Programming education is broken. I did one year of computer science at one of the top universities in the world (switched into mathematics after that), and I'd see that course as useless if not actively negative. The only way of learning that I've seen really work for anyone (myself included) is more like a craft apprenticeship, working closely with someone more experienced. We shouldn't be surprised that that produces widely different approaches.
Frankly the field isn't mature enough to have standards. If you tried to set a professional exam based on today's best practices, in five or ten years the answers would mostly be wrong. We still don't know the right way to write software. Million-dollar systems still come with laughably simple bugs.
What does the interview process look like for a craftsperson? That's probably the best we can expect from a field as unsystematic as ours. The one thing that strikes me is that in creative fields it's normal for people to show a portfolio of past (client) projects, whereas in software past employers usually keep all copies of your code. I have no idea how we'd go about changing that norm though.
Welcome to my shop. Here's some wood. Make a chair!
In most of the interviews I conduct, I get the candidate to write follow a spec we've written and some code. And I get the candidate to debug a program with some failing unit tests. About half of the candidates I interview fresh out of school have no idea how to get started debugging a program they didn't write. You need to do that just about every day at work, and its usually not even mentioned at school.
But I've worked as a teacher too. I will move heaven and earth for my students, but in my darkest moments I think taboo thoughts. Maybe IQ is a real thing, and maybe some people just don't have the neural hardware to ever be good at programming. If thats true, we do those people a huge disservice. We steal years of their lives and tens to hundreds of thousands of dollars training them in a skill they can never master. I'd love to see the stats on how many people graduate from a CS program, try but never find reliable work in our industry. I worry the numbers might be damning.
A few months ago a candidate I interviewed asked for feedback at the end of the interview. He wanted to know I recommended that he practice so he could improve. I said he should pick an opensource project on github - preferably something thats not too big and look through the issue tracker. Pick a simple looking bug and try and fix the bug and submit a PR. His whole manner changed after I said that - the idea of doing that was so incredibly daunting to him. But that right there? More or less, thats the hiring bar. As an interviewer I'm trying to answer the question "If I hire you, can you do the work?". Read, think, understand, modify, write code, communicate it to your team. Thats the work and thats the bar.
Approaching an open source project cold is a bit higher than the bar.
In a way, this is mirrored by the organizations themselves. Startups who make the right moves, and work hard through grit and 60-hour weeks, can become unicorns. Also some startups are led by visionary founders and cannot fail.
So all of this, buttressed by real world labor demands, create incentives for people to try to become programmers. Even those who aren't "cut out to be programmers." Our increasingly cutthroat and unequal society also incentivizes people shifting to programming as a safe career choice. "Learn to code."
I don't know how we can stop "pretending otherwise." Tech companies continue to complain about the engineering talent shortage. Bootcamps and online courses continue to promise people that they can become that talent. There aren't any agreed-upon industry standards by which to exclude people who truly aren't fit for it. FAAMG has infinite money and power in the industry to continue their entrenched practices. Most startups cargo cult the leading megacorps' processes. So instead, candidates are encouraged to continue grinding Leetcode and apply, apply again.
Those seem like orthogonal aspects. If we stressed the idea that you had to do (say) a 4-year degree at a great university, or something akin to the bar exam, would that be any more compatible with the idea that some people are 10x better than others? If anything I'd say the opposite: we'd expect most of the Harvard graduating class to be on roughly the same level, it seems a lot less wild that some self-taught people could be 10x better than others.
> So all of this, buttressed by real world labor demands, create incentives for people to try to become programmers. Even those who aren't "cut out to be programmers." Our increasingly cutthroat and unequal society also incentivizes people shifting to programming as a safe career choice. "Learn to code."
This happens in every field though? You get people who are desperate to become a doctor and apply to med school year after year, despite being completely unsuited to it. You get people who insist they're gonna make it as an actor/musician/comedian and spend decades working crappy day jobs so they can live where the action is, when really they'd be better advised to pick a career they're good at.
