AI-Coded Markdown Editor Puts PDF Markup, Git Commits, and a Terminal in One Window
A developer with no front-end background just shipped a complete, open-source desktop editor by treating AI as the implementer and himself as the product owner. XJ-Editor is concrete evidence that the cost of creating custom, personal software has dropped to describing what you want, and its code is available for anyone to extend or learn from.
XJ-Editor is an open-source Windows desktop application whose code was written entirely by AI. The developer behind it is not a front-end engineer; he acted as product manager and tester, feeding requirements to an AI that turned them into working Electron code. The tool's reason to exist is to stop the habitual window-switching that plagues Markdown writers: editing, previewing, converting, and embedding files like PDFs, Word, Excel, and PPT all happen inside one interface. Word, PPT, Excel, PDF, and EPUB files convert to Markdown locally via anydoc's WebAssembly, so nothing leaves the machine.
The editor also ships with built-in AI integrations for eight providers, including OpenAI, Claude, Gemini, DeepSeek, Kimi, Qwen, Zhipu, and Ollama, for in-place polishing, translation, and proofreading. A Git panel handles diffs, history, branches, stashes, and cloning, and can generate commit messages with AI. A built-in terminal supports multi-tab sessions and full-screen TUI programs. PDF markup is stored locally and travels with the file, and reading plus note-taking happens side by side in the same window.
This is not a plugin soup; the point is to lower friction enough that you actually commit, annotate, and convert instead of avoiding those tasks. The author explicitly positions it as a proof that when you're both the user and the author, AI collapses the barrier to building your own tools. The project is available for Windows from GitHub releases, and the code is open for others to fork or study.
When the person building a tool is also its daily user, the tolerances for friction are much sharper. XJ-Editor's feature set is the result of an author who was annoyed by exactly the switching costs that other editors ignore.
AI-generated code shifts the developer's job from implementation to requirement articulation and testing. The author's role as a product manager suggests that clear, iterative feedback is now the bottleneck, not syntax.
Lowering the friction of committing does change behavior. The author's claim that he actually commits now because the panel removes the terminal step is a lesson for any tooling: reducing cost of a frequent action is more effective than adding power.