lobsters score: 1 · tags: linux
deleted by creator
You’re obviously entitled to your own opinions on where your bar is for “vibe coded,” but I started the project years before “vibe coding” was a thing, and I’ve carefully cared for and hand-coded the entire architecture that libghostty-vt is built on over years of dedicated, volunteer work. And along the way being helped by other real humans dedicated their own real time.
It’s hugely insulting to me and others who have worked hard to produce a high quality library with trivialized dismissals of years of a humans actual craft and labor. A couple of our core maintainers are literally teenagers who can’t afford and do not use any AI assistance, and our non-profit helps fund their work to produce high quality work, and has helped some acquire their first jobs ever. (Note: the non-profit also funds non-teenagers, haha, and at least one also uses zero AI assistance whatsoever and maintains a large swath of Ghostty)
How do you think it’d feel to wholly dismiss the labor of a teenager with your low-effort “ugh vibe coded” comment? I assume you’re an adult. I’m embarrassed for you if so.
I welcome anyone to point to any single line in the entire libghostty-vt code base, pull me offline in front of a whiteboard, and ask me to defend it. I’m confident I could. That’s how well I understand every line (and on some CPU architecture every instruction in hot paths by LLVM version).
Besides my own understanding, I proudly defend the work of the maintainers of the project. It is high quality and I am thankful for the hundreds and thousands of hours most have donated totally for free to the project (some are paid for a subset of their hours by the non-profit).
To be clear, in the past year, we’ve accepted AI-assisted changes with a fairly strict policy. I myself do use AI assistance. Nothing goes into libghostty though without real review and understanding. I know that for some, the smallest atom of AI anywhere yields the “vibe coding” dismissal, and you’re entitled to that interpretation, but if you’re asking me to consider the impact AI usage has, I ask you to consider the real human impact your broad dismissals have.
score 1 · canonical
deleted by creator
I know you started the project years before vibe coding. I was excited about your project for years before you started vibe coding. Even switched to Ghostty on all my machines. I’m extremely disappointed in your shift towards embracing AI slop. You can call my disappointment insulting if you want, but maybe the insult is earned.
I think you’re doing yourself and your team a disservice. I think those teenagers who help you develop and maintain Ghostty, who do not do vibe coding, deserve to be associated with a respected non-slop project. But the choice to turn their project into slop was yours, not mine.
You don’t have to agree with my view, and I don’t expect you to. But I’m far from the only one who used to be excited for Ghostty and libghostty.
score 1 · canonical
Your disappointment itself is not insulting to me. It is the blanket dismissal of the entire project. You can be disappointed in me, you can call my code slop if you want, etc. but its insulting to dismiss the years of work everyone has done that wasn’t AI at all, just because some part of it was touched with AI.
I obviously don’t agree with your view. But I don’t need to argue that. I respect your differing view and have no problem with it in general, only the single part of it that invalidates the work of it. To me, this is like dismissing the value of an entire population for the actions of one individual. The remainder of your views are fine to me, despite my disagreement.
EDIT: I felt compelled to edit this to add that I absolutely belief there are massive swaths of terrible AI-written code out there. I believe “slop” exists and there’s lots of it. But I think we probably disagree on the definition of what that is. I feel no need to argue that point, just want to make sure its clear that I don’t think the usage of AI itself is objectively fine in all scenarios.
Genuine question, I’m really not trying to be tongue in cheek here, I’m just curious how far you take your viewpoints: do you consider the Linux kernel slop now that AI is used within it? Are decades of work invalidated wholly to you becase of this?
score 1 · canonical
You are the one who has thrown away years of work, years of manual labour by loads of people, by turning it into slop. You turned the project from an incredibly exciting new foundation for terminal emulator programs, into just another AI project to be dismissed. It’s incredibly sad and a waste of effort. If you don’t respect yourself, I wish you’d at least have respected your contributors.
My opinion of Linux as a project has definitely diminished after they started wholeheartedly embracing AI code generation. Still, I have a practical attitude towards it; there is no another project which does what Linux does, no other open source kernel with the same level of hardware support, so I keep using it and contributing to it until that changes.
There are plenty of other terminal emulators.
score 0 · canonical
It’s hard to imagine what “slop” is supposed to mean if it can be applied to some of the highest-quality software written in decades.
score 1 · canonical
You are the one who has thrown away years of work, years of manual labour by loads of people, by turning it into slop
Those years of work haven’t been thrown away. I’m using Ghostty today. I love it. I’ve never looked at the source code for it because I’ve honestly never encountered a bug as far as I know. It’s obviously, as a user, software developed thoughtfully by thoughtful people whose work I appreciate.
