Hacker News Reader

Comment by ldjkfkdsjnv | Hacker News Reader

I think that the existence of leetcode interviews, where tech companies ignore your background as much as possible, and want to see the skills upfront, is a condemnation of education.

It's essentially saying, we don't believe your degree says anything about your ability to perform the job.

The only reason academia hasnt been completely uprooted, is:

1. Some sects of society use academic prestige as a product. i.e. this harvard educated mckinsey consultant will optimize your business! Or, this harvard educated financier will invest your money carefully!

2. There's no way to have some leetcode like test for most jobs.

3. There used to be more signal in academic credentials, and older sects of society still believe those signals exist. Its the same with doctors, older generations treat doctors like infallible royalty, younger are much more skeptical

theres a huge caveat that some top people in the world do end up at top academic institutions. its just they keep mixing in these political appointees, agendas, etc

View on HN

Computer Science departments have always said that they do not prepare students for work in IT [0].

As with most development team leaders, I have interviewed and employed CS graduates who are woefully unprepared for the reality of a job writing production-level code.

I don't see this as a condemnation of education per se. More that the common concept of education (that it prepares young people for their careers) is completely misaligned with academia's concept of education (that it trains young people in how to think, and how to conduct research).

[0] attribution for this is interesting and varied: https://quoteinvestigator.com/2021/04/02/computer-science/

This might vary by school, but my undergraduate university had, in my opinion, a great computer science program which attracted many big-name companies to job fairs. There was definitely a focus on teaching practical and relevant things using relevant technologies. For those interested in research, there were also undergraduate research opportunities.

IMO this is what a university education should look like for most people in order for it to stay relevant. There are otherwise far more too many people attending universities relative to the demand for researchers.

This is the best model. I have been advocating for this forever.

The university that I got my actual degree from was something that a SV company would scoff at. But, the education they were able to scrap together was far more effective in my day-to-day work than prior education I had received at a far more prestigious institution.

Perhaps it was just me, but it always seemed like the less money I spent on my education the more effective it became.

I really wish I had understood this when I was 18. Would've saved me the four years and tens of thousands of dollars I wasted on a CS degree.

I dont really have any qualifications, never been to uni, in fact I clashed with the educational establishment, but I've worked on some very high profile systems, pioneered novel uses of technology and continue to do so.

Remember it doesnt take any qualifications to set up a business.

The academic culture of "we teach you to think not how to work" only makes sense for a world where the majority of people entering university are from wealthy, elite families such that the economic impact of attending college is negligible, and just an accepted tax to confirm social status. This was true for almost the entire history of college until after WW2.

At those points in history the degree was worth it because it was a signal of your social standing. You would walk into the office as the boss simply because you had a degree.

Today college is not elite. The vast majority of US high school students now go to college. 70% of those students take out loans to afford education. Because it is no longer elite "teaching you how to think" is no longer a great value proposition. Not to mention that most of the new grads I've met have failed to learn "how to think" as well as having no practical skills.

Academic education is not about what the university chooses to teach but about what the student chooses to learn. If you finish the degree by completing only the minimal requirements, that's your failure as a student. Many people need more guidance in their studies, because their motivation to learn for the sake of learning is not strong enough. For such people, many countries offer a parallel track of vocational higher education.

Also research shows that "colleges you've been accepted to but didn't go to" correlates the same to future income as "colleges you went to". I'd say 80% of the signal value of college is the initial sorting vs the education.

Aren't there even earlier sorting events that have predictive value of what colleges you might be accepted to? I thought I read somewhere that "father's income" is the single best predictive signal of the entire chain of events that leads to one's prosperity and career success. If true, even college admission doesn't signal much in and of itself.

>The vast majority of US high school students now go to college.

I think the number is something around 60%. Not sure that would qualify as Vast Majority.

My school had plenty of "vocational" classes available to CS (and other) students. The school even made an effort to keep the pre-reqs simple for these so that students from other disciplines could give them a shot (and maybe so that students could prepare for internships). Classes like, iOS development, back-end web development, intro to data science, etc.

I'd be surprised if most universities didn't offer something similar. Even students at Ivy League school must want to get together with other students and learn to build useful software during their first year.

Learning how to code is not the same thing as learning how to write commercial software at production level. There are lots of ancillary skills that must be mastered (git, testing frameworks, cloud deployment), lots of concepts that must be learned (dependency injection, the rule of 3, CPS, etc), and soft skills that must be mastered (how to accept criticism of your code, how to review other people's code, how to have a technical argument without getting emotional, how to tell your boss they're wrong).

Like most professions, there's a lot of "extra" stuff around the core competency. Knowing Excel does not make you an accountant.

You'd be surprised how much of that you actually need to actually get a software project off the ground and into customer's hands.

I've come into projects as the architect for refactoring what was built by someone who had only just learned to code. It was bad, sure, but they were bringing in enough revenue to hire me and get things on course. All with 0 knowledge of devops and 0 lines of test code coverage.

The question becomes at what point do we expect companies to train people and what point do we stop foisting that on universities? This ties into the bigger goal of what the purpose of university is and should be, but I personally think companies need to step up and start training again instead of expecting universities to cover most things they want/need.

There should be a year's placement in industry as part of the degree.

Three summer internships is roughly a year of industry experience, especially if they are done at 2 (but not 3) different companies.

Actually, those three 3 month engagements can constitute several years worth of experience if they are done with excellent mentors at good companies. Reasons:

1. You see several teams, several projects, several codebases, without having to spend an awesome amount of time at one place. How many engineers work with multiple teams and companies in their first year (without one of those experiences being terribly catastrophic)?

