hexomancer

I have no idea but I sure hope so.

Iranian here. Most Iranians use Telegram or Whatsapp, both of which are blocked right now and can only be used using VPNs (they have been blocked for years, though whatsapp was unblocked a few months …

Yes, every token is expanded into a vector that can be many thousand of dimensions. The vectors are stored for every token and every layer. You absolutely can not fill even a single research paper in …

Here are some objective issues with this page that I don't think is really up to taste (honestly these are so obvious that I assume you viewed the website on mobile which is fair, I never used th…

It's funny that the people who designed this monstrosity of a web page feel qualified enough to advice other people about design.

This is false. The const of producing a single token is linear but the cost of producing an entire sequence of length N is O(N^2) still (which is always what we meant when we talked about quadratic co…

When I read the title I thought maybe they implemented a typescript to binary (instead of javascript) code compiler that speeds up the program by 10x, it would also have the added benefit of speeding …

Do you not see how your argument could be applied to steam engines putting human laborers out of work? Or computers putting (human) calculators out of work? Do you think inventing the steam machine or…

Isn't the fact the P40 has horrible fp16 performance a deal breaker for local setups?

Here is how I make sense of it (I have no expertise in this subject, please feel free to correct me if I am wrong): I think when the model is pretrained on the internet, it does gain most of the skill…

That's beside the point but you are probably referring to harry potter and the chamber of secrets not the half-blood prince.

Page 31 is interesting, where apparently in the task of creating PRs for an internal repository the o3-mini models by far have the lowest performance (even worse than gpt-4o). What is up with that?

If I publish some c++ code that has some hard-coded magic values in it, can the code not be considered open source until I also publish how I came up with those magic values?

A- Who cares? B- The latex source of the book is available on the ArXiv page.

I wouldn't call the case with two sequence "some narrow version". It is by far the most common instance of the problem and honestly the first thing that pops into my mind when I hear LC…

This works the same in vscode as it does in vim as far as I am aware. e.g. `:vs` to split the window and moving between windows uses the same shortcuts as it does in vim.

All of those are available in the vscode plugin.

Can you give example of vim features that you miss in the vscode plugin? Mainly because I likely don't know about them and I would like to add them to my toolkit (the surround thing is not a vim …

Nope, with reverse-debugging enabled it will be insanely slow, not something that you can miss. I am curious does your project have large external dependencies or is it self-contained.

Have you actually compared the debug performance of a large cross-platform application on windows vs linux? I am not saying that you haven't just trying to make sure that your argument is backed …

How large is your code base? I am talking about 100K+ LOC with many complex dependencies (mainly Qt modules).

Slightly off topic but I think it is a good place to ask: One of the few things from windows that I miss when using linux is the debugging experience with visual studio (not code). When debugging a me…