discmonkey

Agreed that the hour invested may not be worth it for a large percentage of people (again I got lucky, I would feel so differently if I was rejected after 7 interviews).

Google's point is just that they prefer to avoid false positives at the cost of (potentially) a lot of false negatives. The current interview process is probably some local optimum;

I haven't worked at a company yet that is actually good at interviewing. Where good is optimized over high recall and accuracy and a small time investment.

I know that Google does have one big advantage, in that a good percentage of people that get the offer end up accepting. That (unfortunately) gives Google a lot more leeway and possibly less incentive to further optimize the interview process. Software engineers are in high demand, but also Google is in high demand among software engineers.

At my previous job, trying to find a candidate for a role essentially involved lowering the bar until somebody was no longer in demand, since we weren't "in demand".

For what's its worth, I really didn't think the technical interviews were nightmares. It was mostly talking with very intelligent people who were volunteering their time to screen candidates. Though how much one enjoys brain teasers over productive work is of course a matter of personal perference. Everything else (emails, phonecalls, etc) were definitely tedious.

As for working at Google, I think it really depends on the person. For me it was a good gamble in terms of both immediate learning and future opportunities; for others it won't be. I do find the opportunity to have my code (or what's left of it after reviews anyways) reach billions of people a thrill like no other though.

Having recently won the google hiring lottery, I wanted to add my experience

total length of time ~ 7 months emails sent/received ~ 50 phone calls ~ 20 technical interviews: 7 (1 with screen with recruiter, 6 with engineers)

team match interviews: 3 (for me at least these were real interviews, in that I "flunked" the first 2, and decided to take the 3rd one very seriously before getting an offer)

Before getting hiring approval, I honestly felt that I would get rejected, and I can honestly say that being more "personable" and very "communicative" probably had a lot to do with why I made it through.

My honest advice would be to stop thinking of these long interview processes in terms of a a binary outcome that you have complete control over. Instead, think of it as trying to maximize some hidden parameter of a binomial distribution, where every interaction slightly increases your odds in the final coin flip.

Gradient descent is a pretty common algorithm, and I've seen it implemented in everything from JavaScript to c++. I have to say in this case the notation is significantly more distracting than helpful imo. Or to put it another way, the set of people working in the ml field, is both larger and more diverse than the set of people working purely in math. Everyone in the population of people who can both program and understand this type of notation would also be able to understand the less terse but more readable utf8 based text, the inverse is not true.

Join the linux clan. It's the one true way.

[Disclosure work at Google] Unless i'm misreading this, the publisher of the html editor can easily replace links on some pretty large websites. Then it can also hack Google to remain a top html editor. This is a very clever virus!

Just to add a little data from personal experience... About 4 years back I ran a small study in grad school where I was trying to find some alternative to ads. For a survey I basically asked "how much would you be willing to pay to view websites without ads".

I bucketed the responses to something along the lines of greater than $1, $.75, $.5, $.25, $.01 or nothing.

What I was (secretly) hoping was that people would be willing to pay something like a cent for a page view, since I was building a prototype for a more user friendly anti-adblocker, seeing that many websites started to deploy their own back then, and hating that there was no middle ground.

However, what I found was that people were unwilling to pay anything for a page view. I think the author actually mentions this as a failed alternative model.

Point being it really seems like people have decided/agreed that they prefer the ad model to anything else. Even if it means that the cost of serving ads is actually higher (in terms of bandwidth) than what it would cost to just pay a cent or two cents or whatever for pageviews.

Ah yes, didn't think about the use case of unsupported codecs. Though at that point I would argue it would be easier to convert the file with ffmpeg, etc, then to use a custom decoder in the browser.

Why not use something like Webrtc? You’ll be able to use a native decoder + seek to the nearest key frame.

Its been a while since I've read something that entertaining. Thank you.