mgraczyk

27 ms·

481.
by mgraczyk 3y ago
> Jpegli can be encoded with 10+ bits per component. How are the extra bits encoded? Is this the JPEG_R/"Ultra HDR" format, or has Google come up with yet another metadata solution? Something else altogether? Ultra HDR: h
482.
by mgraczyk 3y ago
The content isn't worth nearly enough to cover costs, even if they could somehow prevent scraping. Look around at how much people pay for text data, it's not a lot.
483.
by mgraczyk 3y ago
They said that nothing related to safety or capabilities affected the decision
484.
by mgraczyk 3y ago
Hard to imagine a framework for thinking about this where "hand-dug" matters, but unfortunately I think it comes down to things like this.
485.
by mgraczyk 3y ago
My code wasn't broken, I had a tsconfig file that was formatted by prettier and prettier removed trailing commas. Then I updated prettier without reformatting and my tests failed because prettier rejected the now-incorrectly formatted
486.
by mgraczyk 3y ago
Normally patch updates do not require you to fix your code but often include desirable security fixes. That is the purpose of semantic versioning.
487.
by mgraczyk 3y ago
It was a breaking change, lots of people had to change their broken code, including me
488.
by mgraczyk 3y ago
https://github.com/prettier/prettier/issues/15942#issuecomme... I sort of misspoke, the comment itself is not silly, just the fact that it was that point that got a resolution.
489.
by mgraczyk 3y ago
Interestingly, prettier just made a breaking change in a patch release and refused to undo it for a week or so, until a particularly silly pedantic argument won them over. https://github.com/prettier/prettier/issue
490.
by mgraczyk 3y ago
Looks like the URL is wrong
491.
by mgraczyk 3y ago
I find {} much more readable then dict() because the former is more common. On the other hand list(L) more readable than L[:] because the fact that slicing copies is more obscure, and these are not equivalent when L is not a list. My take w
492.
by mgraczyk 3y ago
Probably not, because neither of them have complex branching strategies
493.
by mgraczyk 3y ago
Yes for sure. In some sense, I learned more from these holdouts than from anything else I've ever done in my career.
494.
by mgraczyk 3y ago
Google and Facebook both use trunk based development for nearly all code at both companies. This guide advises you to consider trunk based development if your release cycles are less than ~4 weeks. I'd suggest doing it no matter what.
495.
by mgraczyk 3y ago
A nitpick, but for your latter point it depends on the question. In some cases it's the intent that matters, but the law often tries to avoid questions about actual intent because that's too difficult to figure out. Instead the qu
496.
by mgraczyk 3y ago
Fortunately the legal system is flexible enough to accommodate questions like this, both with interpretations of the law and with juries to determine the facts. This isn't a new problem caused by emoji. You could make a similar argumen
497.
by mgraczyk 3y ago
If you don't use Facebook regularly, the feed will be bad. They use almost no data older than 90 days
498.
by mgraczyk 3y ago
Yes it would be easy, but when I worked there it didn't exist.
499.
by mgraczyk 3y ago
People have the option, you can choose to use a chronological feed on Facebook and Meta
500.
by mgraczyk 3y ago
That's bad too
501.
by mgraczyk 3y ago
Yes that's definitely possible. It's the sort of thing that Meta has teams to try to avoid, but in fast moving information landscapes it's sometimes difficult to keep up.
502.
by mgraczyk 3y ago
Most people on Meta's product follow too many accounts to see everything everyone posts. In an alternate reality where ranking is illegal, the game would be different and worse. Accounts would just repost the same content over and over
503.
by mgraczyk 3y ago
We did, they do. When I was there people did "calibration studies" for certain sensitive categories to understand and fix any model biases that would rank content higher for undesirable or unexpected reasons.
504.
by mgraczyk 3y ago
It's difficult for anyone who doesn't work there to know, but I would guess based on what I know of how ranking decisions are made that you are mistaken. For example I worked there prior to the 2020 election and on Instagram there
505.
by mgraczyk 3y ago
Well no, the rule is just not what you think it is. Meta doesn't ban supporting tools that can be used for violence, Meta bans support for violent acts.
506.
by mgraczyk 3y ago
From my first hand experience designing these algorithms at Meta and running experiments, and listening to feedback from accounts, I would guess that you are not getting reach because your followers do not interact with what you post. In al
507.
by mgraczyk 3y ago
That's Twitter, not owned by Meta
508.
by mgraczyk 3y ago
Typically less, depends on the surface.
509.
by mgraczyk 3y ago
We ran lots of experiments, including asking people survey questions, monitoring usage, asking people offline to try both, etc You're wrong about the design of algorithmic feeds. I worked on them, that's not how we designed them.
510.
by mgraczyk 3y ago
We ran experiments that lasted up to 5 years. People who have ranked feeds self report a better experience after using the product for years, although you have to be careful with these sorts of things because the chronological users are mor

More ›