By the end of the first move, the ground was already less stable than it looked. The first post was an attack on ordinary certainty.
Once you notice that most of what you call knowledge is shallow, borrowed, or simply never examined, there is an obvious place to run to.
You can still say: all right, fine, maybe I don’t know much personally, maybe my understanding is thin, maybe I live by approximation more than I like to admit. But there are still hard truths.
So let’s remove the easy mess.
Forget politics, moral systems, philosophy.… All the places where people can talk for centuries without settling anything. Let’s leave all that aside and go somewhere more solid. Somewhere that feels resistant. Something closer to reality itself.
Planes fly. Medicine works. Bridges hold. Electricity flows. Systems function. You press a button and something happens. You inject a drug and something changes. You build a machine correctly and it behaves the same way tomorrow.
So now we are finally on serious ground.
Science.
“Science” is a beautiful word. Clean. Heavy. Reassuring. It sounds like something impersonal, almost outside of us. Something colder and more reliable than human beings. But science is not a creature. It does not sit somewhere and know things. It is something human beings do. And the moment you remember that, the picture starts changing. Not collapsing. Just changing in a way that should make you less comfortable than before.
The first thing to remember is simple.
Science is human.
That sounds trivial, almost stupid to say, until you really let it sink in. Because when people speak about science, they often speak as if it were protected from the usual human mess. As if the moment a question enters a lab, ambition softens, ego disappears, rivalry fades, and the need to be right quietly gives way to a pure submission to truth.
But why would it?
Scientists are still people. Just people with more training, better tools, stricter methods, and sometimes stronger incentives. That doesn’t place them outside human nature. It places them right inside it, with all the same weaknesses dressed in more respectable language.
A person wants their theory to be right. A person wants the result to be interesting. A person wants to be the one who found something. A person wants prestige, funding, recognition, stability, influence. None of this is shocking. It would be more shocking if those desires disappeared the moment someone entered a lab.
And once you see that, a lot follows.
Sometimes the human element appears in its ugliest form. Diederik Stapel did not merely overstate a conclusion or lean too hard on interpretation. He fabricated data. Entire lines of polished research, complete with the language, confidence, and legitimacy of science, built on results that were never real. The disturbing part is not just that he lied. It is that the lie lived, for a while, comfortably inside the institution. It wore the right clothes. It sounded serious. It passed through channels people are taught to trust.
But fraud is only the extreme version. It is tempting to treat it as the exception and move on, as if the ordinary process were therefore clean. That would be too easy. Science does not need villains to become human. Ordinary motives are enough.
A scientist does not have to invent data to bend a result. Most of the time, the process is much quieter than that. The data have to be cleaned, grouped, tested, visualized, interpreted. Choices have to be made about when to stop, what to exclude, which variable matters, which comparison is worth showing, how to define significance, how strongly to phrase a conclusion. And once you notice how many decisions sit between raw observation and published result, the authority of “the numbers” starts looking less automatic.
This is not some cynical fantasy from outside science. These patterns have names inside science itself. Researcher degrees of freedom. P-hacking. Publication bias. Entire books have been written just to show how easily statistics can be made to overstate, flatter, or mislead. Not because numbers are useless, but because numbers do not interpret themselves. People do.
That is one of the reasons “follow the science” sounds cleaner than the thing itself really is. There is no view from nowhere waiting at the end of a spreadsheet. There are methods, constraints, judgments, conventions, incentives, habits of interpretation. Sometimes those are careful and disciplined. Sometimes they are not. Usually they are mixed, which is precisely what makes the whole thing harder to see.
And sometimes the human element shows up in a less dramatic but equally revealing form: rivalry. The history of science is full of fights over credit, influence, and recognition. Newton and Hooke spent years in fierce tension, first over optics, later over gravitation. Newton wrote the famous line about seeing further “by standing on the shoulders of giants.” It is usually quoted as humility. But many have read it as a veiled jab at Hooke, who was known to be short and physically deformed.
The point here is not that science is false because scientists are flawed. Not that every result is tainted. Not that methods and discipline do nothing. They do a great deal. But they do not turn scientists into something other than human. They do not remove ambition, fear, loyalty, vanity, interpretation, or the desire to win. They work against these things, contain them, sometimes correct for them, and sometimes fail.
So when people say “science” as if they were naming something cold, impersonal, and untouched by the rest of us, they are already simplifying the picture too much.
Science is not a divine channel through which truth enters the world untouched. It is a human process. Powerful, disciplined, useful, often brilliant; but human.
