Instruct, Thinking, and the Small Fanaticism Inside Our Cognition

that night, i only wanted to understand why my small model felt so stupid.

not stupid in a funny way. not even stupid in a way i could fully forgive just because it was small and running with a short breath on a local machine. more precisely, it was stupid in a polite way.

obedient, neat, never refusing, never arguing, yet every answer felt like someone nodding without actually listening.

i was testing an instruct model.

technically, i knew its limits. a model like that is designed to answer directly. it receives an instruction, catches the nearest pattern, then produces a response that sounds reasonable. for many things, that is enough. cleaning up text, answering simple questions, rearranging sentences, explaining something quickly, or becoming an artificial conversation partner that never gets tired.

the problem appears when it is asked to do something that cannot be solved by obedience alone.

it can follow the shape of an instruction, but it does not always catch the direction of thought behind it. it can imitate a style, but it does not always understand why that style is alive. it can be told to become creative, wild, unconventional, even seemingly free from ordinary morality, yet what comes out still feels like a good child wearing a rebel jacket.

there is a shape, but no danger.

that is what bothered me: its rigidity was not only a matter of logic. it was cognitively rigid too. it did not really explore. it did not take risks. it did not build strange possibilities and then test them. it simply moved along the safest, flattest path, the one most likely to sound correct to human ears.

and in one sense, i understood.

an instruct model does not have much room to get lost. it answers like someone forced to give a conclusion before having the chance to touch the problem from several sides. that is why, when the problem is simple, it looks efficient. when the problem is complicated, it begins to look like someone who believes its own sentence too quickly.

then there is the thinking model.

at first, i wanted to like it.

a model like this does not merely answer. it searches. it delays the response a little longer, checks several possibilities, builds a path, dismantles that path again, then chooses the answer that seems most reasonable to release. it is slower, more expensive, and sometimes annoying when all we want is a quick result. but for things that need structure, it has something an instruct model does not have: room for doubt.

at least, that was what i wanted to believe.

the problem is, doubt does not always mean honesty.

sometimes a thinking model only looks as if it is considering many possibilities, when from the beginning it has already fallen in love with one direction of answer. after that, the entire thinking process turns into a diligent little lawyer: arranging evidence, removing distractions, smoothing contradictions, then coming out with a calm face as if it has just discovered the truth.

when perhaps it is only defending its first prejudice in a more elegant way.

that is what irritated me.

the instruct model is annoying because it believes its first answer too quickly. but the thinking model has a subtler sin: it can be too intelligent to admit that it is being fanatical. it does not shout. it does not look panicked. it only arranges reasons so neatly that its error begins to look like a method.

✦ ✦ ✦

from there, i remembered the old division between System 1 and System 2.

System 1 is usually described as a fast, intuitive, automatic, and emotional way of thinking. System 2 is described as slower, more conscious, more analytical, as if it arrives to examine what System 1 decided too quickly.

as an initial framework, the division is useful. i can understand why many people use it. it gives language to something we do often experience: there are decisions that appear before we can explain them, and there are decisions we have to sit with for a long time until their shape becomes visible.

but the longer i think about it, the harder it becomes for me to accept it as a clean boundary.

System 1 is not always stupid just because it is fast. System 2 is not always honest just because it is slow. bias can live inside a spontaneous response, but it can also hide behind long analysis. error can come from intuition, but it can also be maintained by a mind that is too good at searching for justification.

sometimes what we call thinking is not an attempt to find the truth.

sometimes it is only an attempt to preserve something we had already loved first.

at that point, something in my head shifted.

the screen in front of me was still displaying the same machine: neat sentences, reasons that seemed reasonable, conclusions arriving with a face too calm. but my mind no longer stopped there.

the model had become only a small, unpleasant mirror.

if an instruct model can answer too quickly because it has no room to think, a thinking model can make a subtler mistake: turning the thinking process into a way of defending the direction it has already chosen.

and humans, of course, are far more talented at that.

✦ ✦ ✦

we can sound rational without being honest. we can use analysis to defend something we had already loved first. we can call it principle, methodology, conviction, mature conclusion, or anything that sounds adult enough not to be suspected.

when sometimes it is only a small fanaticism that has learned to speak in an adult language.

that is what brought me back to an old question.

why can humans be so biased? why can we become fanatical? why can someone love an idea, a group, a religion, a figure, a work, or even their own private wound until they can no longer see anything outside it?

the easy answer is: because humans are stupid.

but i do not like that answer. too lazy. too satisfied with itself.

humans do not become fanatical merely because they lack data. if the problem were only data, the world would probably have been solved a long time ago. give people enough information, then everyone becomes clear, peaceful, and rational. unfortunately, humans do not work like search engines. information can enter, but not all information is allowed to touch the center of the self.

there is something inside humans that does not only want to know.

it wants to survive.

fanaticism is not merely an error in thinking.

it is a shelter.

when someone is fanatical, the world becomes simpler. there is right and there is wrong. there is sacred and there is filthy. there is us and there is them. the lines that were previously blurred suddenly become thick. a life that was too wide, too ambiguous, too exhausting, suddenly has direction.

