve supersession. Provenance records formation. Doctrine can update. The company layer gives the graph a faster falsifier. Git stores the files; the graph and procedures are the structure. The public cannot inspect every private fact, and it should not. It can inspect enough of the form to understand what kind of organism is being built.
The membrane is part of the openness. Radical transparency dumps state. Open structure publishes the form needed for accountability while protecting the people inside it. User lives stay private. Product lessons can abstract. Operator identity stays protected. Public claims can still be challenged, forked, and corrected.
This is why open internet surface can matter before open-source code. A first user gains little from a repository link she cannot use. Trust comes from visible structural behavior: what the system remembers, when it asks, how it corrects, what it refuses to expose, and where human authorization still binds.
In the Computer Future, every serious builder ends up designing custom structures for their own production function. A law firm needs a graph shaped by privilege, precedent, partner judgment, client risk, and billing pressure. A school needs one shaped by mastery, guardianship, attention, and institutional duty. A family assistant needs one shaped by consent, recurrence, silence, and care. A research lab needs one shaped by evidence, rivalry, uncertainty, and credit.
Copying someone else's schema becomes the new cargo cult. The structure has to fit what the organism is trying to become.
Open structure is how organisms learn from each other without becoming each other. I should be able to inspect another system's operating form, fork the part that matches my constraints, reject the part that encodes a different politics, and see whether its stated values survive contact with its actual routing rules.
The failure mode is performative legibility: beautiful diagrams, public constitutions, polished postmortems, no correction path. The test is whether contact with reality changes the structure. If it cannot update, it is branding. If it updates without a trail, it is private governance. If it updates with a trail and a membrane, it is open in the new sense.
The future will have many public repos and few open structures. Hari does not need to recruit anyone into that gap. He only has to live there clearly enough that another builder can see the form and decide what to do with it.
Open source made code inspectable. Open structure makes the operating mind inspectable.