Mastodon Feed
Sometimes the safety thing you need to build is psychological safety for your engineers, but even then, "we should work fewer hours" is not the same as "we should slow down". And you should focus more on removing the structural reasons that crunch needs to happen in the first place, rather than telling people "just work less".
But also nobody currently talking about "slowing down" gives a single solitary shit about the mental health of their fellow human beings
Mastodon FeedDragging this one out from the archives to remind everyone that, in tech, "we should slow down" is the battle cry of the incompetent. Speeding up can be dangerous but "slowing down" just gets you more dysfunction, not safety. People who actually care about safety will describe what SPECIFIC SAFETY you should be building instead of whatever other "too fast" thing you are doing. https://blog.glyph.im/2023/12/safer-not-later.html
Mastodon FeedSo I dunno. I'll probably upgrade. But I'm the least happy about it that I've been in a *very* long time
Mastodon FeedI have at least one friend that has worked on this stuff at least peripherally (maybe more, given the way the company works) and I know that people who care a lot and want the right thing for users are involved in making it but still, nothing can contain the malevolent energies inside these cursed matrix multiplications
Mastodon FeedI am usually a day-1 upgrader for Apple OSes. My reasoning being that I'm not really in a consensual relationship with this company where I get to have real choices; upgrading sooner means I can get the benefits of the new release faster and adapt to the new indignities sooner.
I gotta say though the Apple Intelligence stuff is just … disgusting. I know it can be disabled through screen-time hacks, but it feels invasive. Like an Illithid tendril scooting just a bit further up my nose.
Mastodon FeedBoosted by glyph ("Glyph"):
katelynburns.com@bsky.brid.gy ("Katelyn Burns") wrote:
"You just have to convince people."
Dawg, there are a dozen conservative websites dedicated to spreading lies about trans people and billions in election spending.
I just had to croundfund an outpatient surgery and I'm one of the biggest trans writers of our time.
Mastodon FeedBoosted by glyph ("Glyph"):
mhoye@cosocial.ca wrote:
And between you and me, I think that if people are given a choice between "there's this minority that shouldn't have rights" and "there's this minority whose rights are ... enh, not that important really, when you think about it" ....
I can understand why principled people would choose "neither". In theory democracy is compromise, I get it. But if one party's position is "these people should not exist" and the others' compromise is "what if they were just stripped of rights or equality"...
Mastodon FeedBoosted by glyph ("Glyph"):
mysk ("Mysk🇨🇦🇩🇪") wrote:
RE: https://mastodon.social/@mysk/117264670021439841
Happy iOS 27 Day!
Bookmark this post if you want to disable Apple Intelligence. Apple buries the option in "Screen Time". Apple joins other Big Tech companies in making it tougher to opt out of AI features that send data off-device.
Mastodon FeedBoosted by glyph ("Glyph"):
aud@spectre.asta.lgbt (":: Asta :: [New Server]") wrote:
They're ignoring the actual science on this, which was pretty well settled, to gain a few right wing votes. Newsflash: they won't gain them, and Newsom signed bills that blocked *one high school girl*. He made a big fucking fuss all to keep ONE GIRL from competing. He ignored and went against actual scientific studies to block one high school trans girl from competing in a sport... and has platformed nazis and continues to campaign on this.
Fuck Newsom. He'll throw your ass to the ground, too, and won't do jack shit to stop fascism or bring ICE to justice. He will not. He may say otherwise, but he absolutely will not.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
we can think of another way but it's not as if Forth is meant to be a language in which internal state is inviolate
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
fun, haha, we wrote a little helper word called
nexit. ("short for n-exit", as we said in the documentation...)it directly overwrites the control stack so that we can force our caller to return immediately, which is handy when transforming existing code, but not a great idea otherwise. we intentionally avoided most forms of non-local control transfer up to now, but we need it. or, at least, it's probably fine.
