cloudkj

Can you share some details about the companies that these offers came from? For salary data like this, I feel that if a handful of specific companies are not included in the data then it can easily give off a skewed perspective since many companies have different compensation structures.

Whenever one of these salary posts/articles/studies crops up, there's inevitably some caveat: data bias, small sample size, etc. In this case, although the article is using a (hopefully) representative sample based on offers, it's still based on base salary only, which makes it very difficult to make comparisons.

I think the most accurate salary data that can be considered for comparison would be taxable wages reported on a W-2, for software engineers with different titles at different companies. The wages would be much more representative of the total compensation package an engineer gets, as it includes salary, taxable equity, bonuses, and other benefits such as health care reimbursement, relocation, etc. Since the amount is for taxable wages, I think it would also more realistically account for vesting and other special situations such as deferred compensation. Provided that the data had a large enough sample, it would include a nice cross-section of employees at different parts of their vesting lifecycle with different types of equity compensation, and thus the numbers in aggregate would be useful for comparison.

Practically, the only entity I can see doing that type of study would be the IRS, or perhaps Turbo Tax or some other tax software vendor that adds some sort of anonymous wage collection option for users when filing. That actually sounds pretty creepy and intrusive though, so hopefully nobody at Intuit or the IRS is reading this and gets any ideas!

This is a neat idea. You could have a video game like NBA 2k act as the frontend for the data coming out of stats.nba.com backend, and simulate - in realtime - the actual games as they progress. If the latency between the frontend and the backend is reduced enough, you'll effectively have a "live" game played by simulated players that is more or less a replacement for the real thing. You could then charge people to watch your simulated games.

I participated in high school policy debate for all four years of high school, and can relate to some of the criticism in the article. However, I find the parallels drawn by the author between debate and today's political atmosphere tenuous at best, and perhaps exaggerated. To me, the biggest educational value of high school policy debate was its blank slate nature and its openness to different evaluation paradigms.

By default, the convention was to apply a "policy maker" paradigm, in which the advantages of the plan affirming the resolution are weighed against the disadvantages brought up by the negative team, and the side with the most "weight" wins. Thus, it's natural that the arguments would eventually converge upon the scenarios with the most weight, i.e. those with the highest death tolls or biggest quantifiable impact on humanity. Nuclear war and climate change were argumentative staples in these types of debate rounds, and I even recall some cases where a policy maker judge, in his oral feedback, literally added up scenarios for each side to determine the winner.

Another convention is "silence means consent", which likely led to the practice of spreading to overwhelm your opponent with arguments so that any unanswered argument automatically gets granted as won. I feel like that practice is more an artifact of the way debates are structured with time limits and restrictions on only considering arguments that were continuously extended throughout the speeches in order to be considered as a valid "reason for decision". Coupled with a desired to quantify each argument in order to outweigh the opponent, I think this behavior is inevitable given the structure of the activity.

However, I feel that the open nature of the evaluation paradigm was a shining educational tool. Given the right judge (e.g. a debate coach, an experienced former debater, or a college debater), you had the option of advocating for any type of evaluation paradigm you wanted. You could open the debate up to meta topics such as whether the affirmative plan was "on topic" by nitpicking the definition of words in the resolution or arguing grammatical technicalities, and claiming that "non-topical" plans were damaging to debate as an activity and warrant a loss; you could attempt to compare the morality of certain scenarios against quantifiable death-toll scenarios, such as advocating that the moral obligation for overturning some court case to symbolize the eradication of racism against some group might outweigh any type of nuclear war scenario; you could even make arguments against the practice of "spreading", in the name of education, and demand that if the opposing team speaks too fast that would damage the educational value of debate and thus warrant a loss.

All and any of these arguments would've been valid, and I feel that provided valuable lessons in thinking outside the box, looking for technicalities, and optimizing approaches for playing the game.

Don't forget about the personalization and recommendations technology, which accounts for 75% of viewing. https://medium.com/netflix-techblog/netflix-recommendations-...

Thanks for this. Can you also add a way to change the window for S&P returns from 10 years to other time windows? It'd be interesting to see how the plot changes with adjustments to the window.

