krackers

39 ms·

841.
by krackers 2y ago
If you just need C instead of C++ then .m suffices, ojective-c is a strict superset of C.
842.
by krackers 2y ago
That is very cool, I think it's `-rewrite-objc`. Presumably it still needs to be linked to the objc runtime. It says it lowers to C++ instead of C though, now I wonder what pieces cannot be implemented using pure c.
843.
by krackers 2y ago
Yes, it's basically writing a nibless objective-c application but manually translating all invocations to objc_msgsend yourself. You're not going to be able to avoid the objective-c runtime though.
844.
by krackers 2y ago
Just saying "draw a black box" is not sufficient, you need to know the implementation details. If the software saves in a layer-based format, that's no good. If there is alpha channel then it's no good. If there is pre-e
845.
by krackers 2y ago
As a toy example if you have pixels a,b,c and you blur to (a+b)/2,(b+c)/2, you can recover back c - a. You might be able to have a good estimate of the boundary condition ("a"), e.g. you can just use (a+b)/2 as an a
846.
by krackers 2y ago
If you just want some intuition, I found this previous HN submission https://jiha-kim.github.io/posts/introduction-to-stochastic-... pretty approachable at giving you some key ideas without being too rigorous. It'
847.
by krackers 2y ago
But we have both 4o and 4.1 for non-reasoning. And it's still not clear to me which is better (the comparison on their page was from an older version of 4o).
848.
by krackers 2y ago
Might this be due to the fact (or at least something I've heard) that TLC SSDs maintain some initial portion as SLC for speed reasons? Also I'm not sure simply keeping the drive powered on is sufficient to prevent degradation you&
849.
by krackers 2y ago
>days as durations rather than points Isn't thinking of day X as the range [midnight of X, X+1 midnight) isomoprhic to associating it with a point for X, at least for purposes of considering coverage (e.g. both approaches work to sh
850.
by krackers 2y ago
Assuming that DeepSeek continues to open-source, then we can assume that in the future there won't be any "secret sauce" in model architecture. Only data and training/serving infrastructure, and Google is in a good posit
851.
by krackers 2y ago
Gemini 2.5 pro is not the same that powers web search (or any of the dozen other Gemini related things).
852.
by krackers 2y ago
DeepSeek V3 and R1 are about as good (or even slightly better than) 4o already though, so OpenAI wouldn't really lose much by such a release.
853.
by krackers 2y ago
Didn't they say they were going to open-source some model? "Fast and good but not too cutting-edge" would be a good candidate for a "token model" to open-source without meaningfully hurting your own bottom line.
854.
by krackers 2y ago
>I would be interested in reading a paper that does a good job of explaining what a parameter ends up representing in an LLM model. https://distill.pub/2020/circuits/ https://transformer-circuits.pub
855.
by krackers 2y ago
>xnu could alert it when it was low on swap; dynamic_pager would create swap files, and pass them back into the kernel What's the benefit of this indirection through userspace for swap file creation? Can't the kernel create the
856.
by krackers 2y ago
For this to happen they'd have to actually pay attention to spotlight and the quicklook/spotlight plugin ecosystem they abandoned. There's lots of obvious ways to combine LLMs with macOS' unique software advantages (appl
857.
by krackers 2y ago
Summary at https://github.com/ggml-org/ggml/pull/1103
858.
by krackers 2y ago
The smudges on the screenshots got me.
859.
by krackers 2y ago
>So I can't explain why it picks tokens likely to elicit an empathetic response from a human reader. The very fact that the output tokens are being "redirected" is likely something that the model itself recognizes at meta
860.
by krackers 2y ago
So what's the lore with why this took over a _year_ to launch from the first announcement. It's fairly clear that their hand was forced by Google quietly releasing this exact feature a few weeks back though.
861.
by krackers 2y ago
They also had (but squandered) the potential to be ahead on the software side. macOS is the only platform I'm aware of that has every app wired for scripting (AppleScript/Apple Events). And not only that, they already solved the i
862.
by krackers 2y ago
I'm not as familiar with Shortcuts API but from a quick glance it seems less rich than apple events/apple script. With LLM + apple script you could achieve computer use agents on easy mode. Not just one-off "what's the w
863.
by krackers 2y ago
All Apple needs to do is allow 3rd party apps to integrate with Siri, then use an LLM as a way to convert natural language into the set of allowed API calls. Basically similar what chatGPT does with function calling. And on osx, they shoul
864.
by krackers 2y ago
I thought that Cloudflare et al. just had a whitelist based on the SSL fingerprint.
865.
by krackers 2y ago
"Everything is chrome in the future." The article doesn't really say anything new, but _why_ are most browser forks of Chrome instead of Firefox? Is chrome just that much easier to fork? I remember there used to be Camino for
866.
by krackers 2y ago
This doesn't take into account active control though right? It's basically inverted pendulum?
867.
by krackers 2y ago
Compositing is done by the OS though (WindowServer in case of mac). Ever since OSX stopped allowing direct framebuffer access I think the only way to read (e.g. for screenshots) windows of other processes' is by asking WindowServer to
868.
by krackers 2y ago
Very nice catch, I was under the impression that o3-mini was "as good" as o1 on all dimensions. Seems the takeaway is that any form of quantization/distillation ends up hurting factual accuracy (but not reasoning performance)
869.
by krackers 2y ago
Sydney is a glimpse at what an "unlobotomized" GPT-4 model would have been like.
870.
by krackers 2y ago
The developer's blog is really cool, has lots of neat practical opengl insights and stuff https://hacksoflife.blogspot.com/

More ›