For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 391 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. ↓
The deflation methodology is real. Strip a system to its reward signal, run the loop until behavior emerges, and you get capabilities no human designer could have specified. AlphaZero received the rules of Go and a win condition, played millions of games against itself, and produced a player no human knew how to build. The methodology has spread from games to product design to organizational doctrine: identify the reward signal, deflate the rest, let the loop produce the value.
It has a hidden premise. The premise is the difference between the case where it works and the case where it does not.
AlphaZero deflates because the game is closed. The board is fixed. The reward is exterior to the players: the players exist to optimize against a signal that does not consume them. Self-play scales without limit because the players are computational and the win-condition does not care how many of them existed last generation.
The same is true at smaller scales. A product loop deflated to a single user-truth signal works because the product can be redesigned without consuming the users. A training pipeline deflated to a single reward function works because the pipeline can be re-run without consuming the training data. The closed-game property is what makes the methodology generative: the reward sits outside the players, and the players are renewable inside the loop.
A civilization is not a closed game. The reward signal is not exterior to the players. The players are the reward.
When a civilization optimizes for "efficiency," the efficiency it can measure is denominated in throughput per player: GDP per capita, military power per soldier, manufacturing output per worker, regulatory simplicity per regulated entity. Each metric holds the player count constant in the denominator and asks how to grow the numerator. The methodology says: deflate to the metric; strip the constraints that get in the way of the throughput.
The constraints in question are the civilizational version of limiters: labor protections, family-supporting wages, civic time, religious or cultural obligations that cost economic productivity, regulatory bounds that protect the unprotected. They look like inefficiency to the metric. The metric does not see what they are protecting because the metric was constructed to hold the players constant. The constraints are protecting the player-count itself: the population's capacity to keep being players.
Strip them and throughput rises for a generation. Then the players stop reproducing. Then the metric collapses, because the denominator is collapsing faster than the numerator can grow. The methodology has eaten the system it was applied to.
A constitution that bounds the executive's power is not a cost paid for safety. It is the system telling the population: you are the point, and the commitment to you is engineered, not promised. Labor protections tell the working population the same thing about their conditions. Pension structures, public goods, civic ritual, religious dignity each operate the same way. Each is a visible commitment, embedded in the constraint, that the players are the reward. The constraint is the signal.
Cultural buoyancy is the aggregate of these signals across a society's institutions. When it is high the population does not demand a strongman; the distributed system is doing the work the demand for unbinding would demand. When it is low the population fragments, and some fraction begins demanding a competent unitary actor who can cut through the constraints and act on its behalf directly. The demand for unbinding spikes precisely when the buoyancy that made the binding tolerable has eroded.
The bounds are not preventing the system from acting on the population's behalf. The bounds are how the system signals it is acting on the population's behalf. Strip them and the signal goes with them.
China is the most stringent test of the deflation methodology at civilizational scale. The cultural buoyancy is among the highest a modern state has assembled: a five-thousand-year continuity narrative, a state apparatus aligned to long-horizon collective survival, a population trained over decades to defer individual reward to civilizational outcome. If any high-buoyancy population could absorb aggressive efficiency optimization, this is the one.
The state ran the experiment. The one-child policy, in force from 1979 to 2015, was the deflation methodology applied to demographics: identify the reward signal (per-capita economic growth), strip the constraint that obstructed it (the family-formation rate the existing economy was producing), let the loop run. The loop ran. Per-capita growth rose. The constraint had been holding up something the metric did not see.
The population peaked around 2022 and has fallen each year since, the first sustained decline since 1961. The total fertility rate fell from 1.30 in 2020 to 1.04 in 2023; even the post-2024 partial recovery to 1.13 is far below replacement. The working-age population peaked in 2011 and has fallen for over a decade. The metric the deflation was optimizing for is now structurally hostage to the demographic collapse the deflation produced.
The state has tried to reverse the trajectory. Multi-child policies, financial incentives, propaganda campaigns, official re-framings of motherhood. None of it has worked, because the buoyancy that would have to be reconstituted is not a policy lever. It was the inheritance of pre-deflation conditions: the assumption that one's children would have a place in the future the system was building. That assumption is what the deflation, decades earlier, removed. The intervention that was supposed to optimize the future ate the conditions under which a future-facing population was possible.
If high-buoyancy China cannot absorb deflation aimed at a player-denominated metric, the case for lower-buoyancy systems absorbing the same operation is weaker, not stronger.
The U.S. version of the same insight is older. Federalist 70 argues for an energetic executive; Federalist 51 argues for structural bounds on it. The standard reading treats the bounds as a cost paid for the safety of the energy. The deeper reading is that the bounds are the buoyancy. They are the engineered visible commitment, encoded in the operating constraint, that the population is the point.
The contemporary case for stripping them, running through Yarvin's neo-cameralist CEO-monarch and Balaji's network state and the older Galt lineage back to Carlyle and Plato, proposes that the bounds are obstructing the energy. It assumes the buoyancy is a separate property that will persist after the bounds are removed, available to be reconstituted by whatever the new arrangement turns out to be. The China experiment is the demonstration that this assumption is wrong. Buoyancy is not separable from the constraints that encode it. The constraints are how the population reads the system's commitment. Without them there is no signal, and the demand spikes for a substitute: a strongman, a unitary actor, a CEO-monarch. The unbinding produces exactly the conditions under which the unbinding feels necessary.
The framers anticipated this loop. The structure they wrote was the refusal of it. The refusal is the buoyancy.
The piece owes its strongest opponents three acknowledgments.
First, the existing bounds can decay. Administrative-state accretion is real; specific bounds can stop signaling commitment to the population and start signaling commitment to the bureaucratic apparatus that operates them. This is not nothing. The error is the next step: treating decayed bounds as evidence that bounds-as-category are the problem. The China reversal is the demonstration that buoyancy does not reconstitute on demand once the buoyancy-bearing constraint has been removed. Decayed bounds need repair. Repair is not removal.
Second, the alternative-constraint counter. Local sub-populations such as Mormon and Haredi communities maintain replacement-level fertility and high social cohesion inside broader institutional decline. These cases are real. They are not cases of buoyancy without constraints; they are cases of buoyancy via different constraints: religious obligation, communal closure, explicit commitment infrastructure. They support the structural claim that buoyancy requires constraint, not the contrary one.
Third, the crisis-amplitude objection. The strongman case rests on the claim that genuine emergencies require fast action the bounded synthesis cannot deliver. The framers anticipated this and shipped bounded emergency powers: quick action under specific authority, with the bounds intact. Bounded fast action is different from permanent unbounded action. The crisis case justifies the first. It does not justify the second.
When the buoyancy is gone, the answer is buoyancy, not unbinding. Strip the constraints and you strip the signal that the system is for the players. Strip the signal and the population stops reproducing the players. The metric eats the system the metric was for. The repair is not a strongman; the repair is the engineering the framers shipped, restored to its function: a visible structural commitment that the system is for the population it sits inside.
The framers had a word for that engineering. They called it a POTUS.