Why are we emotionally tied to command line interfaces
Desktop apps are a second class citizen that do not get feature parity
Lot of actions on Claude Code seem much more suited for a thoughtfully designed GUI
Even the chat responses and links therein can benefit from judicious use of rich text and formatting and real hyperlinks to other parts of the UI or elsewhere
Favourite Skills can be toolbar buttons or menus if user so wishes.
For one I’m not sure if I’ve ever gotten an ad in my terminal or in a tui. That alone is probably enough. And it’s so much harder to ruin the terminal experience compared to a desktop app. They don’t get pointless redesigns. I can customize them however I want. The terminal is a like an oasis in the current climate. And that before getting into utility.
It was very poor UX (from the demos i saw) -- and i tried it myself in Windows native command terminal and the rendering was horribly broken (windows is a second class OS for dev tools).
A light weight GUI with keyboard shortcuts that mimic CLI experience would have been far better without taking anything away from power users of the terminal.
PowerShell was great, if a little verbose. PowerShell's real strength though isn't as an interactive shell, its as a scripting language. You pass objects back and forth, not text. Its basically an interactive API explorer.
doesn't matter how the output is formatted, you aren't manipulating text directly you're working with .NET objects.
Anyway, PowerShell is the way it is out of necessity of the OS. Microsoft did try to just port ksh to Windows at one point it obviously failed because Windows isn't text based, system state is stored in WMI, COM, etc not text files.
The reason for the CLI is because most editors will support CLIs natively, so its very easy to pop in Claude Code either via your IDE or your preferred terminal in your project directory. It's lowkey the most genius design choice by Anthropic. Why they havent made a real GUI since? No idea.
Because it's all keyboard based. Depending on your field of work a good UI can be very different.
A few years ago I watched an account work through my companies numbers using their accounting software. It's entry method? A windows commander like tool. The menu options like add expense etc were all numbered. So he never left the numeric part of his keyboard.
The tool looked super old and obsolete but as soon as you see a power user use it, you see why.
I wasn't alive in the 70s, but I still prefer a terminal.
I can string together a complex series of text-related tasks far more effectively as a shell pipeline than I can by pointing and clicking in a UI. I can scale that sequence of tasks out to operate on every file on the filesystem if I want, or down to a single character in a single file.
Claude Code being a full-featured TUI is also helpful because I can quickly/easily use it remotely via SSH without having to deal with setting up X forwarding, VNC, Parsec, etc. The remote host doesn't even need to have a window manager. Sure, it'd be nice if it also had an elegant multi-page GUI so I could more easily drill into the actions its performing and make better use of my large screen to watch it do multi-agent things, but if I have to choose between the two, I prefer the TUI.
That said, I'd much rather use a GUI to do things that are actually visual/spatial in nature.
can't agree more, I now run my agents in parallel with "agentbox claude", "agentbox opencode" and it teleport my project and settings to an hetzner VPS
The "vt100 teletype emulation" only concerns the protocol in which the program describes what to draw on the display. It might be inconvenient for the programmer, but for the user it is kinda irrelevant.
> Such statements can only be voiced by those that never experienced what using a proper REPL actually feels like.
Maybe. What do you mean with proper REPL? SBCL just runs fine in my terminal emulator. As do all other kind of programming languages, like Prolog.
When I interface with the OS, I want to start programs, control the processes, setup communication channels (pipes) and tell to computer to combine multiple programs to filter, combine, split stuff and redirect it to files and other programs. These should be also be able to run several times with slight differences. All of that works just fine in my shell. What are you missing?
SBCL REPL cannot do this on xterm, it needs a proper hosting REPL environment like SLIME, which is no wonder, given how Emacs came to be and the interaction with genera.
You are missing integration with live debugging, calling anything on the OS during a REPL session, e.g. OS APIs, calling into automation points of the OS,
Can try out directly on the browser, courtesy of WebAssembly and recovery of Xerox PARC software, https://interlisp.org.
