A desktop made for one
A developer has painstakingly built a custom desktop environment, swapping out almost every off-the-shelf tool for his own, handcrafted alternatives. He even replaced Vim after 25 years in just three days, crediting AI agents like Claude Code for dramatically lowering the barrier to personal software development. This story highlights a growing trend: AI making it feasible for individuals to create perfectly tailored digital tools, fostering debate on its implications for productivity and the future of software.
The Lowdown
The author chronicles his journey of replacing nearly all his standard desktop applications with custom-built versions, creating a computing environment "made for one." This hyper-personalized setup, including a window manager, terminal, editor, and more, is driven by his desire for tools that perfectly match his unique workflow and thought processes. He emphasizes that this ambitious undertaking, once a multi-year project, is now feasible within weeks due to advancements in AI agentic coding, exemplified by his use of Claude Code as a "workhorse."
- The Stack: His custom desktop is built upon two main layers:
- CHasm (x86_64 assembly): The foundational layer handling low-level tasks like pixel rendering and key input, replacing components like i3-wm, i3bar, and kitty with his own
tile,strip,bolt, andglass. - Fe₂O₃ (Rust on crust): The application layer for higher-level tools, replacing Vim, Ranger, Mutt, and others with
scribe,pointer,kastrup,tock,astro, andwatchit.
- CHasm (x86_64 assembly): The foundational layer handling low-level tasks like pixel rendering and key input, replacing components like i3-wm, i3bar, and kitty with his own
- The Vim Revolution: A standout example is replacing Vim, his text editor of 25 years, with
scribein just 72 hours.scriberetains Vim's modal editing but jettisons unused features for custom "writer-shaped tweaks," such as soft-wrap, reading mode, and AI integration within the buffer. - The AI Catalyst: The author argues that AI agents like Claude Code, alongside languages like Rust and well-documented TUI programming, have dramatically reduced the time and effort required for such deep customization. The cost of building bespoke software has fallen "by orders of magnitude."
- Audience of One: Crucially, he states his tools are not for public use or kudos, but purely for himself. This "design without committee" eliminates the complexities of general-purpose software, resulting in small, fast, and precisely shaped tools that bring "quiet pleasure."
This movement toward "Build Your Own Software" (BYOS), even for just one tool, is presented as a newfound avenue for personalization and productivity, offering immense satisfaction without the typical overhead of public software development.
The Gossip
The Dawn of Personal Software
Many commenters enthusiastically echo the author's vision, predicting an explosion of "extremely personal software" or "home-cooked software" driven by AI. They foresee a future where the cost of creating bespoke tools for small audiences (or just oneself) is so low that it will disrupt traditional software development, fostering a new "commons" of individualized applications. This shift could lead to a massive increase in custom software, potentially even open-sourced if cheap enough to create.
AI's Assist: Cost, Control, & Claude Code
A significant thread revolves around the role of AI, specifically Claude Code. Commenters inquire about the actual time and monetary cost of using AI agents for development, with the author clarifying his Claude Max subscription makes the cost "effectively null" for him. Others raise concerns about the "ownership" of AI-generated code, questioning if it's truly "written yourself" or merely piles of code one barely recognizes, contrasting it with traditional hand-coding. The author also clarifies he *monitored* the process very tightly, not just pure agentic workflow.
BYOS: Perks and Pitfalls
The discussion delves into the practicalities of building one's own software. Proponents highlight the liberation of not having to please others, leading to perfectly tailored, efficient tools that enhance personal productivity and offer a great learning experience. They suggest it's akin to gardening or shaping tools exactly as desired. Skeptics, however, voice concerns about security implications, arguing that an individual's custom software lacks the collective scrutiny of widely used OS/browser components. Others question the actual friction AI removes, suggesting prior generations of developers also customized extensively without LLMs, primarily limited by time rather than capability.
Generative Prose: Is the Article AI-Infused?
A tangential but prominent discussion emerges about the article itself. Several commenters speculate that the blog post might be AI-generated due to its "overfitted" language or resemblance to LLM output. The author openly clarifies that he prompted Claude Code for a draft based on his outlined points and then heavily edited (about 50%) and spell-checked it, embracing AI as a collaborative tool for writing. This sparks further debate on the validity and value of AI-assisted content, with some dismissing it as "garbage" and others defending its utility.