2. You get the interesting meaty projects, or get to work on prototyping projects, but are also expected to contribute to the existing code base (or at least build on top of it). You participate in meetings, give presentations, etc. Basically, you see an entire compressed project/product lifecycle but embedded in an existing codebase.

3. Most importantly: Doing 1 repeat internship is super valuable. Spending 6 months at a place in 3 month increments with 9 months in between is an exceptionally unique experience in terms of understanding how orgs work and evolve. Most engineers don't get to experience until they are in the consulting portion of their careers.

University students in CS should do as many internships as possible, and universities should be more proactive about finding a way to mock summer internship experiences for students who do not get any internship offers.

totally agree!

Atlassian made some comments like "Australian universities aren't teaching devs what we need them to know" a few years back. It triggered some discussion, but not nearly enough along the lines of "well maybe you should train people in what you need them to know then?"

As an industry, we're utterly crap at employing juniors, or training our staff. We expect everyone to have X years of experience and then get surprised when devs make technology choices based on what's good for their resume, not what's good for the business.

there's a full rant I have for whenever this subject comes up ;)

As someone who moved countries to do a technical masters (Theoretical Physics and Applied Maths) who has to pivot to tech (sadly doesn't seem Ireland recognizes my US teaching credentials or 5 years experience) for a while, I'd be interested to hear the rant as well as how I could better prepare for interviews, etc.

Xavier Niel, a (self-made) French billionaire, founder/owner/president of a few tech companies, said the same thing. And as a result he founded a series of schools (called 42) to "fix" that and provide a more realistic and practical education, focusing on tools and problem solving and critical thinking and industry experience through long internships. Completely free of charge for students.

As an alumni (kinda), it was awesome and IMHO has served me very well in my career.

College was never meant to be and I don't think ever should be Vocational school. Ironically things like Bootcamps have cropped up that in theory teach the vocational part but are pretty disrespected in the industry because the grads didn't learn DS/Algo.

The trouble with a company providing training is people will take the training, and then leave.

Because freeloading companies can spend the money that would have gone to training on increased salaries to the freshly-trained employees they poach.

To me that sounds like a company problem, and not something that should be foisted on to the universities to deal with. Perhaps it'd be better if we had more trade school focused places for a wider range of jobs instead of forcing the universities to deal with it all.

But, really, if company wants to retain people after training, maybe make it worth it to the people to stay.

The only reason this happens is because the company doesn't then update the employee's salary to be competitive with their new experience.

Lots of organisations seem to think that if they have to train an employee, then there should be some kind of discount on the salary for that employee. This is totally the wrong way of looking at it.

Training your staff in what you need them to know means that they have a consistent view of the subject, practices, etc. No more "but that's how I've been doing it for 2 years in the last place" or "I taught myself and discovered this way of doing it and I think it's cool". A junior employee who was trained in the exact tech stack you're using is more valuable than a mid (or even senior) who has random experience.

Not to mention, if your employees trust that you'll train them in what they need to know, they're massively less likely to pick technologies based on resume need. And massively more likely to stay with you.

No educational institution will build your corporate culture. And how code review or technical discussion looks like has way more to do with your culture then with whatever went on in school or previous employer.

Also, it should not take long to train someone git or test framework.

My degree dedicated an absurd amount of time to software engineering. And I graduated a while ago. Two of those courses required completed projects in groups, with documentation, a project management plan, VCS, deployment, testing, and a presentation at the end. Plus, we had a capstone engineering project which was essentially all of this, but solo.

My personal option was they were wasted courses. The learning was in the coding, not project management and documentation.

That may be, but when I think back to some of the nonsense classes I took as "breath requirements" to make sure I was well-rounded enough to be a CS grad there was definitely some room in there to teach me how to use git.

My biggest issue with new hires is that they can't write.

I can teach you how to use git in an afternoon. I can't teach you how to write a thorough but concise email about a complicated topic. I can't teach you how to give a good presentation. (Or, I can, but it will take a lot longer than teaching enough git to get started.)

(To be clear, all CS departments should teach students how to use git as a component of the intro programming sequence.)

Maybe an English course should've been included in your "breath requirements" ;)

Well that's embarrassing. Truthfully though, an Android keyboard with a functioning "swipe" algorithm might be more effective for this one. I swear this used to work way better 5 years ago.

> it trains young people in how to think, and how to conduct research

But they don't. If they wanted to students to think more deeply, there would be more of an emphasis on formal logic, developing proofs, etc. rather than "write this in assembly." If they wanted you to conduct research, you would...conduct research.

STEM degree fields are so obviously career-based. There's a reason Merck internships go to biology students, Lockheed internships go to aerospace students, Google to CS, etc. But when people call out universities on how poor of a job they do, they suddenly claim career-based education isn't the goal. It's a widespread CYA move by vastly overpriced universities.

>If they wanted to students to think more deeply, there would be more of an emphasis on formal logic, developing proofs, etc. rather than "write this in assembly." If they wanted you to conduct research, you would...conduct research.

The claim that a university education doesn't do this feels very subjective. My CS course did all of those things. Every student was required to do a research project in their final year for a significant part of that year's credit.

HN comments on CS degrees are always a hot mess because quality in the United States is extraordinarily variable.

At the high end, you have places like MIT/Stanford/UIUC//Harvey Mudd. As far as I can tell, none of the bad stuff that's said in these threads is even remotely relevant to the programs at those places^1.

At the low end, you have branch campuses of state universities and small colleges where half or more of the CS courses are taught by mathematicians who did not study CS, have never worked on even academic code projects, and have never worked as anything other than an academic^2. This is pervasive in Data Science. You're way better off with a boot camp than a data science sequence taught by a desperate pure math PhD taking whatever work they can find.

