llama_drama

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 of memory when allocating the Vec. This can be easily demonstrated by comparing it with a version that allocates uninitialized memory.

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?

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

> 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 had an issue with extra margin in one of Flutter's built-in widgets and spent hours trying to figure out why - without success.

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, not arbitrary advantages.

If this is exactly the same model then what's the point of encrypting it?

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.

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.

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.

That's weird because it completely freezes m2 ipad

Would this be practical? Or would it be similar to how printers have separate black ink, which is theoretically unnecessary?

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 Chrome.

Is it possible, in general, to know where something belongs — whether it's an LLM or anything else — before you try?

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.

This might be not a good idea. There are some reports of malware (npm packages, iirc) specifically targeting russian computers since the invasion