Or get either Squeak or Pharo, and see how using the Transcript integrates with the whole platform
Windows, with either PowerShell ISE, or its new VSCode integration, are the closest to these kind of experiences.
To finalise with Xerox PARC view on UNIX, from 1989
"UNIX Needs A True Integrated Environment: CASE Closed"
It might be overcompensation. I think UI, UX and GUIs got better up until the 90s, and early 2000s, but then somewhere GUIs suddenly got a lot worse. So a modern CLI is better and more standardized than a modern GUI.
Electron is the reason, and the elusive dream of "write once, run anywhere" that got us cross platform UIs that are bespoke and don't follow native OS conventions (or keyboard navigation), plus once marketing got involved and GUIs started needing to be branded instead of just fitting in with every other native app on that OS.
I see arguments like this particularly against Electron and the web development sphere in general and I think it's more nuanced than either programmers or "marketing" (read: anyone not a programmer) gives credit towards.
The "elusive dream" of 'write once, run anywhere' is realistically just people wanting to write software with direct product or service use in mind. Native OS conventions are subject to the middlemen of OS vendors, whereas the web (while basically subject to the same vendors) makes a substantial attempt at bridging the gap of writing software for your own purposes without native OS problems. This is a symptom of OSes catering/selling to developers as a platform and hooking them in the 1990s and 2000s.
This attitude that wanting to just make useful code for people and not worry about a windows 11 update breaking everything because they are irresponsible - to think that is not a valid desire is IMO a big problem.
On the other hand, you have a point that it quickly gets out of hand in terms of standards and accessibility and performance bottlenecks. WebAssembly and the WASI are so slow to come out and will by design always be slower than native performance. This doesn't and shouldn't stop us from having decently performant and decently usable program experiences, but it is a prerequisite to care about those things, and the other inheritors of the web development sphere clearly do not want to develop things properly if they take longer than the next fiscal quarter.
There is 100% good Electron code out there, just as 100% there is bad native OS code. The problem isn't inherently the goals of the 'write once, run anywhere' idea; it's more the casualty of other interests pulling away from what developers actually want.
I don't mind a GUI (as long as it isn't an obnoxiously large ribbon or anything) - but if I'm doing work my input device is the keyboard. I don't want to interface with software through moving a mouse pointer when I can just tell it what to do with a few keystrokes.
Why did you lambast it as an emotional attachment instead of a practical preference?
People prefer terminal apps because they run inside our terminal app environments (kitty, zellij, tmux), tend to be keyboard driven, tend to be more lightweight than GUIs, tend to be scriptable, and can be run remotely over a standard ssh session.
I do love GUIs, and use them for most of my workflow. But for Claude, I definitely prefer the CLI.
Since it's a CLI app, I can wrap it in yoloAI for the sandbox protection, and also use VS Code's tunneling feature to reach that sandboxed workdir (with permissions safely bypassed) through my GUI.
I have a similar setup, but I access it directly via iTerm2 instead of VS Code's terminal. I've figured out the right terminal settings to get copying/pasting text to work (including with neovim's + register), but not images. Would be nice to paste images, though. Currently I have to SCP them over.
I suspect the first case worked as intended, and VS Code is greasing the wheels. I'm sure there's a way to get it working in iTerm 2, though I wouldn't be surprised if the solution was some Goldbergian chain of forwarded unix sockets and a helper daemon living inside the sandbox.
Thanks for mentioning yoloAI, though. I started off sandboxing via devcontainers using kata & cloud hypervisor set up as a custom docker runtime. It worked well enough, but nested docker was super slow due to virtio-fs limitations. I recently moved to sysbox and it's a bit quicker. It's probably not as airtight as kata/chv, but good enough to keep Claude from writing a security test that deletes my whole filesystem [1].
