Viewing profile — jiehong

I might be in the minority, but for me the best way is installing with a cli (like homebrew or winget on windows). A UI way to install is only a fallback.

Thanks Neil for taking the time to clarify this. Sorry I brought I point in the first place then! Thanks a ton for Berkeley Mono, and good luck on that Houston Mono project! Kinda …

That's fair. I only have a vague recollection, but I found [0]. I think there were also deleted posts on twitter from @usgraphics which I can no longer find, so that's limited. Tha…

I think there was some backlash about the author of Berkeley Mono at some point. Might be a reason for considering this "copy".

Thanks so much for pandoc! I never heard about djot [0], is anyone using it? [0]: https://djot.net/

We all know letters are expensive ^^

To be fair, the Vega-lite backend is json mostly based on GG ideas. In this vein, I think I prefer ggsql that made it to HN recently [0]. [0]: https://ggsql.org/

Well, finally! I wish they wouldn’t mix mutation methods in there, but ok.

The provided TLDR is the subtitle of the article. Limited helpfulness.

The tooling could be better, though. I wish it came with more batteries included, and it renovated that old Java cli split per commands, following a bit what golang did: - java run…

So, basically it’s a little bit of optimisation everywhere to reduce cost and latency (partially found by Sol). The only interesting part is that they trained the model for both ta…

Congrats on the launch! 1. I see it uses python, but I was wondering if it would use mlx-swift instead. Just curious if you considered it. 2. So far, qwen ASR seem usually less wel…

I think one of the point of the author is making might also be that a window manager could do just more, like managing windows as tabs as well. So that any application could have t…

As a user of atuin, it also keep track of the status code of a previous command, as well as its execution time. It’s quite handy.

If we take a browser, and add things to make it a super app, we would basically get an extensible electron app, right? Maybe, why not?

Pretty cool! Only 10B parameters if I correctly understand some of those histograms, which is tiny in LLM world! It bodes well for the future. The training videos seem to show huma…

If we just correctly called it `synthetic data` training, instead of `distillation` (which it isn't), and that fable is accessible through paid-only API endpoints, it would sound m…

> The continuous part is crucial. Many software will paste whatever I said after I have stopped recording, but that is not useful. It really depends on how one uses transcription. …

It does, calling it Elisp

It’s true in general, but not always. Some tools do not work if you don’t treat them with utmost care (like an optical system that might need a clean room to work, or that gets dam…

well, it's nice, but the multi-lingual diff is rather limited (only European languages).

Also, this test is English-only, while a strong point of other models is to understand different languages without first having to say which one (so you don't need 3 different keyb…

That's a very fair point, and Rust shares this fundamental issue as well. That's where one better appreciate the work being done in Zig to get incremental compilation in millisecon…

It's cool, and all, but I'm exhausted by "No X, no Y — just/only Z". > No screen glow, no keyboard, no chat UI. Just ink appearing on paper.