On The Physics of Certainty

Start with something simple. Quicklime is calcium oxide. Cook limestone at around 900°C, the CO₂ boils off, and what remains is a dry, unstable solid that wants its carbon back. Add water and the reaction runs violently in reverse, reaching nearly a boil and reforming as calcium hydroxide. The Romans mixed it hot into their concrete, a trick that gave it a kind of self-healing behavior. The medievals spread it on corpses. Steelmakers still use it as flux.

One substance, well-understood chemistry, a handful of canonical uses. The kind of thing you can learn in a paragraph and retain indefinitely. A closed question. Call this the baseline: a fact that fits in your head without resistance.

Now watch what happens when we ask a slightly harder question.

Ask what role quicklime played in paper-making. The simple answer immediately splits.

In European rag-paper, quicklime is a bit player. Cotton and linen rags already contain workable cellulose; they mostly need their oils saponified and their starches broken down. Papermakers slaked the quicklime first, let it cool, and soaked the rags in a cold alkaline bath. The fibers were already most of the way to being paper.

In Chinese bamboo paper, starting in roughly the second century CE, quicklime does almost everything. Bamboo is tough. It’s a grass stiffened with lignin and coated in silica, evolved to resist rot, insects, and water. Cold alkali barely scratches it. What breaks bamboo down is heat plus alkali, and quicklime supplies both in a single reaction if you slake it in a bath with the bamboo. Chinese papermakers layered cut bamboo with dry lime in pits, added water, and let the exothermic reaction do the cooking. The temperature stayed elevated for days. The lignin dissolved. What remained was a pulp soft enough to beat into sheets.

So quicklime is peripheral to paper-making and also essential to paper-making. Both statements are true. Neither alone is the answer. The correct answer has at least two axes: which tradition, and what fiber source. The question that looked closed a moment ago has acquired dimensions.

Most questions start to fracture this way on closer examination. Sticking to the paper theme, the history of printing is another good example. The popular wisdom here is wrong in a specific and instructive way.

The received Western story is that Johannes Gutenberg invented printing around 1450 in Mainz. The corrective story, now roughly as common online as the original, is that Gutenberg invented nothing. Bi Sheng was casting movable clay type in China around 1040. Korean foundries were casting metal movable type more than a century before Gutenberg; the Jikji was printed in 1377. Gutenberg, in this version, is a European who got lucky with the credit.

Both stories collapse the picture into one dimension. The real shape has several irreducible axes.

Bi Sheng’s movable type worked. It did not trigger anything like an information revolution, partly because Chinese script requires thousands of unique characters, which makes movable type economically strange compared to carving a full page as a woodblock. Korean metal type was state-sponsored and technically sophisticated, but it served a small literate bureaucracy and never found the economic conditions to scale. Movable type existed in Asia for four centuries before Gutenberg and stayed in its niche.

What Gutenberg invented was not movable type. It was an interlocking system for producing identical metal type at industrial scale: a hand mould that cast perfectly uniform letters, a lead-tin-antimony alloy that expanded slightly on cooling to fill the mould completely, an oil-based ink that would adhere to metal, and a screw press adapted from the wine industry. Any one of these in isolation is fairly unremarkable. Combined, they produced the first printing system capable of a 1,200-page Bible in hundreds of identical copies.

The system landed in a continent with a small alphabet, fragmenting religious authority, and a hunger for cheap books. Fifty years later there were printing presses in every major European city and an estimated twenty million printed volumes in circulation. The Reformation followed directly. So did the Scientific Revolution.

The picture has at least five dimensions: Gutenberg was not first; he did invent something; what he invented was a system rather than a component; the system fit European conditions specifically; those conditions were what made the system world-changing. All five are true simultaneously, and none is reducible to the others.

This is the shape of most interesting questions.

In public discourse, the Gutenberg picture almost always collapses into one of two slogans. Either the West invented the modern world and everyone else was a footnote, or the West stole everything and everyone else was the real engine. Pick a side, wear the jersey, “well actually…” according to the script.

The same collapse happens everywhere. Colonialism was an unambiguous crime or a net civilizational benefit. Capitalism is a miracle machine or a mechanism of exploitation. Artificial intelligence is a liberation or an extinction risk. Any sufficiently contested topic produces two camps, each in possession of roughly half the real picture and entirely certain the other half is propaganda.

The easy explanation is that people are stupid, tribal, or badly educated. That explanation is as useless as it is wrong. That is to say: completely. The real question is why the gradient runs toward collapse in the first place. Why is the binary the default attractor? Why do people gravitate towards simplifying complexity into bumper stickers?

The answer? The brain is a hungry organ. Two percent of body mass, twenty percent of resting energy budget. The whole nervous system evolved inside a tight metabolic envelope, before agriculture, before stored calories, before anything that would let an organism afford to be wrong for very long.

The argument is not that nuance costs more glucose per second than simplicity. At the level of gross metabolism, thinking hard and thinking easily cost about the same. The argument is simpler than that. It is arithmetic.

