Comments by "Vitaly L" (@vitalyl1327) on "ThePrimeTime" channel.

  1. There is no such thing as a "general purpose" language.
    2
  2.  @RomeTWguy  a lot of people build their internal systems this way, myself included. There is no point in making a product out of it, since cloning such a product is trivial. Devin is a scam, it does not add any value because it is so easy to make a better system nearly for free.
    2
  3.  @decator_lol  you're not a very competent developer, evidently. If you fail to comprehend that eDSLs is a zero-cost abstraction and a compile-time optimisation that is impossible to achieve any other way.
    2
  4. For me the biggest argument against java is existence of any versions after Java 1.6. Java 1.6 was perfect. Everything that followed just ruined it, by thousand tiny cuts.
    2
  5. And where are all those magical millions of smart and motivated graduates? Cannot find any. Shortage is massive and there is no way to fix in in sight.
    2
  6. Setting up an environment for debugging can be a hellish task indeed. First, not to forget to expose JTAG (even as unpopulated placeholder) on your PCB. Then, get a compatible JTAG programmer/debugger device for your target platform. Than configure your IDE, whatever it is (STM32Cube, whatever) to work with this particular JTAG programmer. And only after that realise that you cannot get execution traces over JTAG in real time without breaking real-time behaviour of the code you're debugging, so you just resort to debugging via a couple of available GPIO pins connected to a logic analyser. Or even blinking LEDs and capturing them with a high speed camera.
    2
  7.  @BeefIngot  I'm commenting on the moment Primeagen got confused by using GFLOPS to represent memory throughput. This is not an optimisation question, not even a programming-related problem at all, it's a simple arithmetic. I'll tell you more, you can reason about floating point compute throughput in the GB/s units too, though it'll make much less sense. My point is, anyone who was not excluded from a primary school must have an instinct for unit conversion.
    2
  8. Lol. Good luck "setting breakpoints" if a few milliseconds disruption breaks the logic of the code completely. Guess debuggers are only useful for toy code like web and such...
    2
  9. Oh, a ranting Java monkey - that's funny. That guy never really programmed, and suddenly bumped into a real language, which immediately showed him how much out of his depth is is, that all 30 years of his experience was 1 year repeated over and over... Sad, really. Funny, but sad.
    2
  10.  @maushax  given that Crowdstrike is just a scam, as well as the entire "anti-virus" industry, they would not care about any proper practices, all they do is security theatre.
    2
  11. The logic is good, but it's coded in Malbolge...
    2
  12. Write in OpenCL then. C++ is not suitable.
    2
  13. I hope in 3. you meant homotopy type theory
    2
  14.  @isodoubIet  might be "useless" for trash code monkeys, who simply have no idea how to apply it and end up writing crap code. But for 100% of the real engineers type theory is essential.
    2
  15. what are you talking about? Both Vim and Emacs can easily do it.
    2
  16.  @complexity5545  hire humans to solve problems AI canmot solve. Hire humans that are smarter than AI (about 10% of.humans are).
    2
  17. This is the wrong way to use LLMs to write code. When done properly, it generates unique and not seen.before code.
    2
  18.  @luvincste  I'd posit that those who find this puny little git "unintuitive" and hard shoud not be in this industry in the first place.
    2
  19. Again code monkeying being confused with computer science. Sad.
    2
  20. There is an entire industry built around celebration of incompetence and mediocrity. All those bootcamps, "get programming in 6 months" courses, all the grifter youtubers selling the idea that self-taught engineers are not any worse than the proper ones. Let them fail. There's no point resisting, too much money is in this industry. All we can do is make sure none of those bootcamp "graduates" get accidentally hired, and mock them at every opportunity.
    2
  21. Guy Steele Jr, the author of the Java language specification, who also happen to be one of the core Scheme designers, bragged that with Java they dragged kicking and screaming coders half way to Lisp.
    2
  22. There's is SPIR-V though.
    2
  23. ESP32 is huge and far too powerful. Try ATtiny85 instead.
    2
  24.  @matta5749  sadly (for you), there is an objective measure of complexity. And pointers are among the easiest things on this metric. Those who find them confusing are either genuinely dim, or made a mistake of skipping the basics.
    2
  25.  @anon1963  control theory, some basic physics (Lagrangian mechanics), hard real-time programming, and of course some basic geometry (i.e., know what jacobian is). Not much, really, it's just the same programming as pretty much everything else.
    2
  26. With a better thinking and planning you can end up with hundreds of lines of code instead of hundreds of thousands.
    2
  27. It's very easy to see if the answer was memorised or invented on spot.
    2
  28. Ah, a trash developer world view, with equally trash results.
    2
  29. If you cannot reinvent the wheel when needed, you'll end up in bog boring CRUD jobs and will never do anything meaningful.
    2
  30.  @jolting  indeed I do, but Tcl is not only EDA, Tcl/Tk is perfect for everything.