And there's everything in-between.

A lot of these debates are probably happening between folks who went to good programs and folks who went to mediocre/bad programs. Not all, but quite a lot.

So, when reading comments on HN about how useless or wonderful a CS program was, always ask (at least rhetorically) "where did you go?"

I gave the big famous names because that's the easiest way to make the point. There are tons of really high quality programs whose names you won't immediately know. My intention here is not elitism... BUT!...

It is 100% true that the quality of CS programs is incredibly varied. Way more than it was even 10 years ago. The rapid increase in enrollment and unwillingness to pay even half decent wages for CS faculty has caused an explosion of what I would characterize as fraud in CS higher ed.

So, in the spirit of helping parents identify good programs without depending on name recognition/elite signalling alone:

When deciding if/which college/university to send your child to, carefully read through CS faculty bios. Ideally there will be a good number of faculty with street cred in both academics and industry. In addition to checking out CS faculty buios, also check out the names of instructors in previous years' course schedules. For example, make sure that CS/Data Science courses aren't being staffed by math folks unless those people have significant industry experience in addition to their math phd, or work in an area of applied math where they're doing lots of coding, etc.

Checking who teaches the courses in archived course schedules is more important than department bios! The department bios are a first check, not the final answer. Here's why: at many places you'll have a few half decent CS faculty who are just "fronts" for a department that is mostly staffed by way lower quality labor. It's the EXACT same bait and switch as consulting shops who have super experienced principal/senior-level engineers participate in the sales pitches and once the ink dries you're working with cheap grunts in a body shop.

Three things to watch out for in particular:

1. Non-CS (usually math) phds without substantial industry expertise. They're teaching tons of intro DS/CS courses instead of taking a math job in industry or academia for a reason.

2. Folks with an online masters and maybe a year or two of full time industry experience.

3. Tons of CS courses being offered in the evenings or early morning by faculty whose names don't appear anywhere on department websites (these are industry workers doing ad junct labor... they can be EXCELLENT, and if you get a course taught by one of these people in the middle of the day it's often a gold mine. But if they're always teaching around the 9-5 schedule it signals that their employer considers this a side gig and the course's quality is probably going to depend on how busy things happen to be that quarter at the instructor's "real" job. Okay as an occasional stopgap or an option, but a huge warning sign if this is a standard modus operandi for core courses. Teaching well is a big job; doing that plus engineering without full employer buy-in is dangerous.

Again, there's nothing wrong with the first two categories of folks teaching CS courses per se. The problem is that they just aren't worth the tuition dollars you're paying or the time you're putting in.

Again, think about how much you will be spending. Repurposing math phds or junior engineers as CS faculty just isn't appropriate at those price points. At tends of thousands per year, your child deserves PhDs with industry experience or senior/principal engineers who have decided to start teaching full time. That's what is being sold to you! It's what every other deparmtent at the university gets! Don't pay shitloads of money for less.

Footnote 1: Those "good" programs have other huge issues -- cost, stress, elitism. Most importantly, some of the best CS departments even at otherwise accessible universities are extraordinarily selective due to capacity issues. But the curricula are extraordinarily rigorous in terms of both theory and SE skills. The only complaint I've ever had about these hires is writing ability.

Footnote 2: Warning: rant incoming! Honestly, regional accrediting bodies should probably start refusing to accredit CS/DS programs that aren't primarily staffed by faculty with either appropriate academic background or significant industry experience. And parents/students should consider civil suits when they enroll in an accredited program in CS that is taught by pure math phds with no industry experience and who have no business teaching CS or SE. The status quo is embarrassing for the entire industry. At the end of the day, universities need to start investing in good CS faculty or else face displacement by competitors who are willing to invest in human capital. Budgeting $80K for an instructional position pretty much explains why most CS departments are essentially fraudulent.

All programs, not just CS, have widely varying quality levels, probably leaning toward "bad", on average (there are a lot of colleges and universities in the US).

Fussell (and probably many others, but I've read him) blamed the rapid, great expansion of colleges and universities when the GI Bill injected a ton of money and eager new students into the system after WWII. He wrote that in 1983, but the system still seems to be reeling from that, and it may simply have stabilized (perhaps unavoidably, hard to say) at a much lower average quality level than before. Generally, institutions that existed before the war are still very good. Ones founded or converted (from, say, teaching colleges) after are much more hit-or-miss.

CS was "equally bad" to other not great fields in the aughts. It has gotten exponentially worse as enrollments have exploded.

CS is embarrassingly bad even relative to the already low standards.

I'm not exaggerating when I say that many computer/data science programs are 100% staffed by pure mathematicians who both have never done academic work in computer/data science and also have never worked a day in industry.

Find me a single mathematics department that offers a BS in Mathematics and in which 40+% of courses are taught by failed psychology PhDs. That's the state of CS at many institutions.

Again, I take your point that everything is not great. CS is a whole order of magnitude worse. Even at places that aren't great.

You are making a very important point in that program content and quality can vary greatly between universities. However, you then go on to specifically recommend the kind of program that has less to do with classic Computer Science and more to do with Software Engineering or development.

The person above you complained there weren't enough proofs in CS degrees, so they probably attended a program that was more focused on relevancy for the industry, while I and others attented programs very much geared towards academic research and theoretical CS, which contained a large amount of maths, often taught by math professors, and involved proofs in pretty much every course, even things like Intro to OOP.

Both of these programs have their place, I'm very glad I attented a program that was very theoretical, since I was able to pick up job skills on the side, while the things I learned in my degree I would have probably largely not been able to teach myself.

> However, you then go on to specifically recommend the kind of program that has less to do with classic Computer Science and more to do with Software Engineering or development.