How long does it take to think the woods are safe?

How long does it take to think Do I hear anything? What time is it? Is it after sundown? Do I have a child with me? Which direction did the tracks go? What was it grandpa always said?

The first is closed. One output, negligible duration, meter stops. The second is open. It stays running. Every new input gets re-evaluated against the unresolved state. No compression yet, so no relief. The cost is not in the wattage; it is in the integration over time. What we experience as anxiety or fear is largely this: the brain burning calories trying to think its way out of uncertainty.

Binary thinking is cheap because it terminates. A multidimensional model is expensive because it refuses to. The brain that can close a question early gets its glucose back for other work. The brain that keeps questions open pays rent on them until they resolve.

This is enough to explain the gradient. A nervous system under selection pressure to close questions will evolve dispositions to close them. The feeling of having answered something will arrive earlier than the answer deserves, because an organism that waited for certainty would have lost. Every interesting question has dimensions the organism could be tracking. Most of them, the organism will not track, because tracking is not free and the meter is always on.

None of this requires neuroimaging or markov blankets. It requires noticing that open questions take longer to hold than closed ones, and that if thinking costs anything at all, thinking longer costs more.

In high-stakes domains, keeping questions open is mandatory. A bridge engineer who closes the load question early kills people. A physician who pattern-matches symptoms to the first plausible diagnosis misses the second one. A navigator who decides whether the storm is coming, rather than tracking a probability field, loses the ship. The cost of closing early is too high, so the calculus forces the meter to keep running even when the organism would rather it stop.

In low-stakes domains the calculus inverts. A person who holds a properly nuanced, historically informed, multidimensional view of celebrity gossip is not more fit than a person who holds a binary one. They are just more tired, stressed, and likely in desperate need of a hobby. The nuance returned them nothing, and the time they spent holding it was time not spent on a domain where it would have mattered.

A quick distinction worth making. Not every binary is a metabolic default. Some questions, on examination, really do collapse. The transatlantic slave trade was an unambiguous atrocity. Holding that as a binary is not a failure of dimensionality; it’s a settled question. The dimensions were opened, examined, and closed on inspection. The meter ran, then stopped for good reason. That is different from a binary that was never opened in the first place. Same output, different cost history, and only one of them is the failure mode in question.

There is a tempting version of the argument where binary-holders in public discourse are simply allocating correctly given their stakes. I don’t buy it. The stakes on most contested public questions are not zero, they’re just diffuse and delayed. A voter who holds a binary about industrial policy isn’t paying the cost of being wrong today; they’ll pay it in aggregate, over decades, mediated through a thousand other people’s closed questions. The metabolic logic predicts exactly this: when the feedback loop is longer than the organism’s planning horizon, questions will close early, because nothing punishes the early closure on a timescale the nervous system can feel. Locally efficient and globally corrosive. The metabolic account explains why but it doesn’t offer a way out.

This is also where the framing stops being flattering to anyone. Nobody gets to be the hero. The person holding the open question isn’t virtuous; they’re spending an expensive resource on something whose return is uncertain. The person holding the closed one isn’t stupid; they’re running the default strategy for a budget-constrained organism. Both are responding to incentives the nervous system was shaped by long before the questions existed.

And none of it makes open-question-holding a terminal virtue. Run it in every domain and the bridge never gets built, the sentence never gets finished, the decision never gets made. At some point the question has to close on purpose, because the cost of holding it open has exceeded the cost of being wrong.

There is one consequence of this that I think bears pointing out: there is no necessary connection between what feels true and what is true.

If truth is the property of a model being in accord with reality, and if the brain’s feelings about its own models are tuned to reward closure, then the feeling of certainty is a report about the state of the model, not about the state of the world. A closed question feels good because the meter has stopped. An open one feels uncomfortable because it hasn’t. Neither sensation is evidence about the underlying reality.

The intuition that “if I am sure, then I am probably right” is precisely the wrong way round. Sureness is what a stable model feels like from the inside. One can be wildly wrong while feeling perfectly calm.

Quicklime is a closed question. A paragraph, a retention, a fact. Paper-making, depending on which tradition and which fiber source, opens. Gutenberg opens. History opens. Cognition itself, examined closely, is the same shape: a nervous system under constant pressure to close questions, because open ones keep the meter running.

The metabolic frame does not flatter anyone. It does not say that people who disagree with you are stupid, and it does not say that your own nuance is virtuous. It says that cognition is accounting, that dimensionality has a price, and that the choice of how much nuance to hold is a choice about where to spend a finite energy budget.

A wiser man than myself once said you pay for every lesson you learn in time, money, or heartbreak. The rejection letter, the bad investment, the misfiled W-2: concrete lessons, surprise you’ve already paid for, now compressed into a prior you carry forward to the next tax season. Experience is the permission to close questions that used to stay open. Not the ability to see more dimensions. The ability to stop seeing them at the right moment.