Haha yup. yoloAI is to scratch my own itch. I'm getting close to taking it out of beta, but first I'm putting it through a significant architectural overhaul in a feature branch. Normally I'd balk at doing something so heavy, but AI makes it so damn easy to do major mechanical changes (provided you guide it properly and have good tests). So far, so good! And it feels nice to fix the architectural warts before I lock in the interface.
Composability (piping to other programs, or calling them via scripts), reachability (through ssh, for example), focus (not being distracted by all options being present) and universality (cli is more or less the same interface everywhere) are my reasons.
I still use GUI apps too, and actually find claude code to be closer to a GUI app than a cli.
i would gladly use claude code via the desktop app, but it lags behind the cli in terms of supported features, so i just don't bother. last i tried, it didn't support executing CC within WSL while desktop is running in windows.
Personally, I much prefer the CLI. The CLI is a tool that has been refined for over 50 years to excel at text input and output. Once you learn it, it can feel like an extension of your brain.
idk I just like running 6/8 terminal panes and organizing my workflows / projects in an exact space. I even tweaked my theme. and seeing them all on my side portrait monitor.
A text based interface is perfect for interacting with a large language model, and it seems unsurprising to me that it's the most popular way to work with them.
Frankly, the idea of having to decipher what a picture is supposed to represent to use a skill fills me with horror.
>Why are we emotionally tied to command line interfaces
Being a power user, having used computers for more than 30 years, I usually prefer GUI because that's an evolution over CLI.
Going from the basic interpreter on ZX Spectrum to the command line in MS DOS had me mesmerized. Going from the DOS CLI to Windows 95 GUI, had me me mesmerized, too.
I think people in general consider themselves more pro and "hackers" if they use CLI and editors like Vi and Emacs.
There are bonus points for memorizing hundreds of different keyboard shortcuts and not using the mouse at all.
If they absolutely have to use GUI, they not use a desktop environment in Linux but a stacking window manager.
Which is a pity, a real hacker uses a graphical environment inspired from Lisp Machines, Interlisp-D, Smalltalk, selecting code in the REPL with "do it", fixing it on the fly in the debugger and "redo it", changing the work environment in the flow.
Unfortunely they hear that and only understand Emacs.
I am not trying to diminish anyone and I do not have a preferences for how other people use computers. I was merely trying to explain why the CLI gets so much attention.
If you don't believe me, take a look at the leaked codebase from a couple weeks ago. It's the stuff of nightmares, because too many junior devs slopcoded in all places without any plan or understanding of software architecture patterns. They never actually take the time to refactor, there's dozens of outdated redundancies and orphaned modules all over the place.
Without good architecture patterns, there can be no good GUI nor good UX.
Ctrl+V paste works for me on WSL. My secret is that I have given up on WSLg and use a standalone X Windows server. Specifically, the X410 X Server. This removes a whole lot of weird behavior including the ones described by the article.
I have not tried this mostly because I figured it would a resource hog and clunky. Are you describe your experience with X410 on WSL in some more detail? What are the downsides?
you do you, but I've had only good luck with WSLg. my main gripe with it is that it could be doing more. internally (part of?) WSLg uses the RDP protocol, which natively supports audio forwarding, USB passthru and smart cards. yet none of it's wired up.
Why would anyone assume that pasting an image into a CLI interface would work? The fact that this is framed as a bug is wild to me (but very cool that there is a way to do it now)
1) It isn’t a matter of literally “pasting into a terminal” (with the terminal emulator shoveling bytes into the TUI’s stdin), rather it’s “a TUI key-binding tells TUI app to read system clipboard”. No different than any other app.
2) This works on macOS, Linux and Windows, but not WSL. Sounds fair game to call this a bug, or at the very least a feature disparity.
Only tangentially related, but does anyone know if it is possible to ‘paste’ images to an agent harness running inside a docker container?
My current workaround is to paste it inside the working directory on the host machine, then @ reference it, but would be nice to streamline that workflow.