No, I do not. Is that really how you would describe any of the specific institutions I mentioned?

A good program should have some of both, and two equally good programs could emphasize one or the other more. All of the positive institution examples I listed have very rigorous theory requirements in their undergraduate curriculum, for example.

> ...which contained a large amount of maths, often taught by math professors, and involved proofs in pretty much every course, even things like Intro to OOP.

The problem is one of quality, not the "theory vs practice" axis.

A mathematician can do a great job at teaching pretty much any CS course. But, also, within the last 10 years, a CS/DS major staffed largely by mathematicians or folks without terminal degrees has become a VERY predictable sign of a low-quality department. These two things aren't contradictory.

Have you looked at the composition of CS departments at unselective regional institutions, or CS faculty ads for such places? If not, consider this exercise. Things have gone downhill fast in the past 10 years, and the current situation in CS for regional colleges and universities is very much "caveat emptor".

I remember reading an experiment where they gave first-year undergrads a critical thinking test based on a text.

Then gave the same text/test to graduating undergrads.

The thesis being that they just spent 4 years being trained in how to critically think and analyze a text.

There was no difference in their performance

I know many well educated STEM professionals who rigorously apply critical thinking skills to their profession. They use the scientific method to discern truth from nonsense.

Outside of their profession, they uncritically accept silly ideas as the truth and throw all logic and reason out the window.

> there would be more of an emphasis on formal logic, developing proofs, etc

Which is exactly what a CS program from a reputed university is. We never once touched assembly. Heck most classes involved no coding at all.

The courses that teach SE tools (MIT's Missing semester) are well received. I can understand the resistance for focusing on specific frameworks or dumbing down the syllabus.

What I think is there deserves to be one or two full pledged software engineering courses, teaching how to use testing, version control, software structure etc.. As always, you can start from basic commands / usage and reach to general concepts, it isn't that hard.

And actually I believe US and European academia, at least the Ivy league and second tier colleges, do a decent job at teaching many required concepts. Come to a country like India, and you will see a C programming course that doesn't cover malloc or function pointers, a data structures course without proper introduction to algorithm complexity or difference between abstract data type and implementation, a java course that never properly teaches subtyping. I think students in US graduate decently equipped to write real world software compared to us. At least, I learned a lot of things by reading notes and Ppt of US universities.

(most Indian colleges are woefully inadequate in teaching even the basic concepts for writing software, despite the CS courses being named CSE (comp sci and engineering), not just CS. Incompetence runs to the root in education system except few elite colleges. From that perspective I think CS depts saying that is a first world problem. Sorry for the rant.)

The current university system (I can only speak to US here) was made post-WW2 to produce employees for office work in large corporate bureaucracies. The degree insured that you could show up on a regular basis, go where you're told, learn how to perform arbitrary tasks whether you thought they were necessary or not, and communicate effectively about this. They also provided, at the larger universities especially, a lot of training in how to deal with a bureaucracy.

All of this was absolutely relevant in 1950 for working at a large corporation, in a white-collar job.

The more recent job market doesn't look much like this, but then good training for the current job market would probably not call for an institution that looks much like a university. It would probably call for something more like trade schools and apprenticeships, with guilds to protect the apprentices from poor treatment (something that, for example, Ph.D. students are sorely lacking). It seems unlikely that universities can adapt to a situation that primarily requires that they go away or at least become much smaller, to be replaced by a different system.

You could argue the reverse. A lot of leetcode questions are/used to be tests you've studied CS, and largely unrelated to day-to-day software dev (e.g. the infamous "invert a binary tree" question).

Therefore, they mainly ensured companies hired people from similar university programs that the test givers also attended.

As much as I hate leetcode, I think it's imperfect but works. I think the biggest tech companies have rigorously tested it, and come to the conclusion that for mass hiring and removing bias/discrimination from the interview process, it's effective.

Knowing how to invert a binary tree is proxy for knowing how to do other none-trivial tasks in your domain of choice. It's so much better than the alternatives for interviews. Which are questions like, obscure trivia, spot-the-bug, how-many-piano-tuners-are-in-chicago, or two-day-long-mini-projects.

Everyone is optimizing for getting the most information in the smallest amount of time. Algorithm questions do that. The interviewee knows that they are going to get algorithm question, and most places tell them ahead of time what the nature of the questions will be. So interviews should be an easy cakewalk that everyone can pass, but they aren't.

I think the point was how can you say Leetcode is rejecting Academic experience when the majority of questions are straight from undergrad DS and Algo courses?

>So interviews should be an easy cakewalk that everyone can pass, but they aren't. For a person with no CS degree this is definitely not easy and being flippant and saying it is is disingenuous.

It's rejecting academic _credentials_ as a proxy for programming competence. Say what you will about it, Leetcode requires you to demonstrate actual capability, and it's used precisely because a CS degree in and of itself does not accurately signify the presence or absence of that capability.

But it doesn't, it reflects your training in solving DS and Algo problems from undergrad CS curriculum.

I think its almost universally agreed at this point that Algo interviews don't have anything to do with day to day work and don't demonstrate development ability.

> older sects of society still believe those signals exist.

I'm only in my forties but it's incredible how much this has changed in my lifetime.

I remember talking to an intern a few years ago and they made a comment about what a joke Columbia is (in regards to master's students). I was shocked because when I was a teen Columbia was seen as incredibly prestigious. I argued with them that Columbia is a really prestigious, well regarded school, but they and a few other interns just laughed and rolled their eyes.

Looking into it more it became obvious how this view had changed. It turns out Columbia, like many top tier schools from when I was a kid, has basically turned into get fancy version of a degree mill for master's students. Even though the school boasts some impressive faculty in many departments, it's also clear, based on the grads from there I have worked with after this conversation, that very little of that prestige is passed on.

