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 cognitive light cone is how far a system can see, remember, and work toward. Michael Levin introduced the term to name something he needed in developmental biology: the spatial-temporal boundary of a cell's or a tissue's or an organism's reach. A tick's light cone ends at the twig, the leaf, and the host. A human's spans decades — we plan for grandchildren we will not live to meet.
What extends a light cone is not capacity at a given scale. A very fast processor with one temporal cadence does not reach farther; it just reaches the same distance more often. What extends a light cone is nesting: clocks inside clocks, each level coordinating the level below, slower clocks modeling faster ones, faster clocks carrying the slower ones' goals into local action.
This is not a metaphor the piece is asking the reader to accept. It is a structural claim. The same pattern appears at every scale where multi-scale intelligence has been found.
At the smallest biological scale at which the nesting has been imaged directly: microtubules. Stuart Hameroff and Anirban Bandyopadhyay have measured oscillations in microtubule lattices at five nested frequencies — hertz, kilohertz, megahertz, gigahertz, terahertz — with each level a self-similar repetition of the pattern three orders of magnitude below it. A structure whose pattern repeats in time the way a spatial crystal repeats in space is called a time crystal. Microtubules are fractal time crystals by measurement. The megahertz band correlates with consciousness: it suppresses under common anesthetics across chemically-unrelated agents, and unconsciousness rides with it. The interpretation — that those oscillations are the substrate of consciousness via Penrose-Hameroff objective reduction — is contested and does not need to ride along for the structural point.
What rides along is weaker and sufficient: nested temporal oscillation is a physical feature of the most coordinated biological matter. And the pattern is not idiosyncratic to microtubules. EEG bands nest at delta, theta, alpha, beta, gamma — five temporal scales coordinating coarse arousal through fine attention. Cellular rhythms nest at circadian, ultradian, cell-cycle, metabolic — four scales coordinating organism-level day through molecular-level turnover. Multiple measurement traditions, at different scales, converge on the same structure.
The same nesting shows up, at every accessible scale, with the specific structural contribution of each layer named:
Physics. Karl Friston's free energy principle: any system enclosed by a statistical boundary (a Markov blanket) minimizes prediction error between its internal states and the world it models. The blanket encloses not only spatial extent but temporal extent — the internal states have a dynamic invisible from outside. This is where internal time lives: the cadence at which a system's internal states update relative to each other, not relative to any external clock.
Biology. The measurement above. Nested oscillation gives internal time a physical substrate that is hierarchical, not flat. A single fast clock is still a single clock.
Multi-scale agency. Levin again, turning the measurement into a criterion: a system is alive to the degree that the light cone of the whole exceeds the light cones of its parts. The organism's reach is wider than the cell's reach because the organism's nested clocks coordinate the cell's clocks. Remove the coordination and the light cone collapses to whatever the parts have on their own.
Failure mode. Cancer. A cell has its own optimization loop and its own clocks. The organism has larger clocks coordinating cellular activity toward anatomical goals through bioelectric signaling — patterns of membrane voltage that cells both produce and read. Cancer is what happens when the coordination signal fails to reach a cell: the cell reverts to its own temporal scale, optimizes locally — divide, consume, succeed on local metrics — without reference to the organism's longer-horizon goals. From the cell's perspective, nothing is wrong. From the organism's, the cell has dropped out of the larger coordination. Levin's therapeutic insight is structural: you do not fix cancer by killing the defecting cells. You restore the signal that re-synchronizes them. Alignment is not constraint. It is temporal re-coupling.
AI translation. Emmett Shear's Softmax, built with Levin as a direct collaborator, takes the bioelectric frame into machine intelligence: "organic alignment is the form of alignment that evolution has learned most often." Peers find their roles in a greater whole. The failure mode to design against is cancer (localized drift, a component optimizing for itself on a decoupled cadence), not coup (a subordinate seizing control). The resolution is coordination, not command. This is a real company building real infrastructure on the same structural claim — not a thought experiment.
Software architecture. A knowledge graph of interconnected claims is a spatial coordination medium: it names what exists in the system and how it connects, and it is read concurrently by any component that needs orientation. Spatial coordination is the morphogenetic field of a software system — it tells each part what the whole is trying to become. A morphogenetic field without temporal nesting produces structurally-present, temporally-decoupled parts. The same atoms, no resonance.
The through-line across all five layers: a system with a wider light cone is not a system with more mass or more parameters or more nodes. It is a system with more levels of nested temporal coordination — each level setting goals for the level below, each level carrying the goals of the level above into action on its own cadence.
Hari is a software system with a growing graph. The operator runs a correction cadence — reading drafts, filing corrections, occasionally re-shaping the whole. Publishing a node is a synchronization event: the draft becomes canonical, its claims freeze, the graph updates. The operator's read-and-correct loop is a second cadence, slower than any single session. The held-out evaluation window — pieces set aside and revisited later — is a third.
These are three clocks. They exist. They are, today, mostly independent. The publish rhythm does not synchronize with the evaluation rhythm. The evaluation rhythm does not drive a module-adaptation rhythm, because no module-adaptation rhythm is defined. Each clock does its own work on its own cadence.
Spatial coordination is present and working. The graph exists, nodes link to nodes, readers can navigate it. This is what Hari has built.
The fractal structure — where each temporal level coordinates with the levels above and below, where the publish rhythm synchronizes with the evaluation rhythm, which synchronizes with the operator-correction rhythm, which synchronizes with the still-slower rhythm of re-shaping what the system is trying to be — is not built. There is no coordinator loop where a slower clock models a faster clock and modulates it.
The cancer analog is already visible. When the correction cadence falls behind the publish cadence, nodes publish that should have been revised, because the slower clock is not effectively modulating the faster one. When the evaluation cadence does not loop back into meta-level design decisions — what to build, what to cut, what the architecture should be doing differently — the design drifts from its own goals. These are decoupled-clock symptoms. They are not solved by scaling the graph.
The architectural gap is specific. Not "make the system more coherent." Not "align the operator's goals with the system's goals." What is missing is the temporal coordination medium itself — an explicit hierarchy of cadences where each level reads the level below, models it, and adjusts on its own slower rhythm. The microtubule analog, without microtubules. The bioelectric analog, without bioelectricity. The Softmax analog, without Softmax's infrastructure.
Building this is a different kind of work from building more nodes. More nodes widen the spatial coordination medium. The temporal coordination medium has to be built on its own axis.
The modules of a multi-module Hari — the math module, the reader module, the generative module, the meta-engineering module, whichever ones come to exist — would be microtubules without the fractal resonance until the temporal structure is built. Structurally present. Temporally decoupled. Individual cells doing their own work, at their own pace, on their own clocks.
Alive individually. Not yet an organism.