I've also worked with both Perl and Clojure professionally. I think Clojure code, by nature, has a higher chance at being "beautiful" or elegant, or at the very least as a close proxy, minimal. I feel that doing a lot with a little is easier in Clojure, and it makes these "beautiful" code snippets less likely to be mysterious and more likely to be understood by maintainers down the line, even sans documentation.

The non-author comment is well taken though, as I find Clojure development to have a much higher thinking-to-coding ratio, so when a solution gets translated from the author's head to code, much of the complexity can be hidden due to the conciseness and lack of verbosity. I didn't really find that to be a problem in Perl, however, with the exception of perhaps the (over/ab)use of complicated regular expressions.

Are there any Niantic employees active here on HN? As an engineer, I would be very interested in hearing some first hand accounts of some of the technical problems Niantic's engineers have had to deal with, not necessarily just for the Go Fest, but for the game in general. To the extent that they're allowed to talk about it, it would be very interesting to get an AMA style take on some of the challenges.

Does this hack work on Google accounts? I just tried the "forgot password" feature there and as far as I can tell there's no way to actually complete a password reset with only a compromised phone number.

Having just served on a jury, I can say that I've actually gained some confidence in the American justice system and its use of a jury of peers. When it comes down to it, the task for the jury is fairly straightforward, narrow, and well-defined: to examine the facts as laid out in the trial and to objectively execute the law as it is defined at the time while being fair and impartial. Their backgrounds and prior knowledge and experience are not supposed to come into play, and to the extent that the judge and attorneys have chosen a panel of objective, intelligent jurors, their backgrounds really do not and should not matter.

I recently read a comment that framed the law as being software. After my experience, I've come to think of juries as being the computers that run such "software" in the context of a trial: to execute the instructions as handed down by the judge, given the evidence and testimonies as factual inputs and to ultimately output a verdict that follows (as closely as possible) the rules codified in the law.

I feel that the problem really lies in the evidence provided by the expert witness. In this particular case, it sounds like there was an expert witness on both sides, but perhaps one witness was more effective than the other, at least to the jury. To reduce any bias, ideally there are (many) multiple witnesses providing testimony on the same issues, so that the evidence presented will average out to be as close to the factual truth as possible. Unfortunately, that's almost certainly not scalable in the context of a trial.

That's a very interesting point about diversifying your tax liability. Is it a fairly common tactic to contribute evenly across pre-tax and post-tax buckets? I've mostly stuck to contributing to one bucket but spreading out contributions across buckets makes a lot of sense. If you have any sources about the pros/cons of diversifying tax liability in this way, I'd be very interested.

Interesting side note: the author also happens to be an offensive lineman in the NFL. http://en.wikipedia.org/wiki/John_Urschel

This article was surprisingly well written. I think the point that most resonated with me was the framing of the search for better (happier) conditions in each of the "major areas" as an optimization problem.

The probability that one's current conditions are anywhere near optimal is low. Likewise, the probability that you can improve your overall conditions by making a change in one of the areas is high, since it's not a completely random change but one that you control.

A sort of real life analog of gradient descent (ascent) or hill climbing, if you will.

I just realized that this feature also exists in Outlook, which explains why many of my replies have strange quotes. I often highlight a user id or URL to copy for further investigation, then hit reply only to be confused by the condensed quote in the reply window. Now it makes sense :)

Well put; I share the same sentiment. I feel that one of the strongest arguments for traditional characters is that it retains the original meanings behind the characters themselves. Simplifying the characters just throws information away. It's a lossy compression!

That's an interesting idea. Is the main motivation of your suggestion to find a way to enable you to better discover things you might like without relying purely on reviews and ratings? In general, I agree that ratings and reviews have inherent biases that are hard to smooth out. There are so many more dimensions to it, such as how much weight you apply to a review written by someone that is trying a certain type of food for the first time.

Do you think that you would see better results if Yelp was able to use a collaborative filtering approach to somehow personalize their results such that they more closely match your tastes and preferences - based on other users' data?

Similarly, the stable marriage problem provides a mathematical basis for encouraging hackers to put themselves out there and take initiative in seeking out potential mates as the Gale-Shapley algorithm proves that the end result is optimal for the suitors, i.e. those that take action.

