viraptor

6 ms·

61.
by viraptor 8mo ago
Apple is going to Apple. They're as likely to remove the deprecation as they are to restrict it to first party apps only. They don't care about devs.
62.
by viraptor 8mo ago
We're taking about both now, at least one a week it seems. Without the DDoS, we'd mostly forget about the blog. I didn't even know about the blog until the DDoS started.
63.
by viraptor 8mo ago
You normally need to do that anyway. The specific backend host may have been destroyed in the meantime so you have to recover the context. And it's not like they're huge after compression.
64.
by viraptor 8mo ago
This is an impressively unhinged take. I still have no idea what the person is trying to achieve. And I'm sad we're likely going to lose that resource in the future.
65.
by viraptor 8mo ago
It's cool workaround for the problem that preserves the problem. An alternative is to write the query in a more reasonable language like https://prql-lang.org/ which has a representation closer to the semantic meaning
66.
by viraptor 8mo ago
https://bandit.readthedocs.io/en/latest/ can do that for python.
67.
by viraptor 8mo ago
There are a lot of problems solved by tiny models. The huge ones are fun for large programming tasks, exploration, analysis, etc. but there's a massive amount of processing <10GB happening every day. Including on portable devices. T
68.
by viraptor 8mo ago
This is another "in Europe" thing. There's no "in Europe". Germany, Italy, Poland, Portugal, etc. will all have different rules.
69.
by viraptor 8mo ago
Not having bound checks does not make sqlite sqlite. If that was the case, you couldn't compile it with https://clang.llvm.org/docs/BoundsSafety.html turned on and still call it sqlite for example.
70.
by viraptor 8mo ago
You'll need permissions to mount things. It may be restricted in CI environment, so I wouldn't rely on it.
71.
by viraptor 8mo ago
It matters less these days with the thinking models. They'll automatically inject some extra content before the answer for basically the same purpose. But if you're using something simpler with immediate responses - yeah, the orde
72.
by viraptor 8mo ago
Very early, sure. But the later common speeds that most people experienced were between 14 and 56kbps. Preloading was definitely possible even with just image tag ordering.
73.
by viraptor 8mo ago
This is coming to GH soonish. Some clunky alpha version of this UI has been shared on the bad social site. (First class rebasing / stacked PRs)
74.
by viraptor 8mo ago
That's not what preload is for. You don't wait for the extra images on the first page. You start loading them after the page is complete, so that the next page loads faster.
75.
by viraptor 8mo ago
They seem all over the place. Instead of specific colours, you could go around the hue circle without touching saturation: https://developer.mozilla.org/en-US/docs/Web/CSS/Reference/V...
76.
by viraptor 8mo ago
It's nice that this reminder surfaces from time to time. Hopefully there are a few lucky 1/10000ths today. I'd just add some nuance to: > If you cannot reach the untested lines with the public interface of your software, m
77.
by viraptor 8mo ago
It would be actually fun to see where the limit is on echolocation with serious ML processing these days. Apparently people did quite well in 2022 https://pmc.ncbi.nlm.nih.gov/articles/PMC9655721/
78.
by viraptor 8mo ago
> Because the format can't also be strictly defined via structured output, and you have to write it in plain words. That's not how structured output works. Check the docs https://platform.claude.com/docs/en
79.
by viraptor 8mo ago
Nor will they be restricted by the LLVM design. That project is huge and generic trying to be everything and target everything (and takes ages to rebuild if you need some changes). Sometimes it's better to go simple and targeted - time
80.
by viraptor 8mo ago
I do that because there's no point in anyone seeing the pre-release versions of my projects. They're a random mess that changed the architecture 3 times. Looking at that would not give anyone useful information about the actual ap
81.
by viraptor 8mo ago
People still try to use @user all the time, even though it doesn't work.
82.
by viraptor 8mo ago
Looks like you're mixing up two things when testing: the correct answer and format following. If you want both, why not use https://platform.claude.com/docs/en/build-with-claude/struct... ? If you don&#x
83.
by viraptor 8mo ago
I know it's on the sqlite side. I'm familiar with the claim and disagree with it.
84.
by viraptor 8mo ago
While I'm not annoyed by the slowdown that much, what made me not trust them anymore is being careless with the system. For example I did a review recently on a PR where the collapsing sections were not visible and made 2 patch fragmen
85.
by viraptor 8mo ago
You can also save time/tokens if you see that every request starts looking for the same information. You can front-load it.
86.
by viraptor 8mo ago
Auto is not likely to choose the high quality models unless you really try for complex plans. Give the explicit models a try instead. It really makes a difference.
87.
by viraptor 8mo ago
Original post: https://www.linkedin.com/posts/thdxr_everyones-talking-about...
88.
by viraptor 8mo ago
> Are you saying, for example, that generics are not useful? That's quite the claim. That's why I didn't make that claim. If you just want to invent straw men, I'm out.
89.
by viraptor 8mo ago
At the code level it's still rehashing the same ideas over and over again. I wrote lots of things from software 3d on a weird system to jit to websites to telephony software to compilers to firmware for hardware to cloud orchestration
90.
by viraptor 8mo ago
> the expressive type system of Haskell > Expressive type system ARE direct and simple. Those result in a conflict because given expressive types, people will make them not simple. For example, you realise why Haskell continuously get

More ›