Comments by "Vitaly L" (@vitalyl1327) on "Theo - t3․gg" channel.

The very culture of the web code monkeys is disgusting, they fail to comprehend that every third party dependency is a liability and one must think really well before pulling in a single third party library. Yet, they have this npm mess, every project brings in hundreds of dependencies, and they never learned the lesson of the leftpad fiasco.
10
I suspect people keep doing it wrong. End-do-end coding models are a dead end. Training those models on too much data result in overfitting, you're getting the most average, the "always do it the most popular way" solution. Slow down. We have perfectly capable coding models that do not even know the programming languages they're asked to code in. And they work way better than Claude 3.5 did (did not try 3.7 yet).
4
@chess_ramone culture fit is not a predictor of engineering abilities. And these days there is an insane amount of impoators, of pseudo-engineers wbo Dunning-Krugered themselves into believing tbey are capable while tbey are very far from it.
4
 @azizsafudin you can implement the Rule 110 machine in html+css. Leaving it to you as an excercise to prove Rule 110 Turing-completeness.
3
but how? The problems engineers working on in real life are, normally, quite big, complex and entangled. You cannot give any such problem on an interview and see an engineer devouring it bit by bit in a course of one hour. He'll need 3 to 5 months to simply start getting into this problem at work. So how can we assess an engineer, if not by a smaller scale problems that still, hopefully, have some of the properties of the larger ones?
3
@William Henseley ok, you're looking for an engineer capable of writing long term maintainable code, capable of quickly comprehending, often reverse-engineering legacy code, capable of creating optimal or close to optimal algorithms when requirements are tight or producing reasonable solutions quickly when requirements are lax. Your code base is 40 years old and contains millions of lines of code in multiple languages. How do you test engineers for what they're expected to do at work, see the list above? Small, simple coding problems is pretty much the only way to filter out thosr who cannot even code and cannot solve problems, and only then you can think of a culture fit, of a past track record and all that. I noticed a curious pattern. Those who are vehemently opposed to leetcode style interviews are always ones who are consistently vetted out by such interviews. Guess, for a good reason...
2
To start with, CS education does not even need to involve computers at all, let alone trivial minutiae such as git.
1
First of all, why would a CS graduate want a coding job? How about doing an actual *science*, as this is what this degree is for? It's like studying computational hydrodynamics in order to land a plumbing job. Does not matter that 95% or so get it wrong. Science is science, engineering is engineering, you should not confuse one for another.
1
@rhys9336 Nah. No npm-using web monkey will ever deliver stuff I do, and I won't even go into their mess either. But, yes, it is hilarious for the web monkeys to believe they are "productive" when they glue tons of incompatible crappy libraries together, wasting way more time than it was needed to build everything from scratch.
1
@elorrambasdo5233 if people are getting a systematic, high quality STEM education and still don't know how to find out what they're supposed to know are in the wrong place and should have been flipping burgers instead.
1
@elorrambasdo5233 all the proper STEM subjects are taught systematically, from the bottom up, without leaving any gaps in the fundamentals. Anyone who did not receive such an education is incapable of learning any new subject properly, they cannot construct systems out of bits of knowledge.
1
@ohgin12345 leetcode is just a first step to weed out the most outrageously incompetent. Of course it is not enough.
1