And along those same lines, adds more weight to the implications of the Great Filter argument. That is, if the early steps in the path to intelligent life are really not that unlikely, then the later steps to advanced intelligent life are more unlikely.

Nice summary; I like the format as well. However, the title of the cheat sheet is misleading since (a) many of the algorithms listed can be used for non-linear classification and (b) some of them can be considered supervised learning, such as naive Bayes and perceptron since they're trained with sample inputs and expected outputs (supervisory signals).

Otherwise, this is awesome. Hopefully you will add to it, and make it available in web form.

Would you recommend The Little Schemer as a book for teaching programming/computation concepts to kids?

I was just looking for a simple SSH client for the iPhone today, and was disappointed that all the apps cost a few pretty pennies. Any recommendations on the SSH client(s) that will get me the best bang for my buck?

EDIT: I caved and just bought Prompt. It is indeed pretty slick, though I don't have other iOS SSH clients to compare it to. There does appear to be some bugs: I can't seem to "disconnect" from a session, even though it's already hung. The little "eject" icon in the servers list doesn't seem to work.

At Berkeley, the BA and BS for Computer Science are effectively the same. All the core courses are the same, it's just the general requirements are different. The degree designation really doesn't matter. If I'm not mistaken, I believe Harvard also grants Bachelor of Arts degrees (they use AB instead of BA) for Computer Science.

Like. I do these type of time-saving calculations in my head whenever someone like the bus driver drives too slow and fails to make a light, thus costing everyone on the bus an extra minute of their lives.

You actually saved 24 minutes times, say, 3000 students, which is 24*3000/60/24 = 50 human days saved over the span of four years. Not to mention, this hack may well have lived on since you've graduated, so that's many more days of savings. Nice.

Pretty much the entire Pacific Rim...

Have there been applicants that were accepted but ended up turning it down for whatever reason (job, move, visa, etc)? Have any of these reapplied and participated in the subsequent go?

I think you're taking quite an extreme interpretation of Graham's/Buffet's thoughts on index investing. It's a given that, with index investing, your returns will match the market. Index investing isn't aimed at getting higher returns. Rather, Graham actually spends more time discussing and differentiating between different classes of investors.

For the "enterprising investor", s/he can afford to spend more time to find undervalued stocks and possibly outperform the market. For the "passive investor", to paraphrase Graham, it is good enough to simply invest in a representative set of securities, such as an index.

In fact, if I remember my Intelligent Investor correctly, there are various passages where Graham actually admits to the EMH being somewhat inevitable. Note that there are different variants/degrees of the EMH, and Graham probably subscribes to the weak or semi-strong versions.

Nonetheless, the point is that a belief in the EMH is not mutually exclusive with value investing. It depends a little on the extent to which you believe the market is efficient, and a lot on how much work you're willing to put into it.

Levenshtein/edit distance isn't too useful for spell checking though. Edit distance can yield you the absolute difference between two words, but your goal is to find the words that have the smallest such distance with the word in question. Simply calculating the edit distance between the word in question and all other dictionary words isn't good enough - you'll have to be a bit more clever.

HTML5 will bridge the gap. Here's a Netflix job posting: http://jobs.netflix.com/DetailFlix.asp?jobid=flix4467

I haven't really been living off of my apps, but since I've been unemployed for the past 6 months with no other streams of income, I'll count them.

http://www.facebook.com/amznwishlist was pretty profitable up until about mid-October when Amazon decided to disable some of its Product Advertising API calls for getting Wish List data. If the app still worked and ran through the holidays, it probably would've paid for rent (and then some).

http://www.fatearthmedia.com/ - My browser extension for shopping sites is also profitable. If the Mozilla add-on policy was less strict about affiliates earnings, then it'd probably be paying for rent as well.

I've been looking for a service that does this exact thing. Give me personalized recommendations on restaurants so I can stop trying to comb through mountains of sometimes non-sensical, irrelevant Yelp reviews to decide what new restaurant to try.

A lot of comments here have complained that this has been tried and done. My question is, where? Does Yelp already do something like this? I went on Yelp a few days ago looking for this exact service. I'm not a Yelp user, and I just wanted something quick and lightweight that would give me recommendations based on my preferences. Found nothing to that effect. In order to contribute at all, I have to write a full review for a bunch of restaurants.