My main issue was the ability to paste images when using ClaudeCode via ssh on a remote machine, so I solved it by having Claude write a quick bridge that fetches the image in your clipboard, rsync it to the server and paste the correct image path in your clipboard: https://github.com/mdrzn/claude-screenshot-uploader
Unrelated but I have a similar problem with speech to text apps on windows, where due to the funkiness of claude codes (necessary) implementation, it doesn't send the keybindings correctly.
This is still better than trying to paste text, files or images in Linux. In latest Pop!_OS I have to keep the app I copy from open until I paste. To add insult to the injury, pasting in terminal produces weird characters.
tl;dr Use Claude Code in WSL inside Windows Terminal? Copying an image in Windows and pressing Ctrl+V in Claude Code doesn't work. Three things break: (1) WSL only hands Windows images to the Linux side in an old BMP format Claude Code can't read; (2) WSL also keeps quietly overwriting your fixes a moment later; (3) Windows Terminal grabs Ctrl+V before Claude Code can see it. The fix is a small Windows program that converts the image to PNG, a Linux script that puts it on the Linux clipboard (and re-asserts once after WSL overwrites it), and one extra keybinding for Claude Code so the keystroke actually reaches the program.
Valve could actually create a need for native Windows games.
As it stands the only reason I have to pain myself back into using Linux on bare metal laptops, instead of VMs, is independence from US technologies in European soil, which also implies not having to care about Claude.
Not everyone here is *nix-pilled (WSL aside). Despite W11's missteps, Windows isn't a completely terrible OS to work on and has some of the best window management outside of a full tiling WM.
I doubt it, Window management has been where MS Windows has been lacking for decades, at most they have caught up. Can you do "Always on top", "Always on the bottom" yet? That's a very basic and easy to implement feature.
Always on top, yes (with PowerToys). But yeah you can't do "always on bottom" yet, but do able with AutoHotKey. Its still lightyears ahead of how macOS manages windows OOTB
Not a bug, pasting images into the terminal is not supported, do not do this, that's not what the terminal is for or how it is used. The standard way is to pass the path of a file to the program as a runtime parameter or in some config file.
Terminals are not alternative web browsers/graphical application sandboxes.
Sixel came out in the 80's as a way to print on dot matrix printers. If your terminal doesn't support that yet, you might want to look into updating your software.
pass the url (local or otherwise) of the image to Claude code. Otherwise it's not the terminal's problem, please don't pressure Microsoft to introduce an attack vector to wsl for slop's sake.
Right, I could paste my clipboard which takes 1 second and zero cognitive load or I can do your suggestion 100x per day and drive myself crazy. No thanks.
Desktop apps are a second class citizen that do not get feature parity
Lot of actions on Claude Code seem much more suited for a thoughtfully designed GUI
Even the chat responses and links therein can benefit from judicious use of rich text and formatting and real hyperlinks to other parts of the UI or elsewhere
Favourite Skills can be toolbar buttons or menus if user so wishes.
It was very poor UX (from the demos i saw) -- and i tried it myself in Windows native command terminal and the rendering was horribly broken (windows is a second class OS for dev tools).
A light weight GUI with keyboard shortcuts that mimic CLI experience would have been far better without taking anything away from power users of the terminal.
https://en.wikipedia.org/wiki/ASCII_art
> ps aux | awk '{if ($6 > 102400) print $11, $6}'
compared to
> Get-Process | Where WorkingSet -gt 100MB | Select Name, WorkingSet
doesn't matter how the output is formatted, you aren't manipulating text directly you're working with .NET objects.
Anyway, PowerShell is the way it is out of necessity of the OS. Microsoft did try to just port ksh to Windows at one point it obviously failed because Windows isn't text based, system state is stored in WMI, COM, etc not text files.
Snover talked about the creation of PowerShell on the Corecursive podcast a couple years ago, well worth a listen: https://corecursive.com/building-powershell-with-jeffrey-sno...
Started using computers when that was the only affordable way to use computers.
For some reason, some people really love to live in 1970's with their expensive HiDPI monitors.
