4 ms·

This is such a serious problem. It took me a long time to understand why this happens. I remember going through the interview paces at a small company where I

by geebee 11y ago

This is such a serious problem. It took me a long time to understand why this happens. I remember going through the interview paces at a small company where I had a specialization I knew was very valuable to them. Instead, it was all technical quizzes, tree traversal, outer joins, swap two integers without creating a third integer, find the long term state probabilities in a markov chain. I remember thinking, wow, yesterday I was doing a demo of a large, mathematically involved supply chain system to clients at a huge manufacturing company, and here I am, getting paraded from room to room, where recent graduates are asking me how to find the least common ancestor of two nodes in a binary tree. There are a couple of factors going on here. First, a lot of big companies are in a state of constant recruitment, and they aren't really sure they know what they need you for. All they know is that if you can find the least common ancestor for two nodes in a binary tree, adapt merge sort to manage a search where there isn't enough memory to hold all the data at once, and prove that the dual of the primal is the primal of the dual, and you seem normal enough, then they'll have something for you. The other problem is IP - perhaps they view it as inappropriate to ask too much about what you're working on. Part of this problem flows from something very wonderful about our industry - we don't stand (too much) on formal credentials. People don't care all that much about how you learned nonlinear optimization or data structures and algorithms. As a result, the barrier to entry is… well, not low, because any entrance exam that requires proving things about convex sets and doing elaborate algorithms at a whiteboard is not a low barrier to entry, but low in the sense that institutions like law schools can't force everyone to go through three years at $45k+ tuition as a legally enforced barrier to entry (like, you'll be put in prison if you try to practice without the three year, $120K+ tuition degree that many people believe is excessive). The downside is that because software developers don't have a formal system for demonstrating competence in a way that is widely respected by our peers, we essentially have to take our entrance exams over, and over, and over. Actuaries (last I checked, the exams may have changed) have to take a test showing competence in differential equations, vector calculus, and linear algebra, but I don't think that actuaries with 15 years experience are suddenly asked to find the igon values (sorry, couldn't resist) of a matrix or saddle points in a nonlinear optimization problem with non-linear constraints on the spot during an interview. Physicians have to pass boards and exams, yes, but I don't think they're asked questions from their undergraduate organic chemistry midterm every time they interview for a job. Slowly, I am starting to think that it would help if software developers could, as a profession, take greater control over how competence is established. I've heard others here on HN say that while they aren't in favor of legally required licensing, they would be happy to study and sit a very formal exam if it meant they could finally be done with the whiteboard interview exams, at least for a period of time. I say this while acknowledging real fear of regulatory capture, but what we have now is really, really bad.

tomp 11y ago
> The downside is that because software developers don't have a formal system for demonstrating competence in a way that is widely respected by our peers, we essentially have to take our entrance exams over, and over, and over. I don't think it's really possible to have a meaningful and useful standardized formal exam. The issue is that programming, as a profession, is so wide and diverse that it's really hard to formalize what you're good at. Furthermore, even if you're quite specialized, a large part of your day-to-day work will still be "general" skills, e.g. debugging stuff, fixing other's bugs, taking care of backwards compatibility and working around system bugs. This requires a general level of knowledge of OSes, languages, implementations, databases, networking, ... At least in my experience, but I assume it's similar even in more specialized areas, like security.
- geebee 11y ago
  Yeah, it may not be possible. Keep in mind, though, that there are plenty of things about being an actuary that aren't tested in the actuarial exams. That doesn't stop them from requiring that actuaries pass an exam involving calculus, linear algebra, and so forth. And as a result, actuaries don't have to go around re-studying integration by parts every time they want to change jobs. I've been on a few interviews at different silicon valley companies, including Google and Amazon, as well as smaller ones, and the technical screening is remarkably similar. I've noticed it is far more rigorous at some companies than others, but they all more or less test dynamic programming, recursion, trees, lists, hash tables, sorting, and permuting. Even if we all agree that there is no way to have an exam that captures the wide, diverse nature of programming, it could still be useful to have an exam that covers that part that we test over and over and over. There's one other wrinkle that may not work here - many companies say they do this because they want to see how you tackle a difficult programming problem, how you communicate your solution, and so forth. As a result, they might not be satisfied with an exceptional performance on a similar exam, even if they have high confidence in the board that created the questions and conducted the exam. Now, all that said, I do consider this to be a really serious problem. Why? Because people who can study for and pass a difficult whiteboard exam on algorithms and combinatorics can pass the actuarial exams, the bar, the medical boards, and so forth. I think that the software industry loses a lot of talent because people don't like the idea of signing up for an industry where they will have to pass their entrance exams over and over, every time they change jobs, and the people who are capable of this have other options. I think this is costing us a lot of talent, it may cause people to favor other fields, and it may cause people to give up and drop out. There are some risks to a formalized test (cartel building is a serious risk, I recognize this, and it worries me - as a math major rather than a CS major, I might be one of the first excluded). At this point, I don't want to seem like an advocate for this. I'd say I feel ambivalent about it, but that's a change from where I stood on the issue a few years ago. I'm starting to believe that the damage inflicted by what some people believe is "interview hazing" may go very deep. Here's a hint at the extent of the problem... http://www.fastcompany.com/3043082/most-creative-people/why-software-maker-fog-creek-is-helping-its-competitors-hire-women http://www.fastcompany.com/3043082/most-creative-people/why-... "Hall asked a simple question: Why didn't Chipps, a brilliant engineer, ever apply to work at Fog Creek? Chipps said she didn't think she would make it through their rigorous testing process." One last thing - people certainly can agree that there is a problem without agreeing on a solution, especially one that involves a standardized exam.

