Strace-ui, Bonsai_term, and the TUI renaissance
blog.janestreet.com
blog.janestreet.com
There are other great examples of TUIs that i've seen around the web:
https://github.com/ratatui/awesome-ratatui
https://terminaltrove.com/explore/
https://github.com/rothgar/awesome-tuis
But I think we will swing back to using GUIs when we find a performant way of making them, I don't know what it is yet but surely someone is working on this.
TUIs look good to look at though!
I like this new TUI renaissance as well, but if you wanna see what a symbiotic relationship between GUIs x TUIs could look like you need to see what Emacs does with Orgmode and the whole Org ecosystem of org-agenda, org-roam, etc. Lot's of these TUIs from the awesome are somewhat already inside Emacs.
https://orgmode.org/ https://orgmode.org/manual/Agenda-Views.html
- TUI libraries like Lipgloss and Bubbletea really allow users to build a rich experience nowadays. Really anything from https://charm.land is well-polished and a joy to use.
- Kitty image format works great and allows for Avatars, image previews, etc, which helps immensely make slk feel like a Slack client and not an IRC client.
> But I think we will swing back to using GUIs
I've been pushing on BubbleTea Kitty and Ghostty quite a bit to hybridize this. The TUI / GUI distinction to me is about task centrism and delivery. There's an appropriate surface and workflow for every task; beyond TUI/GUI sometime it needs to be a VR headset or an immersive room or a literal sandbox.
A demo of this is web-delivery of our BubbleTea TUI examples ('t' toggles between glyph/kitty):
https://nimblemarkets.github.io/ntcharts/demos/heatpicture-p...
The delivery uses our WIP Booba tool, which is Ghostty-based. The CLI tool can be used to remote or embed any TTY program, but was generally built for BubbleTea. https://github.com/NimbleMarkets/go-booba
I've recently made SVG, PDF, SVG, and OpenStreetMap widgets.... https://github.com/NimbleMarkets/ntcharts-svg https://github.com/NimbleMarkets/ntcharts-pdf https://github.com/NimbleMarkets/ntcharts-osm
Right now I'm working on multiple ds4 TUIs using this stack, for example generating SVGs from a prompt and then rendering it in TUI. Another generates CSG object graphs and renders/composites them in terminal. Here's a gist with screenshots:
https://gist.github.com/neomantra/ae47422c8daf7a458212c93992...
The upstream ds4 project is using C for their TUIs. I have done TUIs in C and C++ (and many other languages) and will not go back to that. Really fun engine though and a great place to stick a TUI. I am a Camel furry (https://cameltopia.org) but wouldn't use OCaml to make a TUI either (makes sense for Jane Street of course).
The only software that is as fast as TUI is the Zed IDE. Apparently they use Rust + their own built GUI toolkit with GPU rendering.
And apparently it's tightly coupled with Zed.
I might be missing something here, but wouldn't any UI toolkit that doesn't live within a WebView work?
And, sometimes, all you want is to output text.
I mean, I don't even think that's true. Many TUIs are bloated, dogshit slow, and it's not trivial to write complex TUIs without glitches or flickering. The more people start making TUIs because it's the current fad, the worse they will get.
I think it's mostly due to the CLI being much more powerful than a full-blown IDE: when you try to make everything simple (in a menu, accessible with one mouse-click or a shortcut), then semi-advanced things (which haven't been thought of in the IDE) become impossible.
They also compose way more easily with the network: it's much easier to SSH/tmux TUIs than to try to network GUIs. That's very important in this day and age of incessant exploits and the need to run things in VMs to try to prevent exploits from pwning your entire world.
TUIs sadly doesn't entirely solve the bloat issue: look at Claude Code CLI... An Electron app because, as the Pi author mentioned, people at Anthropic thought they were writing a game. A full headless Electron app converted, on the fly, to pretend it's a TUI.
The madness is real (shitload of message of users not happy with characters shown on screen not being those saved to source files: and I've been bit by this) and the bloat issue hasn't been solved.
But in any case you cannot go back to a GUI and not have that problem of GUIs being too restrictive: invariably there shall be a need to chain several simple commands that do one thing and that excel at that one thing and that only become powerful when chained with other commands. GUIs cannot solve that.
LLMs have just proved what many knew all along: that the CLI is more powerful.
Instead of GUIs on the contrary I expect more TUIs, more command line, and more, much more, REPL usage by LLMs and by devs using LLMs.
What I do hope though is that we get more lean TUIs (like Pi by Mario Zechner, all the utils written in Rust etc.) and less madness like "headless Electron converted 60 times per second to a fake TUI".
But TUIs can't solve that either? The ability to pipe CLI commands together when using a tool non-interactively is orthogonal to what UI is used interactively.
I think people are confused and think they like TUIs because they like them being keyboard driven etc. But this could all be done with a GUI.