A few years ago I watched an account work through my companies numbers using their accounting software. It's entry method? A windows commander like tool. The menu options like add expense etc were all numbered. So he never left the numeric part of his keyboard.
The tool looked super old and obsolete but as soon as you see a power user use it, you see why.
I can string together a complex series of text-related tasks far more effectively as a shell pipeline than I can by pointing and clicking in a UI. I can scale that sequence of tasks out to operate on every file on the filesystem if I want, or down to a single character in a single file.
Claude Code being a full-featured TUI is also helpful because I can quickly/easily use it remotely via SSH without having to deal with setting up X forwarding, VNC, Parsec, etc. The remote host doesn't even need to have a window manager. Sure, it'd be nice if it also had an elegant multi-page GUI so I could more easily drill into the actions its performing and make better use of my large screen to watch it do multi-agent things, but if I have to choose between the two, I prefer the TUI.
That said, I'd much rather use a GUI to do things that are actually visual/spatial in nature.
For those interested: https://github.com/madarco/agentbox
Such statements can only be voiced by those that never experienced what using a proper REPL actually feels like.
Not something that emulates a vt100 teletype.
> Such statements can only be voiced by those that never experienced what using a proper REPL actually feels like.
Maybe. What do you mean with proper REPL? SBCL just runs fine in my terminal emulator. As do all other kind of programming languages, like Prolog.
When I interface with the OS, I want to start programs, control the processes, setup communication channels (pipes) and tell to computer to combine multiple programs to filter, combine, split stuff and redirect it to files and other programs. These should be also be able to run several times with slight differences. All of that works just fine in my shell. What are you missing?
Inline graphics from 1981,
https://youtu.be/o4-YnLpLgtk?t=376Or
Or using S-PACKAGE used to develop Nintendo 64.
https://www.youtube.com/watch?v=gV5obrYaogU
You are missing integration with live debugging, calling anything on the OS during a REPL session, e.g. OS APIs, calling into automation points of the OS,
Can try out directly on the browser, courtesy of WebAssembly and recovery of Xerox PARC software, https://interlisp.org.
Or get either Squeak or Pharo, and see how using the Transcript integrates with the whole platform
Windows, with either PowerShell ISE, or its new VSCode integration, are the closest to these kind of experiences.
To finalise with Xerox PARC view on UNIX, from 1989
"UNIX Needs A True Integrated Environment: CASE Closed"
https://www.bitsavers.org/pdf/xerox/parc/techReports/CSL-89-...
Electron is the reason, and the elusive dream of "write once, run anywhere" that got us cross platform UIs that are bespoke and don't follow native OS conventions (or keyboard navigation), plus once marketing got involved and GUIs started needing to be branded instead of just fitting in with every other native app on that OS.
The "elusive dream" of 'write once, run anywhere' is realistically just people wanting to write software with direct product or service use in mind. Native OS conventions are subject to the middlemen of OS vendors, whereas the web (while basically subject to the same vendors) makes a substantial attempt at bridging the gap of writing software for your own purposes without native OS problems. This is a symptom of OSes catering/selling to developers as a platform and hooking them in the 1990s and 2000s.
This attitude that wanting to just make useful code for people and not worry about a windows 11 update breaking everything because they are irresponsible - to think that is not a valid desire is IMO a big problem.
On the other hand, you have a point that it quickly gets out of hand in terms of standards and accessibility and performance bottlenecks. WebAssembly and the WASI are so slow to come out and will by design always be slower than native performance. This doesn't and shouldn't stop us from having decently performant and decently usable program experiences, but it is a prerequisite to care about those things, and the other inheritors of the web development sphere clearly do not want to develop things properly if they take longer than the next fiscal quarter.
There is 100% good Electron code out there, just as 100% there is bad native OS code. The problem isn't inherently the goals of the 'write once, run anywhere' idea; it's more the casualty of other interests pulling away from what developers actually want.
