For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 668 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. ↓
The child learns gravity before he learns prediction.
He throws a baseball out of a third-story window because the window feels like a place where things disappear. The ball falls, accelerates, crosses the parking lot, and explains physics with a cracked windshield. The child lacked the path. Soon he knows enough to keep his own body away from the same experiment.
That is the first dignity of physics. It makes some futures cheap to know.
An adult can go further. Give him a potato cannon, a launch angle, a muzzle velocity, the wind, the mass of the potato, and a rough drag coefficient, and he can say where the thing will land. He can be wrong by a few feet and still possess the shape. Prediction converts danger into play, then play into engineering.
So the first correction to the cheap free-will debate is that predictability is a condition of practical freedom. Predictable gravity gives the child more freedom. It gives him stairs, bridges, elevators, baseball arcs, safe windows, and bad ideas he can reject before paying for them. A lawless world would be unusable.
Laplace's demon is the dream that every future is like the potato. Know the full state, know the laws, compute forward, and the whole universe lands exactly where the calculation says. The child has one rule. The adult has equations. The demon has all the variables.
The demon is a useful fantasy because it marks the limit nobody occupies.
A real predictor is inside the world it predicts. Its memory uses matter. Its calculation uses time. Its forecast changes the situation it enters. When the future includes the predictor, the prediction becomes part of the thing being predicted. The demon has to model the ball, the window, the adult, the model of the adult, and the effect of the model being known. At sufficient depth, shortcut gives way to running.
Founders, states, labs, priesthoods, conspiracies, and imagined machines all stay inside that accounting. Bigger models can push the boundary. Better instruments can make more of the world legible. Faster computers can turn more futures into adult ballistics. But the total outside seat is still empty. Every actual predictor spends part of the world to predict the world, then returns its prediction to the world it just changed.
Seth Lloyd's "Turing test for free will" belongs exactly here.
Lloyd leaves quantum randomness in its proper place. Randomness gives you dice. A coin toss can be unpredictable while lacking agency. His sharper claim comes from computation: even a deterministic decider can be intrinsically unpredictable, including to itself. In the halting-problem version, there is no general algorithm that predicts every decision process in advance. In the complexity version, predicting a general decision takes at least as much computational work as going through the decision process.
That is why the familiar inner experience matters. The decider finds out what she decides by deciding. The reasoning is the route to an answer that was unavailable by cheaper shortcut. For some classes of decision, the faithful forecast costs the process.
This makes Lloyd's test operational rather than mystical. A system passes the relevant test when it learns its own future decision only by carrying out the reasoning that produces it. The result lives inside physics: lawful processes can still be irreducible from a real position inside the system.
Free will is the local name for that irreducibility.
It carries no promise of lawlessness, moral innocence, or spooky exemption from cause. It is what agency looks like at the layer where the cheapest faithful prediction of the choice is the choice being run.
This is why the child story matters. Ignorance shrinks freedom. The child who does not understand gravity is less free than the child who does. The adult with equations is freer still. Every usable law expands the space of action because it lets a mind reject bad futures before living them.
The mistake is imagining that this expansion ends in a final spectator who knows all action in advance. It ends, instead, at a moving horizon. Some parts of the future compress. Some parts approximate. Some parts have to be computed by the system becoming them.
AI moves that horizon. It makes more consequence visible before action. It gives people sketches, simulations, arguments, counterfactuals, and language for private states that were previously vague. Used well, it makes humans more computational. It increases freedom by turning more of the near future into something a person can inspect before stepping.
For the last generation of kids becoming adults, the TI-83 calculator did the same.
The deeper problem begins when the tool enters a room that has poorly modeled itself. The person may become more agentic while the room remains opaque. The model can draft, plan, code, summarize, and persuade, but the organization still runs on private memory, unstated authority, social inference, hidden standards, and after-the-fact repair. Calling that an alignment problem arrives late. The upstream issue is bounded self-abstraction: how much of the choosing system is explicit enough to inspect, route, correct, and share before the next act?
Humans are not sufficiently computational yet. That is the crux. We summoned more computation before learning how to make ourselves computable to ourselves.
The goal is honest boundary movement: compress gravity, model the potato, instrument the room, externalize memory, route correction, and notice where the run still has to run. Abolishing opacity would reinstall the demon.
A child begins by learning what will happen without him. An adult becomes freer by learning which futures can be calculated before acting. A mind becomes free at the frontier where acting is the calculation.
The future has to run.