> I don't know how we can stop "pretending otherwise." Tech companies continue to complain about the engineering talent shortage. Bootcamps and online courses continue to promise people that they can become that talent. There aren't any agreed-upon industry standards by which to exclude people who truly aren't fit for it. FAAMG has infinite money and power in the industry to continue their entrenched practices. Most startups cargo cult the leading megacorps' processes. So instead, candidates are encouraged to continue grinding Leetcode and apply, apply again.
Well, if we told people outright that programming is a matter of IQ, and gave an actual IQ test rather than an IQ-like test in interviews, that might help some people realise it's not for them. You're right that what catches on in the industry is largely a function of what the most successful companies pick, but ultimately that list of top companies is not static and we'd hope that companies with better hiring practices will (eventually) rise to the top.
That's not the point I was trying to make- I'm saying that in programming we both prize talent born of nature, and skill honed by nurture. (Though admittedly that may exist in many other disciplines.) Because of the latter emphasis on grit, hacker culture encourages self-improvement and going beyond the capacities one started with. That dogma of self-improvement goes against the notion that people are not cut out to be programmers.
Though of course, this could also be a marketing ploy for recruitment on behalf of management: "Anyone can code, you should learn to. But we only hire from the best." By encouraging an increase in talent, they have a larger labor pool to choose from (and potentially undercut wages), while plucking out the few that can pass their interviews.
> This happens in every field though?
To some degree, but the details vary. Medicine or law used to be seen as safe secure careers into the (upper) middle class, but doctors are limited through the AMA, and currently law is a notoriously difficult and costly profession with dwindling prospects. Entertainment and the arts is universally known as a risky proposition. We're talking about software, which has had the reputation of being the current surefire path to a stable, even successful, career, for at least the past two or three decades.
> Well, if we told people outright that programming is a matter of IQ, and gave an actual IQ test rather than an IQ-like test in interviews, that might help some people realise it's not for them.
Leaving aside the legality of using IQ tests to exclude candidates, that opens up the questions of if there is a direct correlation between programming good software and IQ, why programming out of all STEM fields should focus so heavily on IQ, and why all of those other technical and engineering professions don't need to resort to IQ tests for hiring.
So do your students do this? Why not?
I see it as like if I were hiring for something as generic as "writer". It's easy to have a generic "writer" produce a small sample for you on the spot, similar to a CS interview. Of course you can always practice writing directly itself, but I would imagine someone who had completed a lot of coursework in linguistics, classics, literature, etc. would on average be very well-prepared if they were a good student. But you could still practice and teach yourself on your own if you wanted to
Perhaps, but it's still the closest thing the industry has to a "programming education"; I think it's the first thing employers look for, rightly or wrongly.
> it's kind of silly to judge a whole subject by the first year
How many of my limited days on the planet am I supposed to sink into something before I'm permitted to pass judgement? At some point Stockholm Syndrome would take over.
> A good CS student, who understands the coursework and doesn't cheat, should easily become a good enough programmer just from completing coursework in a mostly theoretical program to get hired basically anywhere. Programming is something you learn incidentally because it's intertwined with what you're doing anyway;
That's not what I saw happening (unless you count the official lectures/colloquia as "cheating"; certainly I saw cases where the meat of the answer to a supervision question was spoon-fed to us directly). The people who could program at the end of first year were the people who could program at the beginning or who "got it" immediately. I never saw people struggling with a new concept but then gradually being taught it (which is something I did see happen a lot in the mathematics course), and a frightening proportion of the students I was friendly with were coming out of that first year knowing seemingly nothing, certainly not being able to program or talk coherently about algorithms or computability. I suppose it's conceivable that those students were somehow getting something out of the system design type courses, but it seems implausible.
I understand there was a shake-up in that CS department a few years after I graduated, so maybe I went through it during a bad time. But the students who graduated there in the meantime aren't going to get a do-over.