People prefer terminal apps because they run inside our terminal app environments (kitty, zellij, tmux), tend to be keyboard driven, tend to be more lightweight than GUIs, tend to be scriptable, and can be run remotely over a standard ssh session.
A conventional GUI is a nonstarter in comparison.
Since it's a CLI app, I can wrap it in yoloAI for the sandbox protection, and also use VS Code's tunneling feature to reach that sandboxed workdir (with permissions safely bypassed) through my GUI.
https://freeimage.host/i/screenshot-2026-05-19-at-141349.ByS...
I have a similar setup, but I access it directly via iTerm2 instead of VS Code's terminal. I've figured out the right terminal settings to get copying/pasting text to work (including with neovim's + register), but not images. Would be nice to paste images, though. Currently I have to SCP them over.
If I copy a file in Finder and paste it into a claude session, it shows in the TUI as [Image #1].
If I do the same, but paste into a claude session running over SSH, it pastes the path to the file, not the data.
If I open the image in Preview, copy the pixels (CMD-A, CMD-C), pasting that into a terminal does nothing.
So it looks like CC just puts UI sugar over top of the image path when it has file access to it? That's not really image pasting, though...
Thanks for mentioning yoloAI, though. I started off sandboxing via devcontainers using kata & cloud hypervisor set up as a custom docker runtime. It worked well enough, but nested docker was super slow due to virtio-fs limitations. I recently moved to sysbox and it's a bit quicker. It's probably not as airtight as kata/chv, but good enough to keep Claude from writing a security test that deletes my whole filesystem [1].
1: https://github.com/anthropics/claude-code/issues/28521
I still use GUI apps too, and actually find claude code to be closer to a GUI app than a cli.
Personally, I much prefer the CLI. The CLI is a tool that has been refined for over 50 years to excel at text input and output. Once you learn it, it can feel like an extension of your brain.
Frankly, the idea of having to decipher what a picture is supposed to represent to use a skill fills me with horror.
Being a power user, having used computers for more than 30 years, I usually prefer GUI because that's an evolution over CLI.
Going from the basic interpreter on ZX Spectrum to the command line in MS DOS had me mesmerized. Going from the DOS CLI to Windows 95 GUI, had me me mesmerized, too.
I think people in general consider themselves more pro and "hackers" if they use CLI and editors like Vi and Emacs.
There are bonus points for memorizing hundreds of different keyboard shortcuts and not using the mouse at all.
If they absolutely have to use GUI, they not use a desktop environment in Linux but a stacking window manager.
Unfortunely they hear that and only understand Emacs.
That's the real reason.
If you don't believe me, take a look at the leaked codebase from a couple weeks ago. It's the stuff of nightmares, because too many junior devs slopcoded in all places without any plan or understanding of software architecture patterns. They never actually take the time to refactor, there's dozens of outdated redundancies and orphaned modules all over the place.
Without good architecture patterns, there can be no good GUI nor good UX.
(disclaimer: I work at MS, not on WSL)
1) It isn’t a matter of literally “pasting into a terminal” (with the terminal emulator shoveling bytes into the TUI’s stdin), rather it’s “a TUI key-binding tells TUI app to read system clipboard”. No different than any other app.
2) This works on macOS, Linux and Windows, but not WSL. Sounds fair game to call this a bug, or at the very least a feature disparity.
My current workaround is to paste it inside the working directory on the host machine, then @ reference it, but would be nice to streamline that workflow.
I sure wish it didn't have to be a console app
https://github.com/ghostty-org/ghostty/discussions/10099#dis...
Code: https://github.com/rajveerb/wsl-clip-bridge
As it stands the only reason I have to pain myself back into using Linux on bare metal laptops, instead of VMs, is independence from US technologies in European soil, which also implies not having to care about Claude.
I'd rather continue to be as productive as possible.
Terminals are not alternative web browsers/graphical application sandboxes.
If it ain't broke, don't fix it