First Principles, Second-Hand Guides

I’ve been thinking about why i’d be (so as not to seem accusing anybody) awkward or straight up stressed out when throwing myself into things i haven’t done before.

my chronic anxiety contributes to part of that, but what contributed to my chronic anxiety in the first place?

There are lots and lots of bot-generated posts on every social media, either created by bad-faith content farms or by teenagers who have no idea what they’re doing trying to be helpful to others (or wanting to be a content creator, or worse, an influencer). not that what they’ve created has nothing of value. but a wrong mental model that has been deduced by inexperienced people with a small amount of (maybe even unconfirmed) experience – or in other words, input/output data – and no knowledge of the underlying mechanisms and principles, can be very harmful to the people trying to learn and understand things. but this kind of “guide” or “advice” is everywhere, because acting like you know what you’re talking about gets traction and attention, and is what people want, whether you are conscious about it or not.

let’s think about a standardized procedure. the most popular one for my peers might be the job interview process, especially campus recruiting. it’s been so standardized that even if companies have different flavours or tweaks to the process, the general structure is still the same. and here comes a billion interview hacks, tutorials, guides, shared experiences. “i landed a job at [insert big company name here] and here’s what i’ve done.” “i’ve been rejected by [insert conglomerate name here] and here’s what i think went wrong.” once you see like 10 of those, you’d realize that most of these posts are just one or several ideas being packaged over and over, the same 10 metrics in resumés being prioritized and weighted slightly differently to calculate a final score of how good this candidate is and how high the standard is for this company (and of course the higher the standard is automatically seen as the company being better).

the bigger a company is, the more inconsistent and varied the jobs and the requirements or hiring process is, because a jungle is not consistent, and big corpos will be even crazier internally than a jungle. one job position might be easier or harder to get than the other, might be better or worse than the other in many factors, and most importantly might suit you better or worse than the other (given that you actually know who you are and what you want, or you are just the adventurer type with no strong attachment to any one thing and just wants to try the best you can get, you have my respect), and might require some different set of skills, or a special trait that the manager deems important. but guides like these emphasize the idea of companies being so standardized (and thus trustworthy) and so consistent that you can follow the same guide if you have the same or better metric values, and you will certainly be able to land the job. and if you don’t, then you must have done something wrong, or you are not good enough, or you are not trying hard enough. real industry veterans know that this is never quite the case, that the hiring process is so inconsistent and so varied that you can never be sure what to expect or what to do to prepare for it, and you need luck on your side on almost anything for it to actually work out. but the guides are still everywhere, because they give people a sense of control and certainty, even when it’s based on a false premise. and this is just one example of many “standardized” (almost, or at least appearing to be, since we all know good jobs will never be released to the general public for grand-scale competition) procedures being oversimplified and overgeneralized by inexperienced people with no knowledge of the underlying mechanisms and principles. real life is so complicated and dramatic that not just ten, but tens of thousands of different factors (and even that could be an underestimate) can affect the outcome of a process. but that gives no illusion of information, no illusion of control, thus no attention to the content creators, and thus they have no incentive to create content based on such viewpoint, only on what people want to hear, which is a simplified and idealized version of reality.

that’s just one example of how a wrongly reduced and deduced mental model could distort reality, especially if people get brainwashed by binge-scrolling the enormous amount of such content on any platform and eventually accept and recreate the same distorted cognition themselves again and again, each more distorted than the last. and in the end the consensus will demand that reality be rewritten to match the distortion. interviewers who have done nothing like this (mostly team leaders who are knee-deep in the mess and are asked to interview five candidates a week), they will certainly ask the ai, and ai learns from this content, and the feedback cycle goes on and on until the negative effects become so visible, infamous and unbearable that people start to “break free of the chains” and land on first principles instead. the cycle goes on, circle of life.