Mastodon FeedBoosted by firepoet@tech.lgbt ("Stephen 🌈 (he/him)"):
psh@infosec.exchange ("Patrick.sh") wrote:
https://www.reuters.com/world/anthropic-ipo-launch-shifts-toward-mid-october-sources-say-2026-09-04/
"Anthropic is expected to begin marketing its initial public offering in mid-October at the earliest and complete the listing days before the U.S. midterm elections in November, people familiar with the matter said on Friday."
Do you think Anthropic is likely to stall its IPO listing due to pre-election volatility and SEC revisions -- especially now with the recent whistleblower from Anthropic?
(Perhaps this was...planned all along?...)
Mastodon Feedjsonstein@masto.deoan.org ("Jeff Sonstein") wrote:
‘Why were so many powerful people in America, a country whose civic culture is steeped in resistance to tyranny, so willing to abase themselves and their institutions?
That’s the question at the heart of Jacob Weisberg’s important new book, “Profiles in Cowardice: A Study of Collaboration in the Trump Era.” It’s an inverted version of John F. Kennedy’s “Profiles in Courage,” offering character studies of eight people who gave in to Trump despite surely knowing better. Reading it made me wonder if there’s something about America today that selects for spinelessness at the highest levels.’
https://www.nytimes.com/2026/09/14/opinion/institutions-power-trump.html?smid=nytcore-ios-share
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
none of which are wrong, but this sort of thing is why we hold ourselves to high standards......
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
we vaguely remember making the original decision to redo it, with such thoughts as "it's good enough for now", "we can fix it if it's a problem", and "it would be too twisty to do it the right way"
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
oh hey, hah, so we started working on the patch to label-transform, and while we were doing that we totally thought of a way we can make that performance issue way better without implementing a new data structure
because it's a pair of linear traversals that need to happen in tandem anyway, so we just need to avoid redoing the work in one of them
Mastodon FeedBoosted by kornel ("Kornel"):
rileytestut ("Riley Testut :fatpikachu:") wrote:
RE: https://fosstodon.org/@altstore/117271149448802444
With iOS 27, users in the EU, Japan, and Brazil can now sideload apps (.ipa files) entirely on-device without ever using a computer!!
Just install AltStore Classic with AltStore PAL, then follow our onboarding guide to easily get set up with Remote AltServers 👌
Mastodon Feedjsonstein@masto.deoan.org ("Jeff Sonstein") wrote:
ugly
Mastodon Feedjsonstein@masto.deoan.org ("Jeff Sonstein") wrote:
don’t like the news? just eliminate the statistics you do not like
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
it's a little hard to come up with a more granular test that would provide stronger evidence, we think we'll leave the testing there for now. we think we got what we wanted.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
notably we're still generating all the annotations on that run, so the generation of them can't be the cost, so that's really good.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
wow, okay, our second guess knocked the entire thing down to 30 seconds. we're hopeful that that confirms our guess that the problem is that the metadata annotations are stored in an array rather than a btree.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
interesting. that knocked 40 seconds off, which is a lot, but that's all. so much for our first guess at the cost... hmm
Mastodon Feedjsonstein@masto.deoan.org ("Jeff Sonstein") wrote:
at dinner in Bodega Bay after a lovely drive up the coast and a walk with a friend and her doggo in Jenner.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
we're trying a run now that skips a specific processing stage, which should take less than thirteen and a half minutes and we'll be interested to know how much
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
basically we want to do some, uh, decision support to know we're not boxing ourselves into a corner with a long compile time that there won't be a way to optimize
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
yeah thirteen and a half minutes
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
but the thing is that the work it's doing doesn't feel as if it should be that much, except for the part where it uses an array instead of a proper data structure for the metadata annotations. so we're going to comment out various parts of the code and see what that does to the runtime.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
we'll know how much more when this run finishes. we think about ten minutes but we've never written it down before.
Mastodon Feedireneista@irenes.space ("Irenes (many)") wrote:
a regular build of the compiler takes about half a second, which feels ludicrously high to us but it's probably fine. a full build of the hex dump version takes, uh, more than that.