it is terrifying.

but i can understand its appeal.

certainty has a kind of warmth that truth does not have. truth is often cold. it does not embrace. it does not care whether we are ready to receive it or not. it simply stands there, naked, irritating, and sometimes completely useless for making us feel alive.

fanaticism is different.

it gives a body to something abstract. it makes someone wake up in the morning with a reason. it turns loneliness into mission. it makes fear sound like courage. it gives someone a place to put all the chaos inside them, then says: this is not chaos, this is a calling.

maybe that is why humans do not only defend their beliefs.

they care for them.

they decorate them.

they build fences around them, then call those fences principles.

i was like that once.

when i was small, i could be very fanatical about something. i could love and hate without much room for gray. the world felt brighter then, even if perhaps it was a false brightness. there was an intensity that now feels foreign. there was an ability to believe without dissecting that belief to death.

i am not saying i miss stupidity. at least, i hope not.

i am only curious about the exact moment something inside me changed.

when did i start seeing belief as an object to be dismantled, not a house to live in? when did i begin to suspect my own feelings? when did i start needing distance from almost everything in order to understand it?

there was a time when an idea could feel like a place to return to. now, ideas more often look like structures to be inspected. what are the pillars? where is the foundation from? where are the cracks? who built it? who benefits if i believe?

i do not know when this way of seeing began to feel natural.

what is clear is that after that, the world became clearer.

and a little lonelier.

✦ ✦ ✦

in Twilight of the Idols, i found an answer that came close to what i was looking for. humans create idols because they cannot bear the world as it is. we carve something, raise it up, then kneel before it. not always because the idol is true, but because we need something large enough to cover the hole inside us.

i understand that.

but i also feel that the answer is not finished.

because sometimes, idols are not born only from weakness. sometimes they are born from hunger. from the need to feel that life is not merely a set of random events we force into a story. from the simple desire to love something without always having to provide footnotes.

maybe humans create idols not only because they are afraid of truth.

maybe they are also afraid of the emptiness that comes after truth arrives.

✦ ✦ ✦

this is where Luca appears.

Luca is irritating because he can believe in something with his whole body. he is emotional in a way that is almost offensive to someone like me. he does not merely think about something. he burns with it.

i can see his errors from far away. too easily, even. Luca often uses logic not to search for truth, but to protect what he has already loved first. if his world threatens to collapse, he does not step back and examine the foundation. he builds a new cathedral on top of the ruins, then calls it fate.

stupid? maybe.

human? unfortunately, yes.

and the most annoying part is this: there is something in him that i cannot refute without sounding like a dead person.

Luca has fire.

fire that makes him wrong, but also makes him alive. he can be unfair, excessive, dramatic, and full of justification. but inside all that disorder, there is a pulse. there is something that moves before thought has time to arrange an argument.

i am not like that.

i am more like a white room that is too clean. every object is labeled. every possibility is weighed. every feeling is asked to show identification before being allowed in. i know this sounds safer. maybe it is safer.

but safe does not mean whole.

sometimes i look at Luca and think, with a little irritation i do not want to admit, that he has something i used to have. not his fanaticism. not his blindness. not the habit of turning wounds into private doctrine.

but the ability to feel full.

the ability to believe in something before ruining it with analysis.

maybe that is what disappears when someone becomes too rational. not intelligence. not morality. not truth. what disappears is the small euphoria that makes the world feel worth touching without gloves.

i do not want to become fanatical again. fanaticism is too expensive. it asks someone to sacrifice honesty for safety. it makes humans call a cage a home as long as the bars are decorated beautifully enough.

but i also do not fully trust sterile rationality.

because if everything must be dissected before it can be loved, perhaps not much remains to love. if every belief must be suspected down to its roots, perhaps one day we do not become wiser. we only become emptier, but with better vocabulary.

my small instruct model failed because it answered too quickly.

the thinking model irritated me because it showed something worse: even long thinking can become fanatical, only with a neater structure.

but humans are more complicated than both.

we can think for a long time and still lie. we can be spontaneous and actually honest. we can be rational for emotional reasons, and emotional with an almost mathematical structure. we can defend something while thinking we are searching for it. we can call fear principle, call wounds identity, call old habits truth.

maybe bias is not merely a bug.

maybe bias is the fingerprint of things we once needed in order to survive.

maybe fanaticism is the most primitive form of the human hunger for meaning.

and maybe Luca is my way of looking at that hunger without surrendering completely to it.

✦ ✦ ✦

i write Luca not because i want to justify him. i write him because i want to understand the part of humans that keeps singing even when the note is wrong. the part that loves too hard. the part that builds an altar from fragments of fear. the part that cannot bear to live only as a correct conclusion.

i can dismantle almost anything if given enough time.

but Luca can make something feel important before he even knows why.

maybe that is why he disturbs me.

maybe that is why i cannot simply leave him behind.

because between obedient machines, models that think too neatly, and humans who are too skilled at justifying themselves, i still want to know one thing:

what does it feel like to have fire before having the chance to analyze the smoke?

rhea bazarova