cbr

If you're visiting a website that runs PageSpeed and you want to see the original markup, add ?PageSpeed=off to the url.

For example, compare http://www.jefftk.com/index to http://www.jefftk.com/index?PageSpeed=off

See https://news.ycombinator.com/user?id=michaelochurch

And specifically https://news.ycombinator.com/item?id=10019003

Are you sure you would pull your money out if all the banks started charging -0.1% annual interest? If you have your money in your house there's a higher risk of theft, more than enough to counter the -0.1%. Add in the inconvenience of moving large amounts of money around by hand, not being able to write checks, and no automated bill pay, and I'm still keeping my money in the bank.

I'd compare Google's ads to consumer electronics as a category as opposed to the iPhone: they're many eggs, not just one. Search ads, DoubleClick, and AdSense are like the iPhone, iPad, and MacBook.

(Disclaimer: I work for Google, on open-source software that doesn't have anything to do with ads.)

> a minimum hourly rate of about five dollars after tax

US federal minimum wage is $7.25 (and higher in many states), with income tax being functionally negative at low incomes. You could adjust downward by 10% to account for the five weeks of paid leave but that still gets you $6.50/hr.

    We’d like issues to gain a first-class voting system,
    and for content-less comments like “+1” or “:+1:” or
    “me too” to trigger a warning and instructions on how
    to use the voting mechanism.

Why bother users with a warning? Turn it into a vote, and then highlight the vote icon so you can see what happened.

You can automate this with the prioritize_critical_css filter in the PageSpeed module: https://developers.google.com/speed/pagespeed/module/filter-...

(Disclaimer: this is what I work on.)

    With a big enough N, the probability of reaching
    significance approaches 1.

Well, not with an A vs A experiment. Then the probability of significance at the p < 0.05 level should stay 5%.

On the other hand, in an A vs B experiment where B is very slightly different from A you're right that with a large enough population you become very likely to reach a p < 0.05 significance level, just with an extremely tiny effect size. The experiment has told you something about the difference between A and B, and you now have high certainty, but it's just such a small difference that you probably don't care.

    Have you ever approached a tool in the field that
    would fail my proposed test?

Your proposed test being, run an A vs A experiment and a tool is faulty if it concludes there's a difference significant at the 0.05 level more than 5% of the time? All the tools I've used for A/B testing would do fine on this test. This includes a custom tool at a computational advertising startup I worked at, Google's internal experiment platform, and Google Analytics' Content Experiments.

(If I didn't trust a tool to pass this test I'd just use it to get the counts for both sides of the experiment and do that stats on my own.)

    statistical significance is meaningless when split
    testing websites (you are guaranteed to archive
    significance if you run the test long enough

Wait, what? Not if you use the right test to evaluate your change. This article isn't showing that A/B testing is bunk, it's showing that Optimizely is using an inappropriate test.

    Do an a/b test with identical versions of you site.
    Spoiler: With 100% certainty, you'll see both significant
    results and a sizable effect size difference.

The whole point of significance testing is that, done properly, it tells you how likely you would be to get these results by chance if the two sides of the test were actually identical. Running your A vs A experiment and evaluating it correctly you'll find results significant at the p < N rate N% of the time. And the longer you run it the lower your estimate of the effect size will be.

> non-stone wall

Drywall is made of stone.

They can pick up a lot of signals, and they have claims data: it seems like they should be able to build a pretty good model.

Even just "number of miles driven" is a strong signal.

Are the "mini" versions of the flagships (64mm) too big? http://www.gsmarena.com/samsung_galaxy_s5_mini-6252.php http://www.gsmarena.com/htc_one_mini_2-6345.php

It's as if we were successfully making cancer less deadly, but at the same time more people were living long enough to get cancer that there were more cancer deaths. We'd still be making the kind of progress we care about in working against cancer fatalities, even though there are more of them. And a suggestion that reduce longevity so that fewer people would die of cancer would be missing the point.