but even setting the bad content aside, there’s a deeper structural issue underneath all of this. a huge amount of the most important knowledge in the world can’t actually be put into words at all – Polanyi called it tacit knowledge. things like “how do you tell if a technical direction is worth betting on”, “how do you tell if someone is bluffing in a meeting”, “how do you tell when a teammate is about to flake on you” – anyone who is genuinely good at any of these will tell you they don’t really know how to teach it, it’s some calibrated intuition built up over years of making judgments and watching what happens, and they can’t even articulate the structure of that intuition themselves. you can’t really post that stuff. if you tried, it’d come out as “pay attention, do more, reflect on what happened”, which is true but no one is going to share that and no one is going to read it. so what makes it into the content stream is structurally selected for being packageable – the tip of the iceberg of all that’s actually worth knowing. and that’s how i end up mistaking the tip for the whole iceberg, without even noticing i’m doing it.

so what’s the actual alternative to all this mediated stuff? as far as i can tell, the only method that builds the kind of knowledge that actually holds up is doing the thing and failing at it, again and again, each time a little more educated and calibrated than the last. the failures are the data. there’s no shortcut, no compressed version, no pre-digested guide that gets me there – the model gets built one failure at a time, and a lot of what ends up in it isn’t even verbalizable to myself afterwards, i just notice that next time i sort of know what to do.

back to my opening, why i’d become awkward or stressed out when doing something for the first time, especially if other people are involved. when all i do instead of actually reading and thinking is to read second-hand guides or mental models to understand the world, i am learning practically something else, a mirage, a leaky abstraction, a distorted version. since i have never experienced the real thing, and i’m so afraid of making mistakes or failing stuff, i’d watch guides and tutorials in order to “be prepared”, “speedrun it”, “hack the process”, “be ahead of the curve”, “be more efficient”, “be more productive”, “be more whatever”. everything i learn about is mediated, what i know about “how to do stuff” is a bad pattern distilled from someone else’s incomplete recall of the experience, a leaky abstraction. and when i get exposed to the real thing, the first thing that deviates from the guides, since i hate failure and want nothing to do with it, i will eventually panic, and appear awkward, and of course, fail at the thing.

and what makes this even worse and self-reinforcing is: when things actually go wrong, the attribution almost never lands on the model itself. it lands on execution. “i didn’t try hard enough”, “i should’ve prepared more”, “i should’ve followed the guide more carefully”. this works because real-world feedback is slow and noisy – unlike a test score that comes back the next day with a clean number, “execution wasn’t good enough” can basically never be falsified, there’s always a plausible story where that’s true. so the model never gets updated, only executed more intensely. and underneath that there’s an identity protection layer too: if the model is wrong, then every piece of advice you ever gave based on it was wrong too, and the whole “i know how this works” identity starts to crack. so blaming execution isn’t just cognitively convenient, it’s protective.

guides and tutorials work, don’t get me wrong, but only in the highly standardized and consistent processes. and even then, things will only work if you have fundamental understanding of the underlying principles and mechanisms, and know what to do when things go off-script. some things can be beneficial to be standardized, but others will not. Shane Parrish talks about the mental model of irreducibility in his book series, the core question being at what point things become not themselves anymore but an abstract substitute if you keep reducing them, or in this case, standardizing them. reducing here means to ignore or isolate the effect of minor factors, stripping out the major and important factors and figuring out an explicit and concrete process to keep it consistent and repeatable. some procedures might be fine or beneficial undergoing such reducing, but hiring, i’m afraid, will never be, because too many people and too many interests are involved. but the guides will still be everywhere, because the mind needs information and it needs a sense of control and it needs it now.

and i think there’s something scarier than just being bad at telling apart what should and shouldn’t be reduced. having been raised in standardized environments my whole life, i don’t just function inside them, after a while i start producing standardization of my own. faced with an open question, my first reflex isn’t to sit in it and think, it’s to convert it into a closed one as fast as possible. ranking universities, sorting companies into tiers, scoring myself against my peers, scoring my peers against each other on whatever metric is fashionable that week – multi-dimensional questions about life and meaning getting compressed into sortable scalars. the operation is automatic, i don’t even know i’m doing it. i think i’m “thinking clearly”. i don’t notice i’ve replaced the question with a substitute. and some questions only exist in their openness – “what kind of life do i want to live”, “what’s worth caring about”, “is this person trustworthy”, “is this thing i’m doing actually meaningful” – the moment i collapse them into a metric, i’m not solving the question anymore, i’m solving something else that looks like it. and it’s irreversible: what made the question real is exactly what got thrown out.

mom was right, it was really that damn phone.