As someone who used work in academia, it's wild how rapidly the entire system has decayed. From publisher or perish culture creating a mountain of non-reproducible work, to schools across the spectrum rushing to exchange credibility for cash. There have been so many systemic errors made that I don't believe academia will recover.

I suspect we'll see a fairly massive contraction in higher ed in the next decade or so.

Columbia also has the School of General Studies for non-traditional undergraduate students. Tuition is $58,440 an academic year, not including their estimated $26,010 cost of living, and they don’t offer much financial aid.

It's frustrating in a way. Perhaps I'm in the minority, but I'd say my well graded Master's Thesis on a complex subject from a top university is much more telling about the type of work I'll create than an hour long leetcode interview will.

It seems no matter how many great projects I make, or quality papers I write, that's second best to 10 minute logic puzzles for a growing amount of companies.

The problem is that many graduate degrees are very research focused and they don't say a lot about the skills many company's may need from software developers. Jobs I've worked that used Leetcode simply wouldn't need you to do the kind of work you did in your thesis. So they shouldn't be requiring your degree in the list of candidate requirements, but they also shouldn't place a lot of value on it if they're not doing that kind of R&D. I've met a few computer science Ph.D's who showed up on day 1 and appeared to have the coding skills of someone who used BASIC for an assignment in their "Technology" class in high school. One wonders if they're even the kind of people who got pushed through a program so their advisor could stop dealing with them. Maybe they are great at researchers, and abstract thought, and maybe they were very well versed in writing code a decade ago. Maybe they're just not good at it. But either way stuff like this isn't made up: https://blog.codinghorror.com/why-cant-programmers-program/. Leetcode comes about to deal with that. Not saying it's a great solution, but don't dismiss the problem.

Same here. I literally got a US PhD in math. Means NOTHING. But can I merge two sorted lists? Now we are talking serious money and benefits.

There is absolutely a way to have leecode type tests for most jobs. We know that the single best predictor of job success and performance is IQ. Academics act mostly as a signal in some industries where you just back into IQ ineffectively by creating certain filters for pedigree.

instead of IQ test -> competence signal

you have:

SAT/ACT = IQ Test -> university acceptance criteria -> degree/credentials -> competence signal

> We know that the single best predictor of job success and performance is IQ.

Citation for this? As far as I know IQ isn’t a measure of intelligence (because intelligence is hard to define, let alone come up with a single number that quantifies it!)

good summary of all the relevant data and studies for the last 50 years:

https://www.amazon.com/Know-Debunking-Myths-about-Intelligen...

> Citation for this? As far as I know IQ isn’t a measure of intelligence (because intelligence is hard to define, let alone come up with a single number that quantifies it!)

All you have to do is get a figure that mostly correlates fairly well to most of the phenomena we might consider intelligence, and it works well enough for many purposes, even if it's not measuring any of those others.

That's a far cry from "We know that the single best predictor of job success and performance is IQ."

I think when people say things like that, what they mean is "what simple number is the best predictor of success", which begs the question that a simple number can be a good predictor at all. Keep in mind that something can be the best merely because every other option is even worse: in practice, the best predictor is a really terrible predictor. So bad, in fact, that a five-minute conversation with somebody will tell you more about their ability to succeed at a job than knowing their IQ.

> that a five-minute conversation with somebody will tell you more about their ability to succeed at a job than knowing their IQ.

Got any citations to support this?

> So bad, in fact, that a five-minute conversation with somebody will tell you more about their ability to succeed at a job than knowing their IQ.

But when you try to put a number on that five minute conversation it turns out that the number is worse than the number provided by IQ tests. Otherwise the simple number from conversations would be well known to be a stronger predictor, but nobody has found such a result.

If we forget about the concept of intelligence for a while, there is a lot of research that points to the direction that whatever the quality that IQ tests measure is, it correlates strongly with financial success, health and a bunch of other things that are often considered good. Whether it is the cause or effect is also anyone's guess.

I've seen rumblings out there[1] that "whiteboard interviews" can be cast as "pre-employment tests" and litigated against in a "Duke Power" sense. Not to endorse that and I'm not sure anyone wants to hear that.

[1] https://twitter.com/richgel999/status/1506870732889985025

Yeah this is a good point. Another issue though is the SAT is a snapshot of your IQ, but also your general life state at one point in time. I think it's cruel to have that follow you for life, especially if you are going through difficulties as a teenager.

> We know that the single best predictor of job success and performance is IQ

Who knows this? I sure as hell don’t and would be extremely surprised if that were the case.

the standard reference for this is ree and earles 1992.

every other method of selection, e.g. structured interviews, work samples, etc, are (noisier) proxies of general mental ability.

"Intelligence is not the best predictor of job performance" (McClelland, D.C. 1993)

Not sure why you would use a paper as a "standard reference" when flaws in its method were identified almost 3 decades ago.

it's not consistently referred to as "IQ" or intellligence, typically the measurement is called "g" or "g factor" or "general mental ability" and over the last couple of decades the meta-analyses just keeps pointing towards g being the single best predictor of performance.

IQ tests are stupid and have large language components to them. The large language components of the tests mean they essentially boil down to culture tests. If your parents were in the in group, you speak scientistese.

Your in group membership is what propels you forward in your career, not your intelligence. Race and Gender are predictors of career success as well.

Any proponent of IQ, I'd challenge to take an IQ test in Chinese, Irish, or AAVE.

This is largely a myth.

