rfoo

33 ms·

631.
by rfoo 4y ago
In my experience it still sometimes makes up schema or outright outputs mansplaining-like plaintext instead of JSON even if I give it correct schema. Happened once about 15~20 attempts, 5% is still too high to be considered reliable :( I&#x
632.
by rfoo 4y ago
The ability to reliably follow such instructions given a JSON manifest likely comes from training on a (small) dataset of these "plugin manifests".
633.
by rfoo 4y ago
Human rights activists are not tech-savvy.
634.
by rfoo 4y ago
That's right. I'm an okay-ish competitive programmer (around 2400 on Codeforces). I haven't played competitive programming for years (though being semi-active in the community, for problem setting etc) and last I tried I can
635.
by rfoo 4y ago
Reading errata for Intel/AMD processors would be an eye opening experience then. There are so many bugs in features you never heard of, just like software. It is inevitable given the complexity.
636.
by rfoo 4y ago
Quite a lot in the old days. Microsoft is a well-known brand in China. Grandma and grandpas are repeating stories of Bill Gates during 00s, they don't know who tf Sergey is. There are very minimal Mac users, so computer = PC = Windows
637.
by rfoo 4y ago
HN Guidelines [1] require using original title, unless the original is misleading etc. It does not matter what's the problem with this title, even if you make up a perfectly fine (and better in any aspect) title, it's still more a
638.
by rfoo 4y ago
I'm pretty sure CAC does not care, as long as they receive a copy of the exploit code.
639.
by rfoo 4y ago
Go out of your way to use stock Android on Pixel devices or GrapheneOS instead. At least if somebody decided to burn their exploit you got a chance to profit over it. iPhone might also work but it's too hard to do forensics.
640.
by rfoo 4y ago
> the thing I find disingenuous is that it came out of fucking no where 3 years later That's how these things happen, at least to the original SARS. Okay, besides killing all masked palm civets for "idk maybe" about 3 mont
641.
by rfoo 4y ago
> written in Rust Thank you, I'm glad to have a more casually hackable implementation in TypeScript. It definitely helps that you project is not written in Rust!
642.
by rfoo 4y ago
For hard links to work you still need to know that the brand new layer you just downloaded is same as something you already have, i.e. running a deduplication step. How? Well, the most simple way is compute the digest of the content and loo
643.
by rfoo 4y ago
> do massive refactoring, validate it using tests Google do have damn good mass code change tools, for human, without AI.
644.
by rfoo 4y ago
Sorry, I didn't say it clearly. I want to use the distro (Alpine) packaged Chromium, which links to musl. Not some random Chromium build I found on the Internet. Having to use a glibc Chromium is already too far for me. Do you suggest
645.
by rfoo 4y ago
There can be only one ld.so in each process. Widevine is a plugin-like .so meant to be loaded into Chrome/Chromium. Because it uses glibc, the entire process hosting it must use glibc. So, what prevents me from ALSO HAVING GLIBC CHROMI
646.
by rfoo 4y ago
When people say glibc, they really mean ld.so + libc.so + libm.so + nss + ... And no, glibc really only work with its own dynamic loader (ld.so), and it has to be with same version.
647.
by rfoo 4y ago
Netflix already labels them. They show only their original content to users which are suspected to be behind VPNs. [1] [1] https://www.whats-on-netflix.com/news/why-can-i-only-see-net...
648.
by rfoo 4y ago
> So, they do store requests for "debugging" purposes No, that's likely entirely made up by ChatGPT.
649.
by rfoo 4y ago
Converting to int8 halves the size.
650.
by rfoo 4y ago
GP could be mentioning quantization aware training, during which the weight and gradient are still computed in fp16/fp32.
651.
by rfoo 4y ago
It would be interesting to see if Identical Code Folding (ICF) can remove the dupe automatically.
652.
by rfoo 4y ago
The more important difference is 58.6mm vs 64.2mm width. 5.6 mm is a lot. I can't hold phones with >68mm width reliably and drop them constantly. I just wish vendors can make small (or long, like Sony) phones.
653.
by rfoo 4y ago
Overall, cool work! Especially the search space reduction part. I haven't dive into the code, but only 818000/60=13633 attempts per second for 64 SHA-256 rounds plus one AES-256 decryption on a 2080 doesn't sound right. Learn
654.
by rfoo 4y ago
They have been just another corporation for a while. The great layoff and such things just made it official.
655.
by rfoo 4y ago
> but what is the source with actual benchmarks and power measurements? Every notebookcheck.net review. For example https://www.notebookcheck.net/AMD-Ryzen-7-6800U-Efficiency-R... They also do the same to a lot more lapt
656.
by rfoo 4y ago
> poor containerization support, an annoying development toolchain Just run an always-on (headless!) Linux VM in the background, and don't use the host macOS for anything besides desktop apps (Slack, VSCode, Browser, mpv, terminal e
657.
by rfoo 4y ago
Fair point. I treated the product as a build system and it just doesn't look right. However it makes perfect sense if it advertised itself as a "task runner".
658.
by rfoo 4y ago
I see that you are targetting the Web ecosystem. IMO using Bazel in this case was pretty tough, and may still be. The way Bazel does things may not make sense for 95% JS projects (and 99.9% FOSS projects), so I wonder whether it's bett
659.
by rfoo 4y ago
Different OS, different storage, different filesystem are all fair as you can say you are testing the laptop as a whole, so what was preconfigured matters. Also generating totally different code (compiler outputting x86_64 vs aarch64) is fa
660.
by rfoo 4y ago
It's not about caching. Beam searching through a search space with no obvious "answers" is naturally hard.

More ›