bla3

How many people are there on windows arm? It doesn't seem super popular.

If anything, this looks like Google thinking that that might be changing.

Does anyone know if there are new windows arm products that might make it more popular?

I parsed this as (X Purges Prominent Journalists), (Leftists with No Explanation) at first and was pretty confused. If that should happen to anyone else, it's supposed to b (X Purges Prominent Journalists, Leftists) (with No Explanation).

PDF had an open spec and oodles of programs supporting it. I don't understand where this comment is coming from.

Which one should I read first?

Or the new Preferences.app.

Also following redirects by default. "Run curl, be confused run curl -L" is something I do many times a month.

I don't know if it's necessary to be competitive about eulogies, but the one you posted was also touching. Thank you for posting it. The picture of Bram's GitHub contribution graph is haunting.

Sounds like your old place had a good idea what's good about c++. It's not the STL and certainly not boost.

Works for me (heh).

Yes, but it is that way because these stories are what people want to buy movie tickets for.

Doesn't this flip all text in the SVG upside down?

The first significant digit of "0.1" and "0.2" _is_ 3! The first significant digit of "0.3" just isn't 3.

In case anyone hasn't seen it yet, if you print a double with enough fractional values, you'll eventually get its "true" value (because 2 divides 10).

  >>> print('%.60f' % 0.1)
  0.100000000000000005551115123125782702118158340454101562500000

It's only zeros after these zeros.

Here are the "true" values of 0.2, 0.3, and their sum:

  >>> print('%.60f' % 0.2)
  0.200000000000000011102230246251565404236316680908203125000000

  >>> print('%.60f' % 0.3)
  0.299999999999999988897769753748434595763683319091796875000000

  >>> print('%.60f' % (0.1 + 0.2))
  0.300000000000000044408920985006261616945266723632812500000000

Turns out 0.1 + 0.2 is actually the number right after "0.3":

  >>> print('%.60f' % math.nextafter(0.3, 100))
  0.300000000000000044408920985006261616945266723632812500000000

If you don't use a ton of decimal digits, by default the runtime prints the shortest decimal number that has the same binary representation as the true value when converting that decimal number to binary. Much of the confusion around floats is due to that convenience feature that aims to make floats less confusing.

Have you heard (intelligent!) non-tech people talking about tech? Why would (intelligent!) tech people fare better talking about non-tech?

Domain knowledge is more important than intelligence almost everywhere.

zeux posted this: https://gist.github.com/zeux/37f8f9dd7840ac0217ee862ba3a269f...

Also features like alpha support, yuv444 support in addition to yuv420, progressive rendering are interesting. Do jxl and avif both have all of these?

In NY at least, you can do into Fintech. That pays _better_ than Google/Meta.

A macro lens is defined as one having 1x magnification, meaning 1cm in real life at the focus plane is mapped to 1cm on the image sensor. A full frame sensor is 3.6cm across horizontally, so a 3.6cm object fills a photo. These things are smaller, so they need a super macro lens with 2x or even 5x magnification. You can get that by buying such a lens, using macro bellows, clipping a separate magnifying lens in front of your normal macro lens, etc.

In any case, with such a high magnification, the plane of focus becomes very narrow, so they're very likely also doing focus stacking.

And you need a lot of light, so they're probably also using a flash.

I only rent a small server from them, but I've been happy with their support. Talked to a real human who could help me with tech questions, even though I pay them next to nothing.

(No affiliation.)

In other words, "this approach is getting to expensive for us and we need to control the narrative for our model sizes going down".

Most increases in Imagenet etc scores came from bigger models. "The Unreasonable Effectiveness of Data" has aged very well. It seems very convenient for OpenAI that this trend should be over a few months after their "eye-wateringly expensive" comment.

Tangential, but it's kind of pleasing that Lars WirzeNius is writing an article for lwn.net.

Starts out insightful bit then pivots into

> If you rebel against some of the precepts of this new ideology by espousing a contrary view as James Demore did at Google when he offered an alternative explanation on female representation

Eh.

I think this is both obsolete and subtly wrong. Big corps want products with global impact. When a product is new, it's evaluated for its potential. When it's been around for a while, it's evaluated on how well its growth is matching the estimated impact. Eventually, growth has stopped. If the product lands at an active base of a billion users or so, it's a success and people definitely can make a career out of maintaining and improving it.

And to make working on new stuff a bit less incentivized, promotions to higher levels need to see at least some realized growth and not just getting v0 out these days as far as I know.

It's also true when learning languages. If you speak a language a tiny bit, locals act politely impressed. If you speak it kind of good but aren't a native speaker, at least in some countries that lands much worse than speaking the language badly.

Looks like Cat Allman was let go in the big layoffs: https://en.wikipedia.org/wiki/Cat_Allman

I'm guessing that's the reason for this. I wonder what this means for Summer of Code.

I'm neither credible nor reputable, but I'm in favor of a weak version of this. When reading older books to my young child, I replace language that suggests that women are supposed to stay at home and men are supposed to go to work. One day soon he'll be able to read, and at that point I'll wish that books that were a product of their time would have been updated for the current time, so that he could just read the book, without us having to have a conversation about how things were different then.

"who don't know how to change it"

Or who choose not to change it. Not everyone who has different preferences than you is incompetent.

FYI, https://www.ranchoelcharro.com/mountain-waterfall-horseback-... says "$129" at the top but "$139" at the bottom.

Do you know the difference between the T5 and the T7? From what I can tell, the T7 has faster transfer speeds, but is noticeably cheaper online. What's up with that?