They're smaller, but not that much smaller. With a combined population of 56M to the USA's 319M, we could scale their shooting numbers by 5.7x to get something adjusted for population. So if they were as big as the USA they would have ~17 shootings to our 984.

> "covering their own ass"

I think you and jessaustin may be using different interpretations of that term, judging by their reference of the McDonald shooting. People are much more ok with police officers doing what they need to do to keep themselves and others safe at the time (your usage) than they are with officers lying about why they used deadly force after the fact to keep from getting in trouble (their usage).

https://en.wikipedia.org/wiki/Shooting_of_Laquan_McDonald

The first graph shows that we have made driving far less dangerous over time. The second graph shows that driving has become far more popular over time, popular enough to outweigh this effect until the early 70s.

The recent improvement in car accident rates is pretty impressive.

Often these aren't really background images, they're just using that to set image sources via CSS. These days you can do the same thing with srcset, which also allows someone to learn your screen resolution.

Trying to match her linkedin profile [1] to the article, it sounds like she worked in reinsurance from 1980-2003 (age 23-46), took time off to care for kids, and started substitute teaching a few days a week in 2007 (age 50). Not working within your field for the last 13 years does seem like it would have a large effect on experience; this is getting close to your "whatever experience another person might have had 20 years ago is most likely either forgotten or completely irrelevant by now".

[1] https://www.linkedin.com/in/lynn-colafrancesco-5800163

Expanded this into a post: http://www.jefftk.com/p/detecting-tanks

Actually, there's also the 1992 paper "What Artificial Experts Can and Cannot Do" [1] which gives this anecdote and calls it a "legend".

[1] http://www.jefftk.com/dreyfus92.pdf

> ended up learning which days were sunny and which were overcast.

The earliest source I can find for this is https://neil.fraser.name/writing/tank/ but it says it "might be apocryphal".

Most people I follow on RSS include the full content of posts in their RSS feeds. When someone doesn't, it's generally an oversight and they fix it when I bring it up.

Anecdote: I've been at one of those companies for three years, and made about $250k this year.

"the only product that has ever been advertised on Google’s front page"

The Nexus One and 7 were too:

* http://searchengineland.com/surprise-googles-home-page-promo...

* http://www.cnet.com/news/googles-nexus-7-tablet-pops-up-in-r...

"One possible reason for this is the availability of salary alternatives as part of an overall compensation package. For example, companies with less than 200 people are usually in the series b or below stage, and can offer more equity, whereas many companies with more than 500 people are publicly traded and can offer stock packages. Those in the middle have less equity to give and typically aren’t publicly traded, so higher salaries may be a way to keep offers competitive"

That's a pretty big caveat, and is a big problem with their data. I'm at a big publicly traded company and my taxable comp is 60/30/10 between salary, stock, and bonus. Comparing just the salary portion with another company where comp is mostly salary wouldn't be very helpful.

I wish people talked more about total comp and less about salary.

Sorry, no:

    Chrome does not perform pin validation when the
    certificate chain chains up to a private trust anchor.
    A key result of this policy is that private trust
    anchors can be used to proxy (or MITM) connections,
    even to pinned sites.

-- https://www.chromium.org/Home/chromium-security/security-faq...

What's wrong with importing power?

And even more so by Levy and Young's 2004 paper: http://www.biu.ac.il/soc/ec/d_levy/publications/jmcb4.pdf

They were willing to lobby for a 7.5 cent coin and update their machines to recognize that, so it sounds like they could have handled dimes. The reason they didn't want to switch to dimes is that it was too big of a price jump.

    [they] briefly implemented a strategy where
    one in every nine vending machine bottles
    was empty ... this effectively raised the
    price to 5.625 cents.

Instead, charge $0.10 but 44% of the time you get your dime back along with the bottle. Effective price is the same 5.6 cents, but people are excited to get free bottles instead of mad and disappointed.

(This might be illegal, since, it's a lot like gambling, but both versions should be equally illegal.)