    2
  31.  @jolting  worse is better. Nobody is using it because the vast majority of developers are ignorant dimwits, not because the twisted crappy web stack is objectively better than Tcl/Tk for GUIs.
    2
  32. Also, allows to weed out bootcamp "graduates".
    2
  33. Yet, most of the small models are now trained on a corpus, significant part of which is generated by larger models. It's not a problem at all. Problem is in a tight feedback loop, you cannot feed the model with a data it generated without adding a noise or some other filtering.
    2
  34. You cannot do it in programming either. It's just the employers desperation for engineers, they hire absolutely unfit and incompetent bootcamp grads and self-taughts, and the outcome is always disastrous, but they keep doing it regardless.
    2
  35. Travis is a grifter selling an illusion that self-taughts and bootcamp graduates have a chance to become engineers. Spoiler alert: they don't.
    2
  36. Pretty much every distributed system requires its own tailored protocol. But dimwit coders use some XML, json or whatever other abomination they can dig out of their arses.
    2
  37. code must be dense. As in, dense as natural language, without any irrelevant semantics leaking into it. It's sort of possible with FP (but far from ideal), and impossible with OOP.
    2
  38. Because it workd. Linus middle finger did more for the quality of NVidia drivers than all the open source contributions combined.
    2
  39. And another good reason to get at least some OCaml is Coq. One might point to Haskell + Agda, but for Coq there is CompCert, and it's a pretty good reason on its own.
    2
  40. Do Rule 110 next, it's even more fun.
    2
  41.  @KyleSmithNH  yes, it's better than sticking to even higher level than C indeed. What I'm saying is that it's not enough to get a comprehensive and systematic understanding. It's an important step, but not sufficient.
    2
  42.  @DemiImp  I am talking about generic knowledge, transferable across platforms, which can only be gained by studying one platform (probably a toy one) thoroughly. Things like ABIs, registers and register spills, caches, cost of memory accesss, atomics, memory operations order, pipelines, effect of scheduling on IPC, alignment, SIMD/SPMD, and many more
    2
  43.  @filippomenchini  Tell Sapir and Whorf that language is not important :) Yes, you need to learn fundamental CS. And it includes learning computer architecture too. Which implies learning the low level languages, HDLs and many more.
    2
  44.  @ItsJustJoseHere  Why? If you do it ONLY in your day job. No side projects.
    2
  45.  @RotatingBuffalo  and I find it hilarious that you think that there are macros in C.
    2
  46.  @RotatingBuffalo  Now, I said "F*", not "F#", but you did not notice. "F*" is miles away from the run of the mill Hindley-Millner that F# does and Rust tries to do. And let me remind you that we're talking about Rust here. Fairly mainstream language, with an ML-inspired type system, procedural macros, region analysis and a lot of other features from the "ivory tower" languages that you believe for no good reason to be impractical. C and Java have very, very similar use cases. I worked on a high-frequency trading system that was largely written in Java. Eating C cake, evidently. I also worked on pure C HFT systems, ones written in C++, ones with large parts implemented in HDLs. There was no real difference in what C and Java did in those scenarios. Just plain, predictable imperative languages with more or less low level control of the memory layout. Nothing fancy. Everything very much the same. Not to mention java running on microcontrollers or even NFC chips (already mentioned JavaCard). For all practical purposes, Java is not too far from C. Yes, GC can and will cause troubles, yes, you need to code in a certain way for real-time, but same applies to C, your malloc() and free() are also not allowed.
    2
  47.  @atiedebee1020  code monkeys are hilarious. Yes, it means exactly that, that Java and C are very much alike, to a degree that an ignorant code monkey can safely switch between thrm without suffering from a severely hurt ego. But when a code monkey is confronted with a significantly different language, such as Rust, they have meltdowns like the one we witnessed in the OP. Ia not it a good measure of similarity and difference to you? Have you ever implemented compilers for any of the languages from this fsmily? If you did, you'd know how similar they are, even the dynamic abominations like Python.
    2
  48.  @HrHaakon  it is not HM in a slightest, it is a simple linear type propagation. It does not unify type equations. And, no, external preprocrssors are no macros.
    2
  49.  @lucass8119  objects are not big deal at all (see GLib / gtk+) - just some puny syntax sugar for implicit v-tables. Reference types also not a big deal, as it's a subset of pointers (sans arithmetic). Generics - yes, the only mildly notable difference between C and Java, but still not warranting putting Java into a different class of languages. And yes, I'm piling the dynamically typed eager imperative languages into the same group too.
    2
  50.  @user-oe4id9eu4v  since youtube removes comments with links, I recommend to look up "real-time java" to see some allocation-free memory management techniques employed there. As for preprocessor - who cares if it's a part of a language spec, if it's a separate pass that have no knowledge of the underlying language whatsoever? And it can be used with other languages easily (e.g., I worked on a project where a C preprocessor was used on a Verilog code base extensively).
    2