Habr Author Vibe-Codes Three Rust Utilities: DirMerge, ProcessTree and Notepad

The author of a Habr series has published a new batch of utilities written in Rust with the help of an LLM: DirMerge, ProcessTree and Notepad. There are three programs and one question: can a neural network help build tools that a person genuinely uses every day, without going back to the old software a week later? The author himself calls what is happening "schizophasia with vibe coding," so the self-irony is established from the start.

An important caveat: the announcement gives no model name, no code volume, no development time and no test results. Everything below relies on the stated setup of the experiment and the names of the utilities. The author's full conclusions are in the article on Habr.

What exactly is being tested

Most vibe-coding demos end at the "works on my machine" stage. Here the bar is higher: a utility has to displace a tool the author already uses. That is a harsh criterion. Users notice lag, clumsy keyboard handling, lost settings and odd behavior on large datasets. A simple app gets no forgiveness for small flaws, because people work in it for hours.

From the names, one can guess that DirMerge merges directories, ProcessTree shows processes as a tree, and Notepad replaces a text editor. These are classic tasks with mature solutions that have existed for a long time. The new tools will therefore have to be compared against very strong competitors.

Why Rust and why simple tools

Rust was not chosen for this experiment by accident. The compiler strictly checks memory ownership and types, so some of the mistakes LLMs typically make in dynamic languages are caught at build time. The result is a single executable without a heavy runtime. The flip side: models handle the borrow checker and lifetimes worse, and the "generate, fails to build, fix" cycle can take many iterations.

Simple utilities suit this kind of test for a different reason. They have a limited scope, well-understood behavior and an easy reference to compare against. If AI cannot pull off a text editor or a process tree viewer, it is too early to count on it for complex systems.

What is known about the experiment:

• language: Rust;
• method: vibe coding with an LLM;
• number of utilities in this installment: three (DirMerge, ProcessTree, Notepad);
• criterion: replacing tools the author uses in practice;
• main risk named by the author: unsolvable problems that force a return to the old software;
• the model, code volume and performance metrics are not disclosed in the announcement.

How to use this experience

If you want to repeat the experiment, start with a tool that is easy for you to verify. Pick a task with a clear reference, such as a folder comparison utility, and define acceptance criteria in advance: how many files it must process, which hotkeys are needed, what counts as a regression. Keep the old tool within reach for at least a couple of weeks.

Watch where exactly the model breaks. Typical trouble spots include handling paths and encodings across different operating systems, correctness on large directories, and clipboard and window focus behavior. These are the details that separate a prototype from a program you can actually use. The series is worth reading for these findings, not for the utilities themselves.