The other factor is probably just fashion. Similar to how some kids are now listening to music on cassette tapes, which are objectively worse than other media in almost all respects. The less cynical take is it's like vinyl: it does come with compromises but gives us back some of the things we lost over the years.
The actual interesting text-based interface is the CLI. I've seen a few examples of TUIs that really should be a CLI and would be much more useful as such.
- frameworks for tui development, including react-like DOM shit
- further ad-hoc specs building on top of ANSI escape codes for finer control
- maybe a scripting language
React-ncurses FTW
[1] https://github.com/jrop/morph.nvim [2] https://github.com/jrop/tuis.nvim
Display technology has seen so much progress in the past decades. Apple marketing has taught us about "Retina" displays with pixels so small that you can't tell them apart without a microscope. We get these very rich and colorful desktop environments but we actively decide to not use any of that.
Now, I get that a TUI can look incredibly crisp with proper text rendering, kerning, ligatures, nerd fonts and so on, but still with all that, at the end of the day we still have a thingamajig that implements a VT100. It is a strict subset of what could potentially be drawn with a proper GUI framework.
I understand that TUIs can run over SSH channels, can be juggled with Zellij/tmux/mprocs -- as such they are composable in the sense that they can be used in a way the author(s) didn't think of. It's been a while since I've done any of that personally, and I for one think it's a bit of a cop-out that the Claude Code integration in PyCharm is just the TUI [1] inside a terminal emulator inside my IDE when it could be so much more, just to provide one example.
The article shows off an strace TUI, and it's not like I can't see the benefits of making strace output more browsable. What I don't understand is why that must happen inside a terminal window where (for instance) all text must have the same font and size.
So what is the appeal? I'm asking in good faith. Is it because the perceived alternative is another run-off-the-mill Electron RAM guzzler, because there aren't any _good_ GUI widget frameworks? Is it the multi-platform aspect?
If all we work in are these super-lean TUIs maybe we don't even need so powerful computers or such high-DPI displays anymore?
I'm genuinely puzzled, but interested to know how TUIs appeal to other people.
1) which, I understand, is itself a React app with a console renderer!
- Optimizing for fast, keyboard-only usage
- Allowing me to customize the presentation according to my preferences
- Not having to leave my terminal, where I spend most of my time (I do realize this is something of a chicken-and-egg situation)
I think also a big problem of desktop apps is that you have to deal with window management. Now that I am on Niri it is really apparent to me how much I hated juggling windows in pretty much any other window manager that I ever used, except, interestingly enough, tmux.
I personally prefer same font (one I choose) everywhere opposed to "whichever font developer likes for his app).
TUI apps are quick, efficient, portable and usable through ssh. If crisp graphics is the only downside I really don't mind.
I do this all the time.
One of my favourite applications is a tool called "autoexpect" and I use it every time I try a new program.
What it does is this: I run a program in it's virtual terminal, and it writes a TCL script that does what I did, and puts little regex tests in for the output of that program for me. I can then edit that program (or not: sometimes the first output is fine).
Once upon a time I used to use a program called DESQview: It had a "learn" feature that allowed you to record and playback even DOS programs, so it was very easy to pick up autoexpect.
DESQview/X was their X11 server, and it also had the "learn" feature, but unless the application could be driven entirely by the keyboard, it didn't work; most similar applications I've seen over the decades since need such care for reliable "scripts".
Yes sometimes you also have the possibility of using the GUI accessibility framework to "script" the app. This is barely ok if it works, but most GUIs that I want to script were designed so that would not work at all, and it is coding that requires me work with the app instead of asking a domain expert for a recording.
autoexpect on the other hand is just text, easy to read and modify, and easy to send by email. It is hard to make a terminal application hostile to autoexpect without a great deal of work that (in the text based environment) can usually be undone just by using tmux and mosh on loopback.
> What I don't understand is why that must happen inside a terminal window where (for instance) all text must have the same font and size.
Modern (as in, since the 1980s) terminals are very capable of multiple fonts and font-sizes. I usually use a non-proportional font for coding myself.
Is that really true? Can I, in one terminal window, have prose with a serif font and code with a monospaced font at the same time? If I hover over something with my mouse, can I have a TUI tooltip at a smaller font size?
There once was a program called https://en.wikipedia.org/wiki/ManaGeR which appears at first blush to be some kind of X11-competitor, except it was using the VT330's regular terminal capabilities to do those fancy pixel-patterns and fonts, and so there's just some weird VT escape sequences you've never heard of in there.
You can also use SIXELs if you want even more control, and you can readily see such things in action because qemu can (in 2026) send its graphical VGA display into a sixel terminal, but in the 1980s such a thing would not have been performant (probably something like 3 frames per minute) because the VT330 was slow, and such a thing would not be popular you would "lose the text" at some layer which would be as inconvenient as using any other graphical application.
The latter is mostly because certain GUI patterns simply cannot be implemented with a TUI.