If you were trying to get a job in a Chinese workplace with chinese coworkers that spoke chinese, taking a chinese IQ test would be a great predictor of success in that group. Doesn't mean you'res 'stupid', but my giant english IQ would be useless reading docs and taking meetings if they were all or majority in chinese.

What about tests that only use patterns/shapes/rotation/etc?

> We know that the single best predictor of job success and performance is IQ

I'm not sure a construction company wants a bunch of people with 140+ IQ standing around all day leaning on shovels waiting for gravel to shovel into little holes. I'm pretty sure those employees will just quit or break something on purpose just for something to do. In fact I've seen this with my own eyes.

Maybe you're proposing binning people by IQ instead? Where groups of people who fall in certain ranges of IQs are sorted into different jobs.

But again, I'm not so sure that Walmart wants to hire a shelf stocker with an IQ between 88-95 if that person has oppositional defiant disorder and/or a tendency to steal from their employers.

If you think that you can distill the essence of an entire human down to a single variable and rank them accordingly for jobs, go for it man, but I have a sneaking feeling that it isn't going to work.

There are a ton of people on construction sites with very high IQs that prefer manual labor, physical problem solving or managing divergent teams and contractors to accomplish a project. The notion of clustering people into professions for anything other than their desire to participate in the profession and ability to perform the tasks is a fools errand.

Thanks for this. I'm in Software because that's the best way for me to provide for my family. If I could have my way, I'd be outside working on making things that last longer than a couple of years.

> I'm not sure a construction company wants a bunch of people with 140+ IQ standing around all day leaning on shovels waiting for gravel to shovel into little holes. I'm pretty sure those employees will just quit or break something on purpose just for something to do. In fact I've seen this with my own eyes.

I'm quite convinced that a typical construction site would benefit a lot from employees who would be smart enough to notice inefficient processes and plain mistakes, and take some initiative to fix the problems. And management who would believe the employees, but that is probably never going to happen.

Ah so something like Chinese Gaokao? where you spend 3 years of your life preparing for one exam? Or am I reading what you wrote wrong?

Sure, let me use Raven's progressive matrices to test whether someone has the ability to make plans, stay organized, keep on task and interface with others in a productive way.

Sounds like a smashing idea, I'm sure you'll kill it in business.

One would think if it's such an obvious failure it wouldn't need to be banned: https://en.wikipedia.org/wiki/Griggs_v._Duke_Power_Co.

I'm going to take a different path from the other responses so far. The claim that IQ predicts job success seems to be contentious among a lot of people, but as far as I can tell, it is fairly well supported and other ways of evaluating candidates are at least as noisy anyway, short of professions with the sort of concrete standards that allow for meaningful licensing or something like being a military officer where they're going to train you and evaluate your performance in context for 4 years before ever giving you a real job anyway.

However, thanks to the way mental traits get distributed, a fairly small proportion of the population is ever going to have a high IQ. Roughly 16% of the population at any one time will have greater than 1 SD north of median. Less than 3% will be greater than 2 SD.

More people than that need to have jobs. If you're going to insist only 97th percentile people can work for you, you need to be offering 97th percentile salary. Currently, that is around $220,000 a year in the US. "Most" jobs can definitely not offer that.

> fairly well supported and other ways of evaluating candidates are at least as noisy anyway

It's on you to back up this claim. And the reason had better not be 'oh a bunch of tech companies do it' when there are entire fields from psychology that do research on this exact question, and they consider things like personalities, etc.

You also might be confusing higher-iq job scopes for job success.

A degree is a signal of IQ + Long term planning + Effort.

Plus financial backing, and/or inside influence on the process. Cheating is rampant is some cultures, and I guess some people even basically but their whole degree

> We know that the single best predictor of job success and performance is IQ.

WRONG. WRONG. WRONG.

IQ predicts how fast you learn shit, not how well you do on the job. Plenty of studies from the fields of personality/social psychology/psychometrics talk about this, get out of your narrow SWE field and read up on these subjects. This IQ=job success belief is simply SWE self ego-stroking.

edit:

Jordan Peterson (psychology phd) - The Mystery of High IQ and Industriousness https://www.youtube.com/watch?v=1C0zS2RAzlI

Actually, I'm pretty sure the GP was referring to this: https://psycnet.apa.org/record/1998-10661-006

>We know that the single best predictor of job success and performance is IQ.

Wow, so many people react so sharply to this, I'm not sure why. May be there is an inherent belief that all one has to do is work hard and success will follow.

Minor nitpick - I think performance has a good correlational with IQ, I think less so a job, because a job also generally requires high conformity (not an enviable trait IMO)

You’re assuming that the purpose of higher ed is to teach job-relevant skills. That is false. Higher Ed is an apprenticeship program for creating new instructors and researchers. Most people stop that process early before finishing for a variety of very valid reasons. There’s no reason to expect a degree to exactly prepare folks for specific jobs in industry.

It turns out that there’s very high correlation between the characteristics that make successful students and those that make successful workers. So a degree is still a useful signal during hiring, but it doesn’t certify a specific level of competency or skill set.

> Higher Ed is an apprenticeship program for creating new instructors and researchers

That is completely wrong when applied to e.g. Engineering. And completely true when applied to e.g. Philosophy.

CS courses on the other hand vary in their position on the spectrum.

There's a few fields, such as software engineering, where bullshit is in the long run difficult to hide. At the end of the day, you can either build a thing that does what people want or you can't. That is not true of most other fields. If your ideas about e.g. how to teach kids to read are wrong, it may be a long time before you're found out: https://www.nytimes.com/2021/09/03/opinion/kids-reading-spel....