When you reject the work that someone has done, it’s nothing more than that: you’re deciding that that work is not something you want to be involved with. It says absolutely nothing about the work that those people have done other than “I don’t like it”. It continues to bring value and joy to those who do like it.
For me, I wholesale reject the notion that the use of LLMs anywhere in a development process instantly and irrevocably earns it the label “slop”. I’ve worked with slop; I spent a decade doing project-technical-rescue consulting and have seen the kinds of slop humans are able to create all on their own. I’ve seen humans using LLMs generate slop, and I’ve seen humans use LLMs to create and modify elegant code that continues to remain elegant. I’ve seen humans use LLMs to solve subtle rare bugs that have existed for years and having fixed them have objectively made the software better; that assumes you accept the notion that an RTOS that doesn’t randomly hang is superior to one that does, all else code-wise being literally equal (it was a 2 line fix).
You have your own aesthetic tastes and we’re all welcome to that. Own it. But don’t fool yourself into believing that some piece of code not aligning with your aesthetic tastes means that it’s worthless and a reflection of the value of that code to others, nor that the people involved are somehow failing morally (“if you don’t respect yourself…”).
score 1 · canonical
Honestly as someone who probably largely agrees with you I think this anger is misplaced and counterproductive. In the current period of insanity about AI, I have seen mitchellh moving the conversation in the direction of being more careful and not just using a firehose of generated code. That’s a good thing. I question what publicly insulting him is going to accomplish.
Contributors who disagree with the project direction are of course free to fork it.
score 1 · canonical
I’m happy you’re using other terminal emulators. That doesn’t bother me.
Thank you for explaining your value system to me. Interesting goalposts. Thank you.
score 1 · canonical
Interesting goalposts.
Can you explain what you mean by this exactly? My sole “goalposts” are that I prefer projects which don’t contain AI-generated code to those that do. Is that so interesting?
score 1 · canonical
I don’t really understand how someone who works in technology and has probably not only embraced technology but created it for the better part of their life decides to just bury their head in the sand on this next iteration of technology. Whats the next step for you? Refuse to use any software that has AI-assisted code? Might has well give up computing and embrace physical media (those two aren’t mutually exclusive)
score 1 · canonical
Maybe I’m misreading/misunderstanding, but I find it interesting you don’t as strongly dismiss Linux wholly as a “vibe coded kernel,” according to your principles I’ve come to understand above. Does it follow that its maintainers have thrown away decades of manual labor by thousands of people and disrespected their contributors, too? That’s how you’ve described Ghostty and the kernel has more AI-written contributions than Ghostty does, by far, and an incomparably larger blast radius of that impact both socially and technically.
Your “sole goalposts” were apparently strong enough to establish my disrespect for everyone who contributed and the value of the project largely, but for the kernel you depend on, there’s room to acknowledge that the project still has value and you take a softer tone (though, admit you have a diminished opinion of the project).
score 2 · canonical
Ya know, this AI hype goes both ways. One can use a bit of assistance on grunt work without losing understanding if they have discipline. Discipline is definitely possible; currently not aware of evidence to dismiss mitchell and co’s work.
score 2 · canonical
I wonder what your thoughts are on the legality of LLM output? I’m thinking of papers like this one, https://arxiv.org/html/2408.02487v3, that find up to 2% of LLM output is “strikingly similar to existing open-source implementations”.
It’s hugely insulting to me and others who have worked hard to produce [high quality libraries] with trivialized dismissals of years of a humans actual craft and labor.
The reason I ask is because I was struck by how easily that quote also reflects the sentiment held by many open source creators. Trivial dismissal is a perfect way of describing the feeling of those humans that have worked hard to produce high quality software over years of actual craft and labour, only for those efforts to benefit the world without recognition.
You invited the challenge of scrutinising your knowledge of the libghostty-vt codebase. I suspect you aren’t aware of the possible 2% of the LLM-generated portion of the codebase that bares a striking resemblance to existing open source implementations. Sure, we all used to copy and paste bits and pieces from here and there in the before times anyway, so maybe that doesn’t seem like a big issue. But what if certain non-pattern matchable aspects of libghossty-vt bear a striking resemblance to solutions from, let’s say, that very same cohort of non-affluent teenagers you mention? Except this time there is no trace of provenance.