by geebee 11y ago

tomp 11y ago

> The downside is that because software developers don't have a formal system for demonstrating competence in a way that is widely respected by our peers, we essentially have to take our entrance exams over, and over, and over. I don't think it's really possible to have a meaningful and useful standardized formal exam. The issue is that programming, as a profession, is so wide and diverse that it's really hard to formalize what you're good at. Furthermore, even if you're quite specialized, a large part of your day-to-day work will still be "general" skills, e.g. debugging stuff, fixing other's bugs, taking care of backwards compatibility and working around system bugs. This requires a general level of knowledge of OSes, languages, implementations, databases, networking, ... At least in my experience, but I assume it's similar even in more specialized areas, like security.

geebee 11y ago
Yeah, it may not be possible. Keep in mind, though, that there are plenty of things about being an actuary that aren't tested in the actuarial exams. That doesn't stop them from requiring that actuaries pass an exam involving calculus, linear algebra, and so forth. And as a result, actuaries don't have to go around re-studying integration by parts every time they want to change jobs. I've been on a few interviews at different silicon valley companies, including Google and Amazon, as well as smaller ones, and the technical screening is remarkably similar. I've noticed it is far more rigorous at some companies than others, but they all more or less test dynamic programming, recursion, trees, lists, hash tables, sorting, and permuting. Even if we all agree that there is no way to have an exam that captures the wide, diverse nature of programming, it could still be useful to have an exam that covers that part that we test over and over and over. There's one other wrinkle that may not work here - many companies say they do this because they want to see how you tackle a difficult programming problem, how you communicate your solution, and so forth. As a result, they might not be satisfied with an exceptional performance on a similar exam, even if they have high confidence in the board that created the questions and conducted the exam. Now, all that said, I do consider this to be a really serious problem. Why? Because people who can study for and pass a difficult whiteboard exam on algorithms and combinatorics can pass the actuarial exams, the bar, the medical boards, and so forth. I think that the software industry loses a lot of talent because people don't like the idea of signing up for an industry where they will have to pass their entrance exams over and over, every time they change jobs, and the people who are capable of this have other options. I think this is costing us a lot of talent, it may cause people to favor other fields, and it may cause people to give up and drop out. There are some risks to a formalized test (cartel building is a serious risk, I recognize this, and it worries me - as a math major rather than a CS major, I might be one of the first excluded). At this point, I don't want to seem like an advocate for this. I'd say I feel ambivalent about it, but that's a change from where I stood on the issue a few years ago. I'm starting to believe that the damage inflicted by what some people believe is "interview hazing" may go very deep. Here's a hint at the extent of the problem... http://www.fastcompany.com/3043082/most-creative-people/why-software-maker-fog-creek-is-helping-its-competitors-hire-women http://www.fastcompany.com/3043082/most-creative-people/why-... "Hall asked a simple question: Why didn't Chipps, a brilliant engineer, ever apply to work at Fog Creek? Chipps said she didn't think she would make it through their rigorous testing process." One last thing - people certainly can agree that there is a problem without agreeing on a solution, especially one that involves a standardized exam.