[hpcjoe]

Not true. I've got a Ph.D. in theoretical physics, am very math heavy, created a number of algos for my work over the last 25+ years (in physics sim, bioinformatics, systems management/orchestration, etc.), run sessions at an ACM conference, yadda yadda yadda.

Two google interviews, and nothing. From what I hear from other people I consider way smarter than I, they also got nothing.

Google has a much copied process, but as a creator of something of huge value notes:

https://twitter.com/mxcl/status/608682016205344768

Mebbe their filter isn't quite as good as they think it is. Talking to a number of absolutely brilliant engineers who didn't get hired, it likely has nothing whatsoever to do with talent, algo knowledge, mathematics, etc. There are other factors.

Being an above 40 guy probably didn't help me, google and others seem to have lots of trouble with ageism.

Dan's article was not specifically about being the "top", rather, what does the "top" mean in context, and how do people judge. What is the opportunity cost of doing this? As he points out, as I point out, it can be very high.

The smartest programmer I met in my first decade of work, was a person who had a high school diploma. No college degree. The guy was brilliant, personable, humble. He is quite successful now, and still doesn't have degrees. Chances are, he doesn't have formal education around the math/algos, but has picked up everything he knows.

At the end of the day, hiring is something of a crap-shoot. Past performance is not a guarantee of future performance, either negative or positive. You are after passion, intelligence, fit, experience if it exists (re-inventing wheels can be time consuming/expensive if you are forced to do it, and getting a guide who has been down that path can save you making some mistakes/time/money).

I know people are telling themselves that google has a good process, but honestly, it looks like it enforces homogeneity more than it brings in needed talent. I am not sure this is a good thing. Poor replication of their processes is rampant throughout the industry. I am not convinced this leads to positive outcomes.

[auggierose]

That tweet just tells me not to use Homebrew.

Just kiddin'. I think, inverting the binary tree probably means mirroring it. I had an interesting Google interview as well a few years ago where I aced the automated coding test but then the first human interviewer didn't get why I said that regular expressions run in linear time :) Our background was just very different.

[Klockan]

> I've got a Ph.D. in theoretical physics, am very math heavy, created a number of algos for my work over the last 25+ years (in physics sim, bioinformatics, systems management/orchestration, etc.), run sessions at an ACM conference, yadda yadda yadda.

> Two google interviews, and nothing. From what I hear from other people I consider way smarter than I, they also got nothing.

Then you and your friends weren't fluent enough with algorithms. That is the point, they don't care about all of your degrees, years of experience, conferences etc, they care about your fluency with maths and algorithms. This means that even a person with a shitty background can get hired at Google while a person with a stellar background gets rejected. Should you have gotten hired? Probably, but their system lets them find a lot of diamonds in the rough who wouldn't get hired anywhere else which is why they use it.

[hpcjoe]

> Then you and your friends weren't fluent enough with algorithms. That is the point, they don't care about all of your degrees, years of experience, conferences etc, they care about your fluency with maths and algorithms.

I didn't fail those portions. Actually did quite well on them. So did my friends.

You are making a number of invalid assumptions, starting from the assumption that their processes are fundamentally accurate or correct. My supposition is from the viewpoint that all systems are fundamentally flawed, and the goal is to minimize risk associated with a flawed system.

I know it is generally hard to acknowledge that google does things wrong, but ... IMO (and I am fairly sure I am not alone here) ... they have a number of significant issues that they haven't quite moved past yet, and this is one of them. Remember, they started out with brain teasers, and school pedigree. The new system isn't demonstrably better IMO, but it helps them convince themselves that it is.

[Klockan]

> I didn't fail those portions. Actually did quite well on them. So did my friends.

Then I don't see your point, what are you saying caused you to fail? I have a physics degree from an unknown school, learned to code in my thirties and got a job at Google by just doing well at their algorithms and maths questions so it is definitely possible to get in without ticking any of the hip boxes.

[skellington]

His point is that Google's and everyone else's hiring process is subject to large amounts of randomness and capriciousness. Google themselves have at various times mentioned how their hiring scores don't strongly correlate to performance.

Don't feel because you got in that you are some ordained snowflake. If you had interviewed on another day or with another group within Google you very possibly wouldn't have gotten in.

There are many variables at work when it comes to getting hired and hiring.

[iainmerrick]

they care about your fluency with maths and algorithms

"Maths" is a red herring -- a physics PhD who's still active in academia will definitely be very fluent in maths.

It's all about "algorithms", but I think a lot of software people have tunnel vision about that. There's a lot of fancy terminology you pick up in a CS degree; requiring people to know that filters out a lot of potentially good candidates, unless they've studied CS in their own time.

That's fine if the special CS terminology is absolutely essential for all programmers. But is it really? Realistically, 90% or more of your time as a programmer is spent working on other stuff (automation, testing, designing friendly APIs, catching sneaky bugs, scripting, just generally plumbing stuff together). If you're on a team, does every single team member need to have a great understanding of data structures? Or is it just nice-to-have, specialized knowledge?

[monocasa]

> That's fine if the special CS terminology is absolutely essential for all programmers. But is it really?

I'd argue that it matters more for Google more than most employers. The combo of their scale, combined with their large amount of custom infrastructure, combined with their desire to be able to retask engineers on a whim, means that individual engineers will have pretty good chance of touching code where the choice of Big O could make or break a product.

[iainmerrick]

Mayyybe, I'm not so sure but you could be right. [Edit to add: even at Google, most programmers are not doing that kind of stuff most of the time.]

On the flip side, though, I think many programmers (even programmers who are up to date on their CS) are fairly weak at mathematics. We think we're good because we can, you know, invert a binary tree, but how about figuring out an appropriate filter to smooth some data, or verifying that some randomized process is unbiased?

For something like digital filters, if you have basic knowledge you can just look up wikipedia for the details. But the same applies to data structures and big-O!

A lot of companies (including Google) can benefit strongly from people with good maths or stats skills. Do those people also need to be strong in CS? Or if not, do they need to be siloed into a separate hiring process, and placed in separate departments?

I reckon CS, maths, stats and other specialized academic training should all be treated as nice-to-have skills, of varying importance depending on the team balance and project requirements.