Im less interested in the memory than the memory bandwidth. The current system with 128GB can load pretty big models, but its meaningless unless you want to wait 40 minutes per prompt.
I've had best results with Qwen3.6-35B-A3B, which uses 40GB of memory, but only uses 3 billion parameters per token which helps with throughput.
Until memory bandwidth significantly improves I just can't see myself wanting to use all that memory. Unless it's just to keep a wide variety of models in memory.
On a GitHub issue page, it will show backlinks to anywhere on GitHub that issue has been referenced. This is quite helpful at times because it leads me to find how other teams have worked around an issue I'm dealing with.
There's value in related information being aggregated and linked across many projects.
Japan has their own communication platforms for this, they're not relying on a US government site. I'm in Japan on vacation, and I got notified of the earthquake within a minute of it happening on the NERV app, which is a common disaster alerting app here.
This gave me a blast to the past to Nightingale, the media player built on top of Firefox. It was a Firefox fork that was aiming to be a more powerful alternative to iTunes/Winamp. But since it was built on Firefox, you could also use it as an all-in-one media player and web browser.
The homepage still exists, but it looks like many of the other pages like the blog and wiki are long gone. It hasn't been active in probably over a decade.
ar_lan specifically mentioned Neovim distributions. Examples would be LazyVim and AstroNvim. These are packages you can install that provide Neovim in a pre-configured and opinionated way. They generally come with language servers, linting, and various other features out of the box, and have their own paradigms for configuration.
They can be easier to get started with than just installing Neovim from scratch. But they add their own complexities. First, you have to know that they exist, and pick one. Then you have to know how to configure them, they may have their own nuances about how things are done. Under the hood they're using all the same packages, so you'll need to learn how to configure those as well if you don't want the defaults.
I would say the distributions to make it extremely easy to get started with a functional IDE experience with LSP features. But they're not without their own learning curve.
It's not compiled in the way that C is compiled. Transpiled would be a better term (though there are debates on where the line is).
Amber code gets turned into bash code, and run by a bash interpreter. So at best Amber's performance will match Bash's performance.
I've seen people say bash is faster than PowerShell, but I don't have benchmarks to back it up. Even so, I wouldn't recommend using it for performance intensive tasks such as writing a web server.
The great advantage I see for Amber is being able to write scripts in a sane language (bash is not enjoyable to write), and have those scripts be able to run anywhere that Bash is installed.
That's quite disappointing. As a Neovim user, I depend on nerd fonts for icons.
Does the license forbid patching even for personal use? Or is it only forbidden to re-distribute the patched font, by using it on a website for example?
I've had best results with Qwen3.6-35B-A3B, which uses 40GB of memory, but only uses 3 billion parameters per token which helps with throughput.
Until memory bandwidth significantly improves I just can't see myself wanting to use all that memory. Unless it's just to keep a wide variety of models in memory.