I don’t doubt for a second that you would be among the first to provide attribution if there were a practical way to do so. And I don’t believe that there is any meaningful way to describe Ghostty as vibecoded. I am just struck by how applicable the overall sentiment of your message is to the very body of weights and biases upon whose shoulders AI-assistance stands.
score 1 · canonical
AI learns from patterns and reproduces aspects of them modified by its different persona (or mirage of a persona, whichever you prefer) in mostly the same way you do. When I explain some fact I learned from a well-written article, if I then look over the explanation later, I find that I unconsciously used the same verbiage as that article and shifted closer to that author’s register. This doesn’t produce something on which a copyright claim could be sustained. If I use a solution I heard from somewhere, sometimes there’s an H/T in the code comment and sometimes there isn’t; it’s a nice touch for something that really stuck with me, but you can’t copyright an algorithm. Without opining one way on the community norms, I find posts like these that mention legality are really talking about community norms without a chance of illegality.
score 1 · canonical
- Wholesale pattern-matchable copying of code without attribution, whether by human or machine, is inarguably illegal. This has nothing to do with community.
- I accept that non pattern-matchable influence between programmers and codebases is legal. This then does indeed become a question of community. So if you accept that it is moral or at least okay, then you must therefore also believe that it is wrong for OpenAI and Anthropic to restrict their models from reverse engineering, cloning copyrighted software, binary analysis, etc?
- My actual point is about how the sentiment of mitchellh’s comment so easily also reflects the feeling of many open source creators. This is not about legality. If one believes themselves justified in expressing such feeling towards trivial dismissal of their AI-assisted work then I would hope that one is also able to accept those exact same feelings from those affected by LLM output’s lack of attribution.
- Why would somebody mark my comment as spam?
score 1 · canonical
So if you accept that it is moral or at least okay, then you must therefore also believe that it is wrong for OpenAI and Anthropic to restrict their models from reverse engineering, cloning copyrighted software, binary analysis, etc?
I did say ‘without opining one way on the community norms’. But if I took that belief for granted, no, I don’t see how those two things connect at all. Regardless of your moral position on whether an AI should be offered that does X, it does not imply that it is immoral to offer an AI that does not do X.
This is not about legality.
That is exactly what I said, yes.
score 1 · canonical
AI learns from patterns and reproduces aspects of them modified by its different persona (or mirage of a persona, whichever you prefer) in mostly the same way you do.
Is this broadly accepted legal theory or your opinion?
score 1 · canonical
I use Konsole. When would I notice slowness, that a GPU offloading terminal emulator would know how to avoid?
I poke around the system, I edit files here and there, I build some software packages (AUR and Debian). I can’t say I’ve been blown away by stuttering screens or anything. Am I misunderstanding where the speed increase is? Or is my pc just so slow in general that I’m not reaching bottlenecks? Or do the speed issues only happen to very specific, heavy workloads that I just don’t do? Just curious :)
score 1 · canonical
If I read Wikipedia in a terminal (which I sometimes do) and stumble into a long interwiki section, Konsole can stumble for around ~1s, Contour terminal is only a bit faster even though it talks about «actually fast», WezTerm draws rectangles and background-loads the fonts when there are too many, and Ghostty seems to just manage or maybe preload a good fallback.
That might even push me to WezTerm, although I am not 100% happy about hinting settings; Konsole just does the right thing for my preference (no antialiasing on FullHD) based on generic Fontconfig (so does Ghostty), WezTerm does something worse looking.
And yeah, with software rendering GPU-reliant terminals are laggy.
score 1 · canonical
Which terminal app are you using to read wikipedia? I’m slowly moving towards more of a terminal workflow (playing with aerc, neomutt for reading email, etc.)…and am genuinely curious what other folks use. Would you kindly share? :-)
score 1 · canonical
Immediately, the app I directly use to read text in terminal is Vim. Wikipedia, Vim (browsing in general, too). Emails, Vim. Web feeds, Vim.
This is of course tongue-in-cheek. There is retrieval over network, HTML-to-text, indexing into PostgreSQL, integration with tmux sessions managing the terminal use, integration with the rofi-based launcher, sometimes somewhat with StumpWM, a FUSE FS for translating SQL index of emails into something Vim can manage comfortably…
https://lobste.rs/c/38ubmj probably counts as an answer to your question specifically about Wikipedia
https://lobste.rs/c/mvbtwa for a summary of email setup in terminal (no significant changes there)
Admittedly, with sixel I am constrained by «doing stupid things with sxiv is easier than doing stupid things with sixel scrolling.» https://lobste.rs/c/gonocx
Please ask more if you are interested after reading that!
score 2 · canonical
Ah-ha, fascinating! I appreciate you sharing those links!
