llama_drama

5 ms·

1.
by llama_drama 2mo ago
Branchless code can indeed sometimes be slower than conventional one, but in this particular case, the article comes to the wrong conclusion. At a 1% kept, the branchless version is slower because it pays the cost of zero-initializing 8 MB
2.
by llama_drama 4mo ago
Does reporting each model at its best performing reasoning effort introduce a best-of-N/multiple-comparisons bias, especially if models have different numbers of effort levels?
3.
by llama_drama 10mo ago
I don't think that's practical, because the subset would still be quite large. I did a quick analysis of the Factorio localiization files and found more than 1200 unique characters
4.
by llama_drama 1y ago
I wonder if instructions like VPTERNLOGQ would help speed these up
5.
by llama_drama 2y ago
> Everything, including styling, is just a widget. And that's the downside. What would be just two or three nested <div>s in HTML+CSS turns into several screens of widgets, making navigation and debugging much harder. I once h
6.
by llama_drama 2y ago
The problem with AI 'cheating' is that it often shifts the challenge away from strategy and more toward figuring out how to outpace artificial handicaps. A well-designed AI should create challenge through better decision-making, n
7.
by llama_drama 2y ago
If this is exactly the same model then what's the point of encrypting it?
8.
by llama_drama 2y ago
Facts: my friend ran a Minecraft server, and his hosting provider once told him it was the most DDoS'd server in the entire datacenter.
9.
by llama_drama 2y ago
Looks really cool. One thing that I found missing is that I can't use mouse4/mouse5 in explorer. "Sanding" really could go on forever.
10.
by llama_drama 2y ago
The best defense is to keep improving as programmers. By constantly refining our skills and learning new technologies we maintain our relevanse and value. AI and automation are tools that can enhance our work, not replace it.
11.
by llama_drama 2y ago
That's weird because it completely freezes m2 ipad
12.
by llama_drama 2y ago
Would this be practical? Or would it be similar to how printers have separate black ink, which is theoretically unnecessary?
13.
by llama_drama 2y ago
Check out this comment: https://news.ycombinator.com/item?id=40862781 . While it's certainly possible to write the same parallel algorithm in C++, it wasn't done in Firefox before Rust, nor has it been done in Chro
14.
by llama_drama 2y ago
Is it possible, in general, to know where something belongs — whether it's an LLM or anything else — before you try?
15.
by llama_drama 2y ago
I don't think we pushed CSS too far; the complexity of CSS is inherently tied to the complexity of the UI requirements. Even now CSS can't handle everything.
16.
by llama_drama 2y ago
This might be not a good idea. There are some reports of malware (npm packages, iirc) specifically targeting russian computers since the invasion