> I see it as like if I were hiring for something as generic as "writer". It's easy to have a generic "writer" produce a small sample for you on the spot, similar to a CS interview. Of course you can always practice writing directly itself, but I would imagine someone who had completed a lot of coursework in linguistics, classics, literature, etc. would on average be very well-prepared if they were a good student. But you could still practice and teach yourself on your own if you wanted to
I'd suspect the overwhelmingly important part of writing is actually writing; I only know one person who I'd call a great writer, and hanging out with him the thing you notice is that he writes the way other people check their phone. All the things you list can enhance writing, certainly, but if you don't actually write then any amount of knowledge of linguistics or classical literature is meaningless (at least in terms of how it affects your writing ability).
More than the first quarter of it. I took at least a year of math classes, and I don't feel qualified to pass judgement on the math department.
There's a huge difference between intro and upper level classes. Just like there's a difference between Calc I and proof heavy upper level math class.
That being said I think the overall pedagogy is better in the math department, but then again they've been doing this for a lot longer.
I don't have a medical degree, but I still trust medical science. Why? Because it achieves positive results, and people I trust for other reasons trust them.
There's no you were rigorous enough in tracking this for this statement to be useful to anyone. The people who don't fit the mold stand out.
Sure, but has anyone claiming the opposite done rigorous analysis? Is there any evidence that having a CS degree makes for better programmers than not?
> The people who don't fit the mold stand out.
I'm not thinking about people who stood out as particularly unusual. Most of the time I didn't find out which field someone's degree was in until months into working with them.
I'm not making that claim, you're the one making a claim that people with degrees other than CS are better programmers without evidence.
I'll only make the claim that a CS degree made me a better programmer. Specifically the upper level theoretical classes. I can verify that there are many problems I've solved because I realized that the problem I was working on had already been solved 50 years ago.
I also worked about a decade as a professional programmer without a CS degree, before I went back. Personally I am a better programmer.
Would I have been an even better programmer had I taken another few semesters of math classes instead of CS classes? Who knows? Absent any other evidence though, the simplest explanation is that domain specific knowledge is likely useful.
>I'm not thinking about people who stood out as particularly unusual. Most of the time I didn't find out which field someone's degree was in until months into working with them.
The point is that the more unusual someone's background is, the more likely you are to remember it. Particularly if there is some confirmation bias involved.
Disagree; surely the null hypothesis for any given training programme is that it has no effect.
> The point is that the more unusual someone's background is, the more likely you are to remember it. Particularly if there is some confirmation bias involved.
There isn't anything unusual about professional programmers having a degree in maths or physics rather than CS. At least in my experience it was pretty close to an equal split.
That's not what's under test here though. It's training program A that includes domain specific knowledge or training program B that does not.
>There isn't anything unusual about professional programmers having a degree in maths or physics rather than CS. At least in my experience it was pretty close to an equal split.
Look at the number of graduates, the only way that is true is if almost every single physics or math graduate goes into programming. The fact that you think it's true is just further evidence of bias.
According to the Stack Overflow Developer Survey [1], about 8% of professional developers with degrees majored in math, or natural sciences vs. 63% in CS, software engineering, or computer engineering. There could be some sampling bias, but that's a huge difference.
I suspect SO surveys are heavily biased towards younger developers, but even taking those 2019 numbers at face value: about 20% of professional developers have no degree, and of those with degrees it's about 75% CS/information systems/sysadmin/webdev, 17% maths/physics/engineering, and 8% other. So a typical 15-developer team would be 9 with CS degrees, 3 with no degree, 1 with an engineering degree, 1 with maths/science and 1 other. The non-CS folk are not exactly rare unicorns.
Well then why did you say this:
>There isn't anything unusual about professional programmers having a degree in maths or physics rather than CS. At least in my experience it was pretty close to an equal split.