So far, I’ve been chaining together the terminal apps that i use in a slightly different manner - e.g. aerc for mail, w3m or offpunk for web content, newsboat for rss/feed reading (though might pivot to offpunk for this too), etc. And, for the last few days i’ve been bitten by the idea that hey if offpunk works for this web content, why not this other content, and that over there too, etc. :-)
But, i love learning from others, both how they set things up, and what tools/apps that they use! Thanks again for sharing!
score 1 · canonical
«Why wouldn’t this be also used for that» is indeed quite how much of my setup is constructed! But yeah, it cuts the things in a different direction than having apps for things.
score 1 · canonical
I have been honestly shocked in the past to learn that my terminal (not sure if it was Konsole or GNOME Terminal, it was a long time ago) was actually causing enough back pressure on stdout to slow down the very verbose debug output from some code I was working on. I triple checked before believing it wasn’t just printf and traced it to write(1,…) blocking. The terminal was pegging a CPU and my program was failing to peg one unless I redirected stdout to /dev/null
score 2 · canonical
I don’t think I’ve ever seen a terminal emulator which didn’t limit stdout pretty strongly.
score 1 · canonical
It’s possible I’m leaving performance on the table by using slow terminal emulators as well, but I rarely feel like I’m waiting on text output. The one case I’ve noticed is tar, where archives with many small files are faster to extract if I omit the output… so I switched to using “xf” instead of “xvf”.
score 1 · canonical
Personally I would say that it’s just something you have to experience, similar to high refresh rate monitors. Obviously a slow terminal and a 60Hz display are fine, but for me personally, once you experience the smoother version, it recontextualises the ordinary version. It’s not a big enough difference that I would give up important features for it, but since I have very few features I require of a terminal, I’ve been happy using the minimal Foot terminal, which is very fast.
It is very much not about the throughput of how much work you’re getting done. It’s about the tactile feeling of a responsive terminal. And it’s just ever so slightly less likely to feel annoying and frustrating. It just makes me happy.
(Although, if you haven’t tried it, I recommend trying some high refresh rate monitors. Or, perhaps, don’t, if you want to avoid the chance of getting hooked as I did)
score 2 · canonical
Oh yeah, I definitely “ruined” standard monitors for myself by going high refresh, high res, and OLED. I might hold off on trying a fast terminal based on that, because I have to use windows terminal at work and terminus on my iPad for personal stuff on trips, so it’s probably good if I don’t see the light there.
score 1 · canonical
Same. When people complain about a terminal being slow I have to wonder what it is they’re doing to run into the problem. Konsole works just fine for me. Perhaps my use (mostly reading mail with aerc, some Vim usage, very very light Git usage… that sort of thing) just isn’t taxing enough.
score 1 · canonical
This looks awesome. Looking forward to seeing this maybe become part of the standard loadout of KDE applications in the future.
score 1 · canonical
Possibly off-topic, but a fun fact from a native Korean speaker:
This is also how it came to bear its name: kwisin, or more typically gwisin, is a romanized form of 귀신—Korean for ghost. It’s pronounced a little bit like kwee-shin.
The word is Sino-Korean, written 鬼神 in Chinese characters. The hanja corresponding to Kwi, the name of this terminal emulator, is 鬼, which on its own means ghost, spirit, or devil. Both 鬼神 and 鬼 mean nearly the same thing in Chinese and Japanese as well as Korean.
score 2 · canonical
Does the source code link work for anyone? I clicked
If you would like to take Kwi for a spin today, check out its code repository on KDE Invent.
And got back “501 Not Implemented” LOL
score 1 · canonical
Maybe they updated it? This is the link that currently shows in the blog post and it works for me.
score 1 · canonical
I have the same issue. invent.kde.org always gives me 501(!) errors when I use Orion (WebKit), but opens fine in Firefox. Didn’t look into it further right now.
score 1 · canonical
i get 501 in firefox and the chromium browser but actually “not implemented” made me think of something… http/2, the new browsers try to use that and maybe that’s what is not implemented. If I open it in
linksit works… well “works” it says “loading” in a minimal html skeleton cuz there’s no script support in that browser, but it isn’t the 501 error. So maybe cuz of http2 in the other browsers? And it works for people whose browsers fallback? idk though if i do curl and command it to use http2, it does work. so what else could it be?Anyway, I know this is tangential and don’t wanna derail too much but I do like to at least glance over the code of these projects
score 1 · canonical
Weird. And it’s still working for me. 501 is a strange error code for something intermittent, too.
score 1 · canonical
it’s probably caused by a misconfigured bot detector. The internet is a hostile place these days
score 1 · canonical
So the terminal with the best Unicode support out of sixel-supporting ones is getting discontinued soon? (Konsole supports sixel, Ghostty doesn’t and won’t)
score 1 · canonical
Out of curiosity, do people actually use and rely on sixel? I’ve always considered it more of a curiosity but never actually seen it used in the wild
score 1 · canonical
My unvalidated understanding is that a number of apps, libraries, etc. with use cases in the terminal, tend to us or rely on sixel libraries. I say that only because i’m moving a bit towards using the terminal for some aspects of my computing, and i see mention of needing sixel to help display images here, or support another app or library there, etc. So, at least to apps that i’m playing with (e.g. aerc mail reader, etc.) they seem to mention using or needing sixel suport.
score 2 · canonical
There are quite a few tools supporting sixel. Like GNUPLOT.