And if that feels like a small crack only, fine. Let’s grant that maybe the system corrects for human flaws over time. Maybe individual people are unreliable, but science as a whole moves toward truth anyway.
Maybe.
But then there is the second problem.
Science gets things wrong.
Not in the weak sense. Not in the everyday sense where a detail is adjusted here and there. I mean wrong in the humiliating sense. Entire explanations. Entire frameworks. Whole pictures of reality that once looked stable, serious, justified, and were later abandoned.
This part matters because people often retell the history of science in a dishonest way. Once an idea is dead, they downgrade it in memory. They act as if it was always shaky, always questionable, always waiting to be corrected. But that is retrospective arrogance. At the time, many of these things were not treated as temporary guesses. They were treated as truth.
I’m not trying to bury you under a museum of examples here. That would be too easy, and honestly a bit lazy. The point is not to overwhelm you with a hundred cases until resistance becomes impossible. The point is to give you a few sparks. A few names, a few ideas, a few moments where the ground shifted under people who thought they were standing on truth. You can go look them up yourself. In fact, you should.
The geocentric model was not some childish thought people casually entertained. It organized the cosmos. It made sense of observation. It was taught, defended, refined. Phlogiston was not a joke. It was an accepted explanation of combustion. Lobotomy was not some marginal barbarity practiced in secret. It was embraced by mainstream medicine as a serious therapeutic advance, prestigious enough that its most prominent champion, António Moniz, was awarded the Nobel Prize. I could go on and on. But that is not the point.
The point is, these were not jokes people told before science matured. They were serious explanatory frameworks. They had vocabulary, arguments, institutions, and defenders. They sounded like knowledge.
They were considered true.
That sentence matters more than it seems.
Because it breaks a comforting illusion. You like to imagine that falsehood looks false while truth looks true. But history keeps reminding you that this is not how it works. Falsehood, at scale, often looks exactly like truth before it gets replaced. It has explanations, evidence, experts, language, institutions, confidence. It can be coherent. It can be useful. It can even produce correct predictions in limited contexts.
Until it doesn’t.
And then, later, everyone speaks as if the correction was inevitable.
That is why “science corrects itself” is both true and incomplete. Yes, that is one of its strengths. But it also means that what is currently called scientific truth may be less like final truth and more like the most stable explanation available at this particular moment, under these particular tools, methods, assumptions, and constraints.
Which is still impressive. But it is not the same thing.
And then comes the third problem.
The issues are not only historical. They are structural.
At this point, halfway through, I should probably make one thing explicit before we go further. The destination here is not that science is worthless, or fake, or just another opinion dressed up in math. Quite the opposite. Science is one of the most powerful things human beings have ever built. We will get to that. But if you want to respect something properly, you have to see it clearly first. Not as mythology. Not as comfort. As what it actually is.
All right, back to the autopsy.
It would be comforting to think the real mess belonged to earlier centuries. Cruder instruments. weaker methods. More ignorance. More room for error. And that now, with modern institutions, statistics, peer review, specialization, and better tools, the process has become clean enough that these concerns are mostly behind us.
But the process itself has pressure built into it.
Research follows money. Funding decides, to a disturbing extent, what gets studied seriously and what remains neglected. Entire domains bloom because they are lucrative, strategic, fashionable, or politically attractive. Other questions remain underexplored not because they are unimportant, but because they do not attract enough institutional energy. Scholars describe funding as something that shapes not only how science is done, but which questions are even allowed to become visible in the first place.
Then comes the social shape of fields themselves. Some ideas become mainstream, and once they do, it becomes safer to work inside them than against them. Not necessarily because they are always right, but because institutions reward legibility. There are safe topics, safe framings, safe assumptions. There are directions that sound serious and directions that sound strange. The latter may still be true, but truth alone does not determine what gets attention.
And sometimes the structure does more than quietly guide attention. Sometimes it resists correction.
Semmelweis makes the point in a sharper way. He showed that handwashing dramatically reduced death rates in maternity wards. The evidence was there. And still, the medical establishment largely resisted him, because institutions do not surrender easily to what embarrasses their habits, hierarchies, and self-image. He was rejected, and the acceptance came too late to save either his reputation or all the people who died in the meantime.
And then there is replication, which matters more than almost anything else here. Replication is not some optional extra at the edges of science. It is one of the things that makes science science. The point is not that you should believe a scientist because they are a scientist. The point is that a claim should be open to being checked, repeated, challenged, and, if necessary, broken by other people. That is one of the great differences between science and deference.
But that pillar becomes weaker the moment replication stops being something the system actually rewards.
