sbinnee

Thanks for sharing

I have been staying on vim-plug where everyone seems to be moving on to lazy.vim. I was considering a weekend to migrate to lazy.vim sometime. But with this news, I might just have to wait for this plugin manager to drop.

Does anyone know what’s the ground they want to make a builtin plugin manager? Did they think there should be some sort of standard?

I happened to listen to a podcast episode yesterday where the topic was Git rebase. There were 4 people in the panel. One person made a point that he did not understand why they even care, saying basically Git is broken, all the VCSs are broken because they are too complicated and they get in the way. He said a lot of things but my take was that all the existing VCSs are just a distraction. I personally don't agree with him. But while listening to his argument, I started imagining a perfect VCS which is not Git, JuJutsu, perforce, or whatever.

This got me thinking why I have not tried Jujutsu yet (which was mentioned in the episode briefly) even after I saw it multiple times on HN over the years. It is probably because it just looks like another Git, maybe a less complicated one. But nonetheless it's going to be Git-like, and I know enough about Git and am competent with it. So there is little incentive to learn and try another Git-like tool.

I would love to try a VCS vastly different from Git. But it cannot be just different, it should be better than Git. That, I think, is hard to come by.

By the way, if anyone is interested which podcast I am talking about, this is the episode. "The Real Problems w/ Git" on ThePrimeTime. https://youtu.be/t6qL_FbLArk?si=GOkixm86rFADoc4x

I am surprised that the walrus operator had to do with Guido's resignation. The walrus operator is awesome. There are so many cases you need a temporary variable in a control flow. The walrus is a beautiful way to handle it.

I always knew that there should be a way to have a button to change light and dark mode on my website without javascript. But I couldn’t figure it out (pre llm era) and gave up and just used javascript. Now it’s time to make an update on that. This blog motivated me.

Maybe that’s how we are supposed to work with current LLMs. Having a template or an example we understand is a big advantage. It can prevent LLM from many pitfalls, and more importantly we can easily review the output.

I would to try Pixel phones. But it's simply not available where I live.

Yeah I loved it so much. Thanks for sharing.

>> There are actual opportunities for career development. > Does "career development" just mean "more money"?

To me it seems it’s related to the specialist-generalist point that it’s easier to focus on one thing rather than to do all the chores you face in SME environments.

I can't agree more

I think it’s a good monetization strategy for astral team. I just hope that they do not lock uv only on pyx and that uv seamlessly works with pypi.

Linux is fundamentally appealing to those who want to tinker with OS. Even with this kind of full-featured ricing or package, users will have hard time to figures some features out, if they do not have enough knowledge how to navigate Linux system.

I personally use all three major OS, Mac at work, Windows for gaming, Linux for everything else. For me different OS serves different purposes and gives me task-oriented productivity boost.

I started seeing a lot of dhh talking about Omarchy on twitter. It's been a joy to see fellow hyprland user, vocal about hyprland and linux. Linux users have been talking that every year that it's going to be the year of linux desktop, but now it feels so close when a person with big influence talks loud on social media.

I checked my git-log from my dotfiles. It says I installed hyprland 2023-07-23.

> 319649c 2023-07-23 (sbinnee) install(hyprland): wayland wm tag: hyprland

I used to run i3, dwm, bspwm (my favorite on X11) and tinkered with other wms. Since this commit, I have been full-time on Wayland. When Hyprland were young, it certainly had rough edges. But these days I don't feel overwhelmed even when I update several releases at once. It is stable and just works. The creator of hyprland, vaxry, is an incredible developer and maintainer. He's made so much progress on usability of Wayland.

This post caught my eyes immediately because I have been sort of benefiting from StarDict project. Although I do not use it directly. I have been using sdcv, a CLI tool that reads StarDict dictionary. It’s minimal and serves me well.

My impression is the same. To train anything you just need to have CUDA gpus. For inference I think AMD and Apple M chips are getting better and better.

I didn't know that OpenAI added what they call organization verification process for API calls for some models. While I haven't noticed this change at work using OpenAI models, when I wanted to try GPT-5 on my personal laptop, I came across this obnoxious verification issue.

It seems that it's all because that users can get thinking traces from API calls, and OpenAI wants to prevent other companies from distilling their models.

Although I don't think OpenAI will be threatened by a single user from Korea, I don't want to go through this process for many reasons. But who knows that this kind of verification process may become norm and users will have no ways to use frontier models. "If you want to use the most advanced AI models, verify yourself so that we can track you down when something bad happens". Is it what they are saying?

When they were about to release gpt4 I remember the hype was so high there were a lot of AGI debates. But then was quickly out-shadowed by more advanced models.

People knew that gpt5 wouldn’t be an AGI or even close to that. It’s just an updated version. GptN would become more or leas like an annual release.

vllm and ollama assume different settings and hardware. Vllm backed by the paged attention expect a lot of requests from multiple users whereas ollama is usually for single user on a local machine.

I think the key was multimodality. Meta made a big move in combining texts, audio, images. I remember imagebind was pretty cool. Allen AI has published some notable models, and Matt seems to have expertise in multimodal models. Molmo looks really cool.

Cool. I applied through the career page, although it wasn't for the positions you laid out here. I was surprised to see a position in Korea.

My fingers are just too used to tmux and I can do all I want to do. But I hear the scroll issue. When I have to copy multiple chunks, I have to go to cp mode visual a chunk and as soon as I enter tmux goes to the bottom line and I have to scroll up again…

At this moment it looks clear to me that Apple won’t go that way. It’s enough for them to focus on inference and actual application not the heavy training part. They have been probably training models on a cluster with non Apple silicon and make them available for their chips only for inference.

I really enjoyed the three body problem. At some point it became a bit ridiculous but I liked the ending.

I like the very idea of applying this transformative AI technology to make values. I have been doing research and always been a member of research team ever since I started the ML journey. But, I've always valued applications that can solve problems and found myself inclined to building apps. I would love to work at Answer.AI. How can I apply?

I like --amend a lot too. Usually a new feature requires a few steps. I commit every step and make a list in git commit body. It's basically almost same as rebasing at the end, but amending seems more logical to me.