I somewhat used sixel, but interaction with terminal multiplexers is obviously imperfect (and I want my detachable sessions), so I needed to run just one specific thing in a raw terminal. So now I moved the regular workflow to driving sxiv which is side by side to the terminal. But if I decide that having an exceptional session is worth the hassle, I guess I will still use sixel as the more classical standard sufficient for what I am likely to want.
It looks like now WezTerm has Unicode coverage incomparable with Konsole (stronger in one area and weaker in another), and sixel, and slightly incomparable but similar featureset otherwise. So if I migrate off Konsole I can keep sixel option open and go to something that does the unique part of the work in a fully memory-safe way (although of course rendering is horrible and needs libraries, so it’s not much safety gained).
Apparently Contour terminal has sixel and better Unicode feature coverage. But not sure how much I trust all of their crash reports to be from assertions…
score 1 · canonical
The article literally says:
Please also don’t mistake Kwi as some kind of announcement of the future KDE direction
I you want to believe that Konsole is getting discontinued soon, you may choose so but you might be alone.
score 1 · canonical
The decision has not been formally initiated, let alone made, and Kwi might initially be an experiment — but its implementation already changes the chances of different outcomes, and some (although surely not all) of the factors influencing the result of the experiment are predictable.
score 1 · canonical
Kwi is being worked on by KDE developers, so it does give credibility for it to not be abandoned quickly. Like another person who replied, I also think that over the long term it may replace Konsole. Though, who knows, maybe the Ghostty backend can be integrated into Konsole?
score 1 · canonical
Well, I configure my terminals to be white empty rectangles with black text border-to-border, so for me «Konsole» is the backend!
And frontend-side, I guess the question will be whether Kwi will manage to use modern abilities to integrate things people want with less effort than Konsole needed for the original buildout? On integrating with Plasma, Kwi will be ahead by the virtue of starting with the current-for-Plasma components.
score 1 · canonical
Hopefully not anytime soon. I love Konsole and I hope it’ll live on forever.
But, IIRC their original orange site post mentioned this is possible avenue of reducing KDE’s maintenance burden. Maybe long term, if Kwi proves to be good enough it’ll take over the default terminal role. The winds are blowing in that direction. Konsole is a codebase few enjoy modifying (very generational).
If that happens though, I will mourn Konsole’s death. It has the best name out of all terminals and the logo is very fitting among rest of KDE/Plasma.
score 4 · canonical
Sounds great. Konsole is ok but not my favorite due to its cluttered interface, specifically the toolbar and context menu. The lack of toolbar and speed is welcome, let’s see how the rest pans out.
score 2 · canonical
That’s interesting! As @jmmv noted, the menubar can be disabled…Maybe because i’m in the hybrid/transitional phase between adopting more terminal workflows…but i rather like having the menubar displayed from konsole (it doesn’t take up that much space for me). Of course, when i tried kitty and alacritty, their speed is ridiculously faster than konsole…and, i only went back to konsole because i prefer the menubar…and admittedly, the slowness of lkonsole compared to kitty, alacritty and others did not hamper stuff that i needed to do…but if in the future Plasma will offer sort of the best of both worlds (speed and customization) for a terminal app, well then, alright I’d welcome that!!! :-)
score 1 · canonical
You can disable the menu bar and all toolbars, which I do, and then it looks slim. I’m not sure why the context menu is an annoyance though.
But Konsole is slooooow. I switched to Kitty a while ago and the difference is palpable. Looking forward to testing this Kwi option because it’d be nice to have something more KDE-native.
score 1 · canonical
Because it has dozens of options I don’t use (including the main menu in the context menu), like changing the encoding per pane, which I’ve never done in an entire career. It doesn’t have “new tab” entry which I use on every other emulator. Small but paper cuts that shouldn’t exist at this stage.
Honestly I only want maybe six things tops in an emulator context menu, new tab, copy/paste, split, and edit profile. Old gnome term was close.
score 1 · canonical