i’ve always been trying to learn more from first principles about topics i care about. not because it sounds intellectually impressive, but because i’ve noticed that most shortcuts stop being useful the moment someone asks me a slightly deeper question.
analogies are probably the worst offender.
they’re supposed to make unfamiliar concepts easier to understand by connecting them to familar ones. most of the time, this shit actually works. but, they also quietly replace the thing being explained (AI explanations do this a lot). instead of understanding the actual mechanism, i end up understanding a watered-down version that behaves similarly under some narrow set of conditions.
then, eventually, the analogy breaks.
i’m not saying that its a completely bad way to learn, and that we should ban these connections. i’m saying that they tend to smuggle in assumptions that aren’t actually true. when reality starts behaving differently from the metaphor, you’re gonna be left trying to debug a mental model that was never accurate in the first place.
i’ve probably lost an order of magnitude more time unlearning convenient explanations than i would have spent learning the underlying mechanics properly, especially when i started learning about topics i actually care about (e.g., low-latency systems, networking, kernel internals, etc…). as a beginner, i usually don’t know which parts of the analogy are useful and which parts are fiction. i can’t see the boundary because seeing the boundary already requires understanding the concept.
i feel like this makes analogies slightly backwards as a teaching tool. they become more useful after you understand the subject, which is when you need them least lmao.
first-principles reasoning has been a better fit for how i want to learn. take something apart to its atomic axioms, identify what is actually known, separate constraints from conventions, and understand the smallest pieces well enough that the larger system stops feeling arbitrary.
elon musk talks about this approach a lot, and so do the folks that i worked with at tesla autopilot for 8 months. which is partly what pushed me to take it more seriously than i used to. the useful part isn’t imitating their conclusions, it’s refusing to inherit a conclusion just because it arrived in a familiar package.
i think this matters even more now that ai-generated explanations are everywhere. ask for a beginner-friendly explanation and you’ll usually get a metaphor within the first few sentences. it feels fulfilling because comprehension arrives quickly. but often what arrived was recognition, not understanding.
i see analogies as semi-decent compression algorithms. once i understand something, an analogy can help me communicate it, remember it, or view it from another angle. i just don’t want one to be the foundation.
if i care about a subject, i’d rather begin with the inconvenient version: definitions, mechanisms, constraints, and all the details that don’t map neatly onto something i already know. it takes longer for the first bit, but after that the ROI is usually unmatched.
~ a.k