Comment by FabHK | Hacker News Reader

Comment by FabHK | Hacker News Reader

That is a terrible list. Many of those questions are perfectly fine.

10 - Fermi problems. Good to get an indication of a candidate's problem solving skills.

9,8,7,6 - The candidates's motivation, weaknesses, strengths, plans. What's wrong with that? Evidently, you have to modulate the answers somewhat, but the extent to which you do that well is indicative of your understanding of the job requirements, the competition, etc. Perfectly legitimate.

5 - Best candidate. Again, this is a fine question - it allows you to highlight your understanding of the job, and your fit, and your accomplishments, and so on.

4,2 - If you've had unusually short tenures at prior jobs, why? Again, perfectly legitimate question. There might be some reason pertaining to the prior company, or there might be a character trait with the candidate. Insofar as training a candidate incurs some fixed costs, it is legitimate to aim to find out whether they'll stay around for a while.

1,3 - "bad habits", family. Agreed, irrelevant.

But 2/10 is a pretty bad hit ratio for a top 10 list.

> 10 - Fermi problems. Good to get an indication of a candidate's problem solving skills.

Well, you could ask them to solve a problem that might realistically come up on the job.

There are a lot of professions that require problem solving, but they don't use Fermi problems.

Interviewees for car mechanic or doctor positions don't get asked how many piano tuners there are in Chicago. They get asked questions like "A 2021 Prius is emitting a high-pitched whine and exhibiting poor acceleration from stop lights. What is the likely cause?" or "This patient has a blood pressure of foo/bar, a creatinine level of blah and has this other symptom blugh. What would be your next step?" If you gave them a Fermi problem they'd probably walk out the door.

Many of these "interview questions" are basically just cargo cult stuff. The company has heard that Microsoft or one of the FAANGs has used some weird-assed question at some point, and thinks that if only they do the same, they'll be the next FAANG. It's a tacit admission that the hiring manager actually has no clue as to how to evaluate a candidate based on metrics that are actually relevant to the job.

Do mechanics and doctors actually get asked those questions? I don't think doctors do, they have credentials and references. I'm pretty sure mechanics around here wouldn't get asked that, just references and maybe credentials and "how soon can you start?"

Software is a weird industry where you have to quiz every applicant to make sure they know how to program even though they've been the industry for years, and we seem to get it wildly wrong anyway.

Coincidentally, although I've been offered jobs after doing these kind of interviews and technical tests, I've only ever accepted jobs that are on the basis of "we know you, just come along" and I guess I now don't have any trust in a full interview process.

You're right that doctors have a well-established credentialing process, but yes, I believe that they are asked questions like that. Or maybe "For a herniated bupkis, some doctors prefer the Bloggs reduction, while others prefer the Quimby resection. Which do you prefer and why?"

There are nearly as many poseurs in mechanics as there are in programming, though. I think reputable shops grill candidates pretty hard. And not with Fermi questions.

I guess the only doctors I've known are ones going through the state run education and health system and it seems pretty streamlined. You go and work where you're told to based on your educational choices and achievements, there aren't interviews as such. It must be different going into a private health system.

Nah, software is weird because they so rarely ask people to do anything related to the task at hand. "solve a toy problem without using libraries. Please don't paste off SE/AI".

I prefer to actually think about what I want and devise a problem set to work through with the last candidates. You know, "here's a simple TCP server with a minor bug in the handler for one packet. Let's fix it". Real job: here's a 100k LoC server. Add a report/packet type/etc or fix a bug...

Uses up my time as well as theirs, but shows me a lot about how they work as well as what they'll be like to deal with.

I'd even be happy to simply talk out concepts, issues, troubleshooting and possible things to check.

Generally, if a person can talk in detail about a topic, they know the topic.

But it requires you also know the topic.

Many software applications actually require estimates and thinking similar to how you would solve Fermi problems. Like, how many disks are needed to backup our database every week with 6 months retention?

It's actually a very straightforward equation, but way too often people just trial-and-error or make an assumption out of nowhere, instead of plugging in the disk-size, user-count and growth rate multiplied with size per user.

Same goes for understanding what a computer is capable of, instead of throwing in some complex kubernetes cluster because requests are processing slow. If you actually knew the order of magnitude of queries per second that is normal, you would question the application first before questioning the number of servers.

> Like, how many disks are needed to backup our database every week with 6 months retention?

Then maybe ask that question instead?

But even this question is not relevant to the vast majority of programming jobs. It's more of a devops question.

> Interviewees for car mechanic or doctor positions

Because those are different jobs. A good software engineer must be able to reason around all sort of real world problems. Software does not exist in a vacuum.

I'm afraid I'm going to have to disagree with you that a software engineer needs better problem solving skills than a doctor. Strongly.

I actually disagree on 10, I just estimated the population of New York then looked it up and was off by an order of magnitude. Good chance I'd do worse in an interview situation than sitting at home. A few more estimates with errors of this size and we've learned what? I have a poor grasp of US census data? Unless this is a realistic approximation of the kind of thing your doing with this job, the time in the interview would be better spent on a problem within your domain, but that means the interviewer has to spend time coming up with something instead of Googling "interview questions". See also: "why are manhole covers round?" Which is a real question we asked all of our candidates at a previous job right up until I got control of the process (fun fact, invalid assumption. They come in all shapes and triangles are surprisingly common)

But agreed on the rest. OP thinks they are bad, when really they are poorly worded, and they would be better off taking a moment to think what kind of information the interviewer is trying to get at with their questions. Heck if the job involves interfacing with other people the ability to do this can even be an important signal on if the candidate is right for the job. A lot of people are really bad at asking what they are looking for.

"why are manhole covers round"

So they don't fall in. Next question.

2023 answer: Excuse me? You mean a street-level maintenance access lid? So it can't fall in. And why are a bunch of the other similar lids not circle? I leave that one up to you vape hit

http://sellsbrothers.com/12395

Round shapes handle the pressure of the surface and soil better is the true reason I have heard.

The goal isn’t to see how close you can get to the correct answer, it’s to see the extent to which you’re able to break down a complex problem you haven’t seen before into steps you can take to solve it.

And are you sure the person on the other side of the table can assess your problem solving skills correctly? That would make them an expert on Fermi-like problems. OTOH, perhaps HR really knows how many civilizations are out there in the universe.

There are no guarantees the other person knows anything, but they don’t need to be an expert to confirm you can break down a problem

Correctly? I could be bullshitting the heck out of it. I think I could teach anyone to pass such questions, if not vetted. It seems to boil down to: initial estimate, then multiply by a fraction for each narrowing specification in the question (how many red-haired, left-handed, gay piano tuners are there in New York), dependence between factors be damned, if it doesn't matter if the outcome is 1 or 500.

the goal of these tasks is not to provide the most correct answer, but to show how you think about problems. Even the fact that you googled the right answer after may set you apart from candidates who just don't care, and that's relevant for the job. Of course if the recruiter judges candidates based on how close they were to the correct number, then they are doing it wrong and you might want to stay away.

> 10 - Fermi problems. Good to get an indication of a candidate's problem solving skills.

Calculating the number of pencils used by Chicago's public school system in one year or whatever is pointless, though. I could "solve" this problem, or at least make a reasonable estimate, I suppose, but I'm a web developer. Wouldn't you get a better indication of my problem solving skills if you asked me about a problem related to web development?

A question like "what are some problems you ran into on previous projects and how did you solve them?" will be one I can answer instantly as well as give you a better indication of the competence I have doing the job you actually want to hire me for.

Reply on news.ycombinator.com