Don't discount how long it takes for a company to uncover a mistakenly hired "Brillant Paula Bean"[1]. I know someone who literally can't code but can talk a good talk and has those soothing and smooth "Ivy Leaguer" mannerisms that executives and hiring managers just love. A total bullshitter. Most BigTech can filter these people out at the interview stage, but evidently smaller companies can easily be fooled. As soon as you're discovered, you can pretty easily hop to another "victim" company, especially during times like now when the job market is hot.

1: https://thedailywtf.com/articles/the_brillant_paula_bean

This may be a phenomenon limited to the tech field.

In many other fields that have advanced degree requirements, companies care far more about where you went to school and who you studied under.

For example - I have a friend that is an occupational therapist. They've had 5 different jobs in the last 10 years. Every single interview they've done has been mostly just a meet and greet. No real difficult questions, no tests, no trial periods. They just applied, the company saw their degree and experience, and the interview was simply there to make sure they could work together.

Even if academic qualifications lose their value, there is still value in the actual education provided by academia. If I want to learn a topic really well, especially a more theoretical rather than applied topic, going to do a 3 year degree on the topic at university is a pretty good starting point.

Speaking as someone who tries to elevate the careers of others based on their potential independent of their credentials…

I think if you’re working with really advanced topics, the credentials start to be meaningless, because the important fact is whether a candidate can truly work with complex concepts/technologies. You expect the candidate to come in already knowing what they need, and you don’t need any assurance that they’re “trainable.”

But for most of us here in the middle of the pack, the degree is more meaningful because it shows that at the very least, a candidate has been exposed to slightly higher-level concepts. As an example, when you’re trying to get someone to step up out of a help desk role into an analyst/developer role, that’s a lot easier to achieve with someone who’s had at least a degree’s worth of exposure to concepts of strategic thinking, lifecycle management, development methodologies, etc. Most of us don’t need to be passionate geniuses to do good work. For us non-geniuses, the conceptual “scaffolding” provided by academia opens up a lot of potential. I definitely believe that some people don’t need the degree, but they probably aren’t the average Joe.

> It's essentially saying, we don't believe your degree says anything about your ability to perform the job.

Unfortunately I have worked with people with a master’s in computer science from an american university who couldn’t write a coherent sentence in English. Let alone write code or design a system.

I’ve also interviewed people with illustrious careers who just weren’t a fit for the role they applied to.

Credentialing is in poor shape in our industry.

Yeah coding is funny, because the code has to work. Theres no getting around it, and most resources for understanding how to get things to work are available on the internet.

When the organization is big enough there is plenty getting around it. Also there is a lot of wiggle room around the definition of "works".

I once met an “angular 5” developer. Coding interview I started on the whiteboard.

var a = 1; a = 2;

Before I could write anything more (this was just setup) the person pipes up “that’s false!” We went through the rest of the interview but I just couldn’t believe it. The individual had been working for a government office as a software developer for 5 years or so before this.

I have no idea what this persons work day looked like but still shake my head when I think about that interview.

Wait what about that is false? It’s not even really doing anything yet

If you think like a mathematician, "a" can not be equal to 1 and 2 at the same time. There are languages that actually use := for assignment to emphasise it's not equality.

This is due to the warped incentives in Higher Education. Universities get more tuition fees from accepting more students, and achieve better in national rankings by having higher graduation rates.

Professors are stuck in the middle: asked to teach unprepared students, and forced to pass many who are not suitable.

The people it hurts the most are the students who are actually capable, and put the time in to excel.

Not sure if this is a European thing but in the German speaking countries of Europe there are universities and vocational universities. The first focus prepare you for an academic career and the second prepare you for entering the work force (with some academic knowledge in case you need it in the industry or want to pursue an academic career).

I think this makes a lot of sense and we have a system that allows you to get a well paying middle class job by completing a vocational education and offers paths to get higher education from there or even switch to the academic track. This always made sense to me and I think it's crazy how this is in other countries where the choice is going to University or learning your profession on the job.

The purpose of education should not be the attainment of credentials, but rather the attainment of the ability to "do a leetcode interview". Institutions being given license to grant credentials is a more or less completely separable problem from this.

No university has any incentive to say no to someone who wants to pay 200k to have you grade some homework for a couple of years, as long as they aren't at capacity, and moreover very small incentives to fail them.

The irrelevance of the paper as a signal is the solution to this problem.

Leetcode interviews don't showcase software engineering skills, they showcase your ability to solve leetcode problems. Those are two very different things

The uprooting of universities and their subsequent erosion as research university in the Humboldtian sense (in the US the John Hopkins University can be considered its first model) of "unity in teaching and research" based on a holistic concept of Bildung (education) in contrast - but also in addition - to a mere vocational training has very much to do with the neo-liberalization of higher education.

In Europe this became known as the "Bologna process" in which established "academic degrees" literally - in a EU-bureaucratic hunch and self-aggrandizement - got thrown away overnight (in Germany: Diplom, Magister) and then modernized, accordingly rated and standardized in a modular fashion, repurposed for the international/European (labour) market. The faculties and universities themselves from then on were run in a more and more managerial fashion with "operating numbers" and by winning state-funded budgets ("exellency" programs) and by efficiently recruiting third-party sponsoring.

For a typical (german) undergraduate student there isn't much a difference between school and university, anymore. Most Gen X'ers (and Baby-Boomers for that matter) will readily tell you how they for the first time had to learn to resort to self-reliant learning in want of detailed instructions (like in school) but also began to appreciate the less strict scheduling not only for partying but for developing their critical thinking in heated arguments with peers or when confronted with challenging ideas. This is mostly gone, now, a big part of that former public space has vanished.