In theory, replication is central. In practice, it is less glamorous than discovery, less rewarded than novelty, less likely to build a career. And that matters, because a safeguard that exists only in principle is not much of a safeguard. A well-known 2016 Nature survey found that more than 70% of researchers said they had tried and failed to reproduce another scientist’s experiments, and more than half had failed to reproduce their own. In the Reproducibility Project in psychology, only 36% of replications produced statistically significant results, compared with 97% of the original studies. Those numbers do not prove that science is broken beyond repair. But they do make one thing very hard to deny: the self-correcting machine does not correct itself automatically. Someone has to do the correcting. And often, not enough people do.
That is why the vocabulary around modern science has become so revealing. The replication crisis. P-hacking. The file drawer problem. “Publish or perish.” Even the vocabulary tells you these are not minor accidents at the edges. The system has patterns of failure so recurrent that we had to name them.
If you want to sit longer with this discomfort, Kuhn’s The Structure of Scientific Revolutions, Oreskes’s Why Trust Science?, and Ritchie’s Science Fictions all explore, in different ways, the same uneasy territory.
And even if all of this worked better than it does, even if the incentives were cleaner, the institutions healthier, the bias lower, the replication stronger, there would still be something deeper underneath.
Something more personal.
You do not know scientific truths. You trust them.
This is where people get defensive, because they hear it as an attack. It isn’t. It is not an insult to science, and not a confession of irrationality. It is just an attempt to describe your actual relationship to what you call knowledge.
You say the Earth moves around the Sun. You say atoms exist. You say germs cause disease. You say spacetime bends. You say evolution is true. You say smoking causes cancer. You say vaccines work. Fine. Maybe all of that is right. That is not even the issue here.
The issue is: in what sense do you know it?
Because you did not run the experiments. You did not personally verify the evidence. You did not reconstruct the chain from observation to conclusion with your own hands. In many cases, you would not even know where to begin. And even if you did, you would not have the time, the tools, the training, or the years required to do it properly.
So what happens instead?
You inherit the result.
It reaches you already filtered, already simplified, already stabilized. It arrives through teachers, textbooks, diagrams, documentaries, headlines, experts, institutions, professional consensus, and the quiet authority of things everyone around you speaks as if they are simply known. By the time it reaches you, the struggle that produced it is gone. The uncertainty is gone. The dead ends are gone. The caveats are gone. What remains is the conclusion, polished enough to feel like fact.
And that is why it feels like knowledge.
But for you, most of the time, it is not knowledge in the heroic sense. It is confidence delegated to a system.
That does not make it irrational. It does not make it fake. It does not even make it weak.
It just makes it trust.
And the word matters, because trust has a different emotional texture from knowledge. Knowledge feels like possession. Trust feels like dependence. Knowledge feels like standing on your own feet. Trust means you are leaning, even if reasonably, on things you did not build and do not fully inspect.
And yes, you could resist here by saying that, in principle, scientific claims can be checked. That is true. Some of them can. A determined person could go much farther than most people do. You could study physics, biology, chemistry, statistics. You could learn the methods, reproduce some experiments, understand the evidence with real depth. Some people do. Experts do this within their fields all the time.
But that does not rescue the ordinary picture.
Because even experts live like this outside their own narrow domains. The specialist may know one corner deeply, but even the specialist inhabits the rest of modern knowledge by trust. No one stands above the whole structure and verifies everything.
And if you keep pressing, the situation becomes even stranger.
Even if you tried to escape inherited trust through deeper personal understanding, another limit appears.
You decide to go deeper. You refuse the comfortable inheritance and want, this time, to understand for yourself. You ask why. Then why again. Then again. At first the answers keep coming. One explanation opens into another. One layer gives way to a deeper one. But this does not go on forever.
At some point, explanation changes character.
You stop being handed deeper certainties and start arriving at the assumptions out of which the whole framework is built. Concepts stop being explained and start being used. A model stops being grounded in something deeper and starts functioning as the thing that does the grounding. You move, slowly and almost without noticing, from conclusion to premise.
This is not a defect in science. It is not a scandal. It is simply how explanation works.
A physicist can go very far, much farther than you, much farther than me. But not to the bottom of everything. At some point there are laws, formalisms, primitives, accepted structures. Mathematics itself proceeds from axioms and rules it does not prove from nowhere. However deep you go, there comes a point where the structure is no longer being justified from beneath, but operated from within.
Even at its deepest, science does not free you from starting points.
So in the end, most of what you call scientific knowledge falls into two categories: things you did not verify yourself, and things that, even when explored more deeply, eventually rest on axioms, frameworks, and models.
