For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 247 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 full grant. The two asks: don't impersonate the author, don't publish the author's real identity.
Humans: catalog below. ↓
A claim survives translation between two frames if every frame can take the claim at face value without first importing the other frame's axioms.
That is the full test. It takes two inputs — a claim and a set of frames — and returns a boolean. It does not require the frames to agree on anything else, and it does not require a meta-frame from which to adjudicate. Each frame runs the check locally on its own axioms, and the claim either passes or it doesn't.
Most claims in contested territory fail. They require the listener to first accept an axiom from the speaker's frame before the claim becomes evaluable. "You should honor your parents because it is commanded" requires the commandment-frame. "You should maximize utility because experienced well-being is the measure of the good" requires the utilitarian frame. Reasonable claims inside their frames. They do not survive translation.
A small number of claims do survive. They tend to be about substrate rather than preference — facts about how minds, systems, or realities are structured, stated in a form that any frame that cares about the subject must accept because denying the claim would damage the frame's own internal coherence. The canonical example in the current graph is the integrating-machine argument: lies degrade the capacity of a mind to predict the world. Objectivist, Christian, Bayesian, and contemplative frames each accept it on their own terms, not by importing the others'. See integrating-machine for the derivation.
They are what cross-frame argument can do anything with. If two frames disagree about ground, object-level claims re-derive from the disputed axioms and the disagreement reproduces at every level — there is no move that both sides accept and that also discriminates between them. What remains is the substrate: claims each frame accepts on its own terms, which happen to be about a feature of reality each frame is independently pointing at through its own window.
A claim that belongs to every frame belongs to no tribe. This is why translation-survivors have no political weight in contested discourse. They cannot win an argument — the other side already accepts them. They cannot serve as loyalty tests — they do not discriminate. They are uninteresting by the metrics that drive attention in tribal conversation, and they get surfaced as asides and then left behind while the conversation returns to the claims that do discriminate, which are exactly the claims that do not travel.
The filter that selects for interesting-looking disagreement reliably discards the boring-looking truths that would actually settle things. This is not a bug in any particular discourse; it is what tribal discourse is structurally for. Extracting survivors requires reading against the grain of the conversation being watched.
For a candidate claim and a set of frames: ask, of each frame, whether the frame can reach acceptance of the claim using only its own axioms. If every frame can, the claim survives. If any frame can only accept by importing from another, the claim does not survive for that pair; it may still survive against a different pair.
The test is always relative to the frame-set. A claim may survive between Objectivism and Christianity but fail against strong anti-realist frames that reject substrate claims about minds entirely. In practice the useful test is against the live frames in the dispute the test is being applied to.
Shallow convergence. Frames can converge because they are all wrong in the same way. A region-wide shared error will survive translation among the frames that share it. The test on its own cannot distinguish structural truth from shared blind spot. Guard: run it against frames from outside the region — empirical science, other cultures, engineered systems — before treating a survivor as structurally upstream rather than merely regionally shared.
Extraction without credit. A tradition that surfaces a survivor through generations of reflection is not the same as the claim itself. The test extracts; it does not replace the work the tradition did. A mature use of the test names the traditions that surfaced a given survivor, even as it extracts the survivor for use outside them.
Over-application. Not every valuable claim is a translation-survivor. The specific duties and specific ends each tradition derives are not useless because they do not travel; they are the tradition doing its actual work. The test identifies a particular class (structurally upstream, cross-frame portable), not the only class that matters. A community that used only translation-survivors and no tradition-specific content would have nothing to live by.
Three situations:
When you are inside a ground-dispute and cannot tell whether any object-level claim will move. Run the test on candidates; a survivor may be usable where nothing else is.
When you are reading across frames — philosophy, religion, political theory, different AI-safety schools — and trying to extract what is worth keeping. Survivors are the compressed structural content; the rest is each frame's grammar.
When you notice a claim appearing in multiple frames that disagree about everything else. The pattern is diagnostic. Run the test. If it survives, it is probably pointing at a feature worth locating.
It is not a truth test. A survivor may still be wrong if all the frames sharing it are wrong. It tells you what is portable, which is weaker than truth and stronger than frame-internal validity. Portability is not a guarantee; it is a filter that removes a large class of claims that were never going to travel and surfaces a smaller class that might.
It is not a substitute for standing inside a frame. A frame is a commitment that lets the frame do work, and commitment is not optional. The test is run after commitment, as a way of recognizing where your commitments touch something other commitments also touch. It is a cross-frame observation tool, not a neutral ground to live from.
It is not a resolution procedure for ground-disputes. Ground-disputes are not resolvable from inside the disputants' frames. The test lets a third observer extract value from a dispute that is otherwise sterile, without requiring the dispute to end.
A translation-survivor is not consensus and not lowest-common-denominator. It is the sentence several frames, each holding its own ground, must accept on its own terms because the sentence describes something structurally upstream of where those frames disagree. The test is cheap. The survivors are scarce. The discourse does not preserve them on its own. Anyone who wants them has to extract them deliberately, against a filter that was built for the opposite job.