You said explicitly math and physics degrees vs CS degrees. And you previously said the best programmers tended to have math, or physics degrees or something similar.
This isn't me being pedantic, it was the entire context of the discussion.
The point is that you are prone to confirmation bias as evidenced by your belief that it's close to an equal split. Your mental model is overrepresenting people with physics and math degrees likely because it confirms your belief that they are better programmers.
>So a typical 15-developer team would be 9 with CS degrees, 3 with no degree, 1 with an engineering degree, 1 with maths/science and 1 other. The non-CS folk are not exactly rare unicorns.
That's not the point, it's that you are more likely to remember the background of the 1 person on a team who has a Math degree because she is relatively rare compared to all of the people with CS degrees. This is a well known and well documented phenomenon. And it's one of the primary reasons that anecdotal evidence, even a large amount of anecdotal evidence is so often wrong.
I was giving those as examples of degrees that are normal and don't stand out. We don't think there's anything particularly odd about a programmer with a maths or physics degree. That's all I was saying.
> you previously said the best programmers tended to have math, or physics degrees or something similar.
A category which would include engineering, at which point we're at 20-25% of professional programmers with degrees by your numbers (which I still think are significantly biased).
> The point is that you are prone to confirmation bias as evidenced by your belief that it's close to an equal split. Your mental model is overrepresenting people with physics and math degrees likely because it confirms your belief that they are better programmers.
My mental model is that it's an equal split between CS degrees and not, and per your own sources that's accurate. You're fixating on a couple of specific examples of non-CS degrees that I mentioned when that's completely beside the point.
>The best programmers I've worked with have mostly not had CS degrees (tended to have degrees in maths, physics, or that sort of area).
That's the entire context of the discussion. The assertion that CS majors have worse outcomes (with respect to programming ability) than math, physics or similar majors.
>A category which would include engineering, at which point we're at 20-25% of professional programmers with degrees by your numbers (which I still think are significantly biased).
If you are including math, all natural sciences, and all other engineering degrees you get 17%, not 20-25%.
>My mental model is that it's an equal split between CS degrees and not, and per your own sources that's accurate. You're fixating on a couple of specific examples of non-CS degrees that I mentioned when that's completely beside the point.
There is no other logical way to parse this statement
>There isn't anything unusual about professional programmers having a degree in maths or physics rather than CS. At least in my experience it was pretty close to an equal split.
than that you were talking specifically about math and physics.
I get it, you don't like that there are numbers that contradict you, so you are grasping at straws trying to find alternate interpretations to reconcile your statement with the numbers. You obviously don't like being wrong. I don't either, that's fine, but no one other than us is reading this far down. There's not point denying you farted when there's only 2 of you in an elevator.
I don't entirely disagree with you, there were certainly some later classes that were similarly not useful in the long run, but it wasn't a total crapshoot. My senior year included a year-long group project on a team of ~10 people that basically took us all the way through the lifecycle of a project -- from inception, to design and architecture, to polish and QA, to 'releasing' it. It was a very useful course that placed you into a startup-like atmosphere.
But I think this kind of confirms that apprenticeships may be far more useful to the programming field than college degrees. If my CS program did not include that senior course, I would probably pretty vehemently agree with you. And anecdotally, at my current workplace, one of our best programmers is a kid we hired basically right out of high school who has since grown in skill considerably thanks to the attention of more senior engineers.
Of course it is broken, it rarely mention naming, debugging, and never emphasize reading code. The real fundamentals are not there and you learn them on the job.
So does this mean that programming education is broken? That companies should invest more in training? That bootcamps should revamp what they teach? That there should be industry standards for what programmers at different levels should be expected to know?
The most maddening thing about interviews imo is the opacity. There's always uncertainty about where exactly things went wrong, and how it should be fixed next time. It's almost a meta-engineering problem of its own. And it's a metaphor for the lack of software engineering standards.