And then everyone has their own idea of what keyboard shortcuts should be like. Yuck
For those workflows (as opposed, to say, Photoshop), we could do without that. That's the whole benefit.
>and being self-explanatory and same-y.
GUIs are quite less same-y that TUI. Not to mention the same app GUI can be widely different between 2010 and 2026, whereas any TUIs from 1990s I still use look and work the same.
Yeah, I think that is 90% of it. And the whole related ecosystem aspect. All the major ways of building GUIs suck right now, especially for tiny apps. And to further exacerbate the problem, GUI frameworks are generally tied to their programming languages, Qt is C++, SwiftUI is Swift, Flutter is Dart so on; spewing some terminal escapes to stdout is something that can done from basically any language with relative ease.
A macOS AppKit/Cocoa app uses like 30mb memory and is quite nice to develop or vibe-code. But the whole UX of it is the opposite of a TUI you can run in yet another pane of your terminal emulator next to the other panes you're working in.
GUI apps certainly have their place. But sometimes you want a UX like lazygit where you can launch it in cwd and be done rather than alt-tab to a GUI, open pwd inside it, then alt-tab back to the terminal.
[1] https://github.com/jrop/tuis.nvim [2] https://github.com/jrop/morph.nvim
Bypassing all that crap is the main benefit. I don't want any of that superfluous, and ever-changing stuff, when the terminal and, if needed, a TUI, is enough. As a bonus, it works everywhere I can ssh to.
>If all we work in are these super-lean TUIs maybe we don't even need so powerful computers or such high-DPI displays anymore?
For a lot of stuff, you never needed it. For others, like an IDE, DTP, gaming, 3D modelling, video editing, photo editing, they're amazing.
Yes, there are things like https://github.com/ocornut/imgui, and some (especially open source) applications try and muddle a long with Qt or GTK, but many (most?) serious professional or power user applications have built their own GUI frameworks or at least custom controls to deal with this.
Whatever route you take, as a dev it's painful, especially for someone who remembers adding a couple of libraries to a Delphi project back in the Office 2000s era and getting full docking, configurable toolbars, etc. with little to no work.
So the easy fallback (especially with the recent proliferation of libraries) is TUI and CLI applications with the layout/docking and tabs provided by the terminal emulator itself or one of tmux/zellij/etc.
I've been thinking on and off for a few years now about the idea of a "graphical terminal", sitting somewhere between a GUI toolkit and a terminal emulator and a full blown OS for building inter-composable apps and tools and components that could replace TUI based workflows/apps/layouts. I have a vision of every "pro" app just being a different curation and configuration of underlying components rather than actually separate software.
Are you thinking about emacs?
No amount of cartoons or colourful bouncy animation of pictures on the screen can replace simply displaying text when what you need to do is display text on the screen. And it turns out that text is a really good way to display information.
A picture is worth a thousand words but why would you use 1000 words when only one or two will do?
It is hard to remove the rise of agentic flows and LLMs generally from this discussion. The fact that LLMs understand text in a first-class way and have had more time to get good at them, means that they can iterate and build apps and features that might bog down a model in the context of a GUI app and all of the additional context required for the larger framework and deployment and testing story.
These attributes and the current environment with LLMs mean that it's really easy to build ephemeral UIs that serve a purpose right now, and then consider productionizing or generalizing at a later date. If you decide to do this, you have the really high fidelity tests that make it easier.
The article shows off an strace TUI, and it's not like I can't see the benefits of making strace output more browsable. What I don't understand is why that must happen inside a terminal window where (for instance) all text must have the same font and size.
Somewhere along the way, GUIs seemingly went from economy of use to a graphical resume for the developer (or their manager).
To be fair to Jetbrains, users would get charged API prices if they built their own integration. Zed recently added support for terminal sessions in the agent chat history for precisely this reason. Blame Anthropic.
- It opens to the editor mode rather than the gameplay mode on launch
- It makes a .run/ directory next to the executable if one doesn't already exist
- It makes a timestamped directory within .run/ for this current debug run
- It automatically records stdout to stdout.txt, stderr to stderr.txt, and a crash.txt if the game crashes, in the directory for this run
- When the “take debug screenshot” function is invoked (which can be done by pressing F12), it saves a timestamped (based on time since executable launched) screenshot in the directory for this run
- Editor actions and 3D camera movements are recorded to playback.txt in the directory for this run
With all of this in place, I can do a debug build, run the game, do something in the editor, and take one or more screenshots where things went wrong. Then, Codex can see the log files and screenshots and try to diagnose the problem. When attempting to fix the problem, it can automatically recompile the debug build and rerun it with a launch option that plays back the latest recording file, which does the same sequence of editor actions/camera movements and takes screenshots at the same points in the process. Then it can compare this to the initial recorded run and see what needs to be fixed.
We could be having a GUI renaissance right now but for various primarily aesthetic reasons people are churning out TUIs, and personally I think it's a huge mistake.