A New Trend: When I Need an App, I Build My Own with AI

I’ve been using my own dictation setup for more than a month. I’ve also built a screenshot editor in Rust and created CutEase to plan cuts for sheet materials. These projects have changed how I look for software: I now consider building a small tool with AI alongside the usual search for an existing app.

The news about ArtCraft brought this into focus. Its open-source creative apps include PhotoCraft and VectorCraft, which recreate familiar Photoshop and Illustrator workflows. PhotoCraft’s developers describe it as early alpha and explicitly say it isn’t yet a Photoshop replacement for daily professional work.

My projects are much smaller. I want a few useful actions arranged around the way I work, and AI has made me more willing to build them.

Finding the right app takes more than downloading it

Downloading software is usually the easy part. First I have to find the options, decide which ones look suitable, install them, and try them on the actual task. An app can look right on its website and still turn out to be awkward on my machine or require a different way of working.

Some annoyances only become obvious after a few days. Then I either work around them or start looking again. When I weigh up building something myself, that whole process belongs on the other side of the comparison: searching, trying, configuring, and getting used to someone else’s choices.

CutEase started with a shelving project

I needed a cutting layout for particle board for a shelving unit I was building. The requirements were straightforward: enter the sheet dimensions and the sizes of the parts, then see how they fit. I wanted a simple tool without registration or subscriptions, but after an evening of searching, I decided it would be easier to write my own.

That became CutEase, a cutting layout tool that runs in the browser. I wrote about the project in my article on building CutEase over a weekend. It began with my own pile of measurements and a practical job to do; making it available to other people came out of solving that job.

My dictation tool connects existing software

For voice input, I wanted to use my chosen speech recognition model with the application I already had. Vocalinux handles recording and inserting the resulting text. I built a small API server, a service that passes the audio to the recognition engine and returns the transcript, to connect those parts.

AI helped me understand the components and make them work together. I didn’t have to build a recording interface or a speech recognition engine from scratch. The Vocalinux setup I described earlier has now been part of my routine for more than a month, which matters more to me than a demo of generated code.

A screenshot editor can stay focused on screenshots

ImgEdtr, my screenshot editor in Rust, follows the same approach. I wanted to capture part of the screen, add an arrow or some text, crop the image, and copy the result. It uses the system’s screenshot capture and adds the editing actions I need around it.

With a personal tool, I can keep the scope small. I can build for my own desktop and leave out features that would make sense in a general image editor. A clearly defined task gives me something specific to describe to the AI and something concrete to check when I run the program.

Custom software feels like having clothes tailored

Choosing an app reminds me of buying clothes off the rack: find the closest fit, try it, and live with the parts that aren’t quite right. Building a personal tool feels more like having something tailored. The shortcut, the sequence of actions, the recognition model, and the way the result is copied can all follow my preferences.

Developers have always written scripts and utilities for themselves. What has changed for me is how many tasks feel worth doing this for. I can describe the behaviour, get a starting point, try it, and ask the AI to revise a particular part.

Owning the tool also means maintaining it

A successful first run doesn’t finish the job. I still need to check the output on real tasks, fix mistakes, and deal with anything that breaks after a system update. That work belongs in the comparison too.

I’m a developer, so reviewing code and investigating problems are familiar territory. My experience doesn’t establish that anyone can describe a complicated application and immediately get reliable software. For a small tool with a clear purpose, though, building my own has become an ordinary option.

I expect more people to have tools like these: small programs for dictation, screenshots, files, or calculations, shaped around their own habits. Some will stay on one computer; others, like CutEase, will find people with the same problem. Either outcome is useful.

I’d start with one task where existing software keeps getting in the way. Describe how it should work, build a small version, and use it long enough to find the rough edges. That’s the scale at which I’d try this approach: one real task and a tool I can check myself.