Stop Making TUIs

The Catalyst: A Seemingly 'Niche' Advocacy

Recently, two influential figures in the developer community—Simon Willison and Thomas Ptacek—put forward an interesting advocacy: "Stop making TUIs (Text User Interfaces)." They aren't against command lines or terminals per se, but are targeting a very specific scenario: those "throwaway" personal tools, scripts, and small applications developers build for themselves or small teams. Thomas Ptacek bluntly stated that if you haven't tried turning one of your "500 throwaway CLIs" into a native app, you're "doing yourself a disservice."

This advocacy is worth discussing now because it's powered by a strong technological force: AI coding assistants (particularly "vibe-coding") have reduced the cost of developing a "good enough" GUI to a surprisingly low level. Previously, creating a graphical interface for a small script might mean learning frontend frameworks, dealing with cross-platform compatibility, debugging layouts... This "extra work" often deterred developers, leading them to think, "the command line works, so it's fine." But now, with AI, a descriptive prompt might generate a fully functional, decent-looking desktop or web application. Simon Willison shared his own experience of developing macOS menu bar apps for bandwidth and GPU monitoring this way, which he uses daily.

Deconstructing the Core: Beyond 'Looking Good' to Interaction Evolution

The core point here is definitely not just that "GUIs look better than command lines." It touches on the efficiency and mindset of human-computer interaction.

1. Cognitive Load & Intuitive Operation: A well-designed GUI, through visual elements like buttons, sliders, and charts, turns complex commands and parameters into intuitive actions. Users don't need to memorize command syntax; they can see available operations and current states at a glance. This lowers the barrier to use and reduces the chance of errors.
2. Richness of Information Presentation: Command-line output is a linear, scrolling stream of text. A GUI can present information concurrently in multiple ways—charts, dashboards, tree structures—giving a clear overview of system status. Like Simon's monitoring app, data is displayed directly in the menu bar without needing to open a terminal and type commands.
3. Changing the Mindset of Tool Building: As Thomas said, trying to build a GUI "will probably change the way you think." When you start conceptualizing a graphical interface, you naturally think from the user's perspective (even if it's your future self): How should this feature be discoverable? How is this data most intuitively displayed? This process encourages deeper design of the tool's interaction logic, not just its functional implementation.

Trend Insight: AI is Reshaping the Definition of 'Developer Productivity'

This phenomenon reveals a deeper trend: AI is shifting the core of 'development' partially away from 'writing implementation code' towards 'describing requirements' and 'designing interactions.'

In the past, the long-term cost (development, debugging, adaptation) of maintaining a GUI tool was the main barrier for an individual developer or small team. AI coding assistants drastically cut the upfront "coding implementation" cost, making it feasible to maintain a more user-friendly interface over the long term. This could spark a minor wave of "GUI-fication" for personal tools.

We might be witnessing a quiet upgrade in the developer toolchain: Command-line tools are the Swiss Army knife—compact, powerful, but requiring skill. AI-assisted GUI tools are like a customized multi-tool for a specific task, with clear buttons and icons. They might not be as "geeky," but for specific tasks, they're ready to use and more efficient.

Practical Value & Reflection

For interested IT practitioners, the takeaway is direct: For your next personal efficiency script, data processing tool, or monitoring gadget, don't hesitate to use an AI coding assistant to "wrap" it with a simple graphical interface. This is no longer a time-consuming, extravagant idea, but a quickly testable practice. You might find the tool's usage frequency and your satisfaction with it increase as a result.

However, there's a counterintuitive angle to be wary of: A "good enough" GUI is not the same as an "excellent" GUI. AI can quickly generate a usable interface, but poor design (chaotic layouts, unintuitive operations) can also be generated just as quickly. As developers shift from being "CLI designers" to "GUI designers," they still need to learn basic interaction design principles to avoid creating a "monster" harder to use than the command line. Reduced cost is an opportunity, but design taste and user-centric thinking remain scarce skills.