But let's zoom back a bit. When I started cargo-spellcheck, I figured it'd be "an easy thing" to do. Little did I know about the complexity of spellchecking acronym riddled source code documentation, interacting with the odd #[cfg(..)] flag.

This is me in 2021.

I was wrong. Very wrong.

It's not a fully solved problem still. I won't give you the whole AI spiel. For the particular tool I don't want to add any hardware requirements really. It should stay a small~ish helper tool anyone can run with an of-the-shelf laptop. Not a 10 billion parameter GPU neural network.

That said, there is a neural network embedded in the release binaries.

Now the real deal here is, it's still an ongoing effort across the board:

1. feature development
2. continuous integration maintenance
3. release cutting
4. code review and quality control

All of them driven by me or the people that want to see features to be implemented.

It's going to be three years soon since its inception. Featurewise, the project is close to completion. There are still issues specific to very very large codebases that have yet to be addressed and the potential for some buggy legacy C++ code to be replaced.

The Art of Maintenance

To be continued...