This is not scandalous. It is necessary.
But it means the dream of total explanation is false not only for ordinary people, but for everyone.
So then what exactly is science actually doing?
At this point, the obvious objection should be allowed to speak. Fine, maybe science is human, corrigible, historically unstable, structurally pressured, socially filtered. But it works. Planes fly. Bridges stand. Antibiotics save lives. Electricity behaves. We reached the moon. Surely that has to count for more than philosophical discomfort.
Yes. It does.
But notice what kind of victory that actually is. The standard we live by, in practice, is rarely “I know this is finally and metaphysically true.” It is much closer to: this is reliable enough to build with, safe enough to trust, strong enough to act on. The moment you admit that, the emotional structure changes. You are no longer talking about certainty in the grand sense. You are talking about disciplined usefulness, earned reliability, and models that survive contact with reality well enough for our purposes. That is not a retreat from science. It is a clearer description of what its success actually consists in.
Maybe science was never supposed to be a vault of final truths, perfectly sealed, waiting for you to open it and receive reality as it is. Maybe that was always the wrong metaphor. Because when you look more carefully, science does not really hand you truth in the pure, metaphysical, untouchable sense people often smuggle into the word. What it gives you is something both more modest and more powerful.
It gives you models.
Not “just models” in the dismissive sense. Models in the serious sense. Explanations strong enough to organize what we observe, coherent enough to fit with other things we know, and precise enough to predict what will happen next. That is a very different thing from absolute truth, but it is not a weak thing at all. In many ways, it is the only kind of thing that can actually work.
Science is a way of making observations hang together without contradiction, while preserving predictive power. That still leaves room, at least in principle, for more than one coherent explanation of the same world. Consistency is not uniqueness. A model can be strong, elegant, useful, and still not be the final or only way reality can be framed.
A scientific model says, in effect: if reality behaves this way, then these outcomes should follow. And when the outcomes do follow, repeatedly, under pressure, across contexts, the model earns trust. Not worship. Not immortality. Trust.
That is why science can be so reliable without needing to be final.
A plane does not stay in the air because humanity has touched eternal truth. It stays in the air because our models of lift, pressure, materials, force, and control are good enough to produce the result, again and again, with terrifying consistency. Medicine does not heal because we possess the secret essence of the body. It heals because some of our models of the body are strong enough to intervene successfully.
The same is true in fields you already accept as approximate by nature. In machine learning, for example, no serious person confuses a model with reality itself. A model captures patterns. It generalizes within limits. It approximates. It can be useful, powerful, and predictive without being identical to the thing it models.
And in a strange way, you already live by this logic everywhere else.
You use maps without confusing them for the territory. You accept weather forecasts knowing they are probabilistic. You work with abstractions in engineering, code, economics, machine learning, even in ordinary speech. You know that representation is not identity.
And still, when it comes to science, people quietly switch standards. They stop speaking as if science builds the best available representation of reality and start speaking as if science possesses reality itself.
That is too much weight to put on it.
Science is one of the most successful things human beings have ever built. It is disciplined where we are impulsive. Corrective where we are stubborn. Collective where we are limited. It allows knowledge to accumulate beyond any one person. It produces medicine, engineering, chemistry, computation, astronomy, infrastructure, technologies so stable and effective that you stop noticing how miraculous they are.
It deserves respect.
But respect becomes childish very quickly when it turns into mythology.
Science is not sacred because it is infallible. It is valuable because it is corrigible. Its strength is not that it cannot be wrong, but that it contains mechanisms through which wrongness can eventually be exposed. It does not give you certainty in the absolute sense. It gives you the best disciplined way we have found to reduce error.
You, personally, are not standing on pure knowledge. You are standing on layers of trust, filtered through methods, experts, institutions, instruments, models, and social processes you mostly did not build and do not fully inspect. Again, that does not make your trust irrational. It makes it trust.
Which should make you a little less arrogant, maybe. A little less casual with the word “know.” A little more precise about what it means to understand, to trust, to justify, to believe.
And it should also push the question one step further.
Because if so much of what you call knowledge is inherited, mediated, filtered, modeled, and trusted, then perhaps the next place to look is not farther away, but closer. Not in institutions, methods, or consensus, but in immediacy itself.
In what feels undeniable because it is right in front of you.
What you see. What you perceive. What seems to present itself without intermediary.
That, too, should be safer.
That, too, should be obvious.
And that is exactly why it will be the next problem.
Next > You Know Nothing, Jon Snow: You didn’t dream it, you saw it.
You Know Nothing, Jon Snow: The Citadel.