It Started with Slop
A few years ago as the AI revolution (or not) took hold, people started espousing the idea that anyone would be able to write their own software just how they want it. Every little thing would be customized just how they want and it would unlock a new era of software design.
What starting actually looked like was NextJS apps spammed to all corners of the internet with the same design and a bunch of poor implementation. The app store exploded with bare-bones apps that never left the ground and all looked like the same collection of SwiftUI. But models got better, as there was so much economic incentive to do. A lot of engineers took the wrong signal and didn't take a real inner look to decide what they did or didn't like about something and what they actually wanted out of their toolbox.
Toast!
After coming back from a grueling paternity leave, some spark in me realized all the IDEs that I had used over time, I didn't really use their whole feature set. I didn't use the git version control. I didn't use the built-in terminal. I didn't, I didn't I didn't. And over my leave, AI coding models got good. Like really good.
So I paid for a $20 claude subscription and broke into what I wanted. An "IDE" that had things like syntax highlighting, file icons, lsp support, a file tree. And no matter how I fiddled with vim, it was never simple and elegant to make it do what I wanted. And I would be its customer. Because it was the experience I wanted. I've worked on it, added new features, gotten my first open source issue request! And now I'm sitting here, writing this blog post using a (what I consider) fully featured IDE I built to write this post.
At the end of the day, I'm not really concerned if I get users (though I'd love if I did and if this filled a niche that people wanted). I'm happy with the software I've built and find it fulfilling to really listen to my gut to intuit what I want in this tool I will probably use at work for a good chunk of my career. Because I'll likely always be looking at files and text on my computer. And if I am, this is just the experience I want.
Different
The idea someone writes software to fill the niche isn't new. Casey Liss is doing it with CallSheet because IMDB sucks butt. The difference is that I'm doing it with two twins, a house to keep up with, and a full time job. Building the software I want right now wouldn't be possible without LLMs because the time commitment wouldn't be a priority. But if I can have claude/gpt-5/deepseek-flash work on a problem while I step a way to be with my family. Or even better, if it's made available so I can text it from the couch.
I don't think everyone will suddenly be interested in designing their own software, mostly because they're not interested or whats out their fills that niche. I hope other people use it because even amongst software developers, folks may not be interested in spending their time or money on pursuing this idea. But they may be interested in picking it up to try.
Anyway
Try out toast. I think its wonderful software that I'm trying to improve as often as I can with care.