For attaining craftmanship a particular skillset for a particular field, universities were never a good place to begin with, that's what internships and later your chosen vocations are for. But if they don't prepare you for the labour market and don't provide you with some kind of Humboldtian ideal - a universal starting point with sufficent leisure in pondering about the right/critical questions to ask in a nourishing environment - so what are they actually for?

Higher educations origins are in insulating Bourgeoisie kids from working class life.

As it became clear the people had lost belief in god kings, aristocrats began paying the church to educate their kids and confer nonsense degrees the proles could not falsify.

It was not hard learning guitar on my own. We create a lot of carbon mess for junior to get to an official guitar degree.

There is a whole lot of unnecessary LARPing deference to wealthy people who don’t do much of the real work keeping themselves alive.

It’s not a sustainable social meme to tighten the screws on the masses while handing their own kids the keys to the castle. 800,000 sworn LEO in the US, beholden to a paycheck and retirement account. None of the people at my local bank own that bank anymore. Why is my agency propping up someone else’s grandpa?

>> I think that the existence of leetcode interviews, where tech companies ignore your background as much as possible, and want to see the skills upfront, is a condemnation of education.

The funny thing is, all those leetcode questions are just the kind of thing that's taught at CS courses (algorithms and datastructures) and the kind of material that CS grads are likely to know well. And the people who interview you are more often than not graduates themselves.

Because, duh, tech companies are part of the same establishment as academic institutions. But don't tell them that because they all want to be disruptive, instead.

> It's essentially saying, we don't believe your degree says anything about your ability to perform the job.

I had a group project in 4th year university, in a very difficult math-heavy advanced graphics course, and my partner provided me with his half of the project and it was all global variables named A through Z, with a few A1,A2,AA,AB thrown in, and reused extensively for different purposes.

His code actually worked but it surprised me that someone could do 4 years of university (in a good CS school) but have almost no talent for actual coding.

Except that universities now prepare you for leetcode and those who studies in them do better at leetcode. By that I mean that there are actual classes to train you leetcode like exercises.

Everything that is measured will be gamed. No exceptions.

On one hand, I agree.

On the other, preparing for the test is not gaming it. Fundamentally, if there is a test, it is entirely legitimate and expected to prepade for it. Especially when the test measures skills you rarely use outside of test.

I've understood that the original idea of the Google interviews and others was to see if the applicant can write code at all, has been awake at second year CS class and can apply the ideas on the fly. The idea would have been that good candidates would be able to pass the test anyway. Of course it doesn't work if people start preparing a test that's supposed to be a proxy for other skills. For what it's worth, I've heard a rumor from Google that their interview system used to be quite OK at predicting employees' career development but competitive programmers consistently perform under the expectations because they've been specializing for events that are just like those interviews.

> I've understood that the original idea of the Google interviews and others was to see if the applicant can write code at all, has been awake at second year CS class and can apply the ideas on the fly.

This requires waaay more easier test. This goal does not requires escalation of leetcode exercises difficulty at all. If this was actual goal, the test could have been truly simple instead of requiring a lot of preparation to pass.

Especially:

> has been awake at second year

What is actual difference between learning those in second year of the school vs on your own when you find out employer wants it?

> This goal does not requires escalation of leetcode exercises difficulty at all.

Ok, looks like we've done different leetcode exercises. The ones I have done seemed to test mostly that I was able to sit down and type for an hour. Which, to be honest, is too much for many.

> What is actual difference between learning those in second year of the school vs on your own when you find out employer wants it?

The idea is that the employer should be able to give you real problems and trust that you have the basic toolkit to solve them without too much handholding. At least I have never been asked to implement this or that algorithm at work, but there have been quite a few times when one was needed to solve a problem.

most people are pretty average (tautologically so), and average is pretty unimpressive against the pace at which technology (in the broad sense) has changed and knowledge has so rapidly expanded in the past century or two (which is not to say that we're all that much more advanced yet against the backdrop of everything we could ever collectively know). as such, and compounded by sociopolitical contentions, most people are barely functional in their jobs, let alone adequate or good. that specifically includes so-called "experts", who tend to be further out on the edge nodes of our collective tree of knowledge and therefore more vulnerable to rising to your level of incompetence. once at that level, the competiton turns toward signaling your competence, rather than working on becoming more competent, since that's harder and less socially rewarding.

we need nothing short of a cultural shift toward incentivizing real progress and productivity, and valuing real knowledge rather than education, which we only had for a fleeting moment in the US (middle of last century), that only comes from really constraining markets toward both fairness and risk tolerance. we lost our way when we loosened our belt in that regard (for instance, we now instead reward to act of having capital rather than deploying it strategically).

Or the leetcode interviews are an attempt at a second layer of screen. Looking at the school people went to may not be a great predictor of their differential success versus everyone else who went to a similar school, but it can give a good signal (on average) versus someone applying without one.

In the same way that height among professional basketball players may not be a great predictor of their performance relative to other players, but you don't see too many 5ft players eithers.

Are you seriously suggesting that we not determine a candidate's ability to find all palindromes in a string for a Django gig? Fucking rebel you are.

> It's essentially saying, we don't believe your degree says anything about your ability to perform the job.

It also says:

1. We don’t trust your former employer to restrict promotion to quality employees only. After all, people have to do the leetcode interviews all over again when applying for their next job.

2. If you’re new to the industry and weren’t enough of a chump to take out student loans, you’re welcome to try your luck here

> It's essentially saying, we don't believe your degree says anything about your ability to perform the job.

That says more to me about the companies practicing this style of interview than it does your degree or the institution granting it.

I think that the existence of leetcode interviews, where tech companies ignore your background as much as possible, is a condemnation of my 20+ years of building real-life systems. And a means of filtering out people over 40.

Reply on news.ycombinator.com