Not sure whether this is quite on topic for FoC, but I made a toy and I'm pleased with it.

I was inspired by Dasher, which uses the low entropy of natural language to make it possible to enter text using only a 1-dimensional input device. The core idea is to make the most likely next input take up a lot of space and thus make it easy to reach.

Most interactions we have with a computer involve low entropy input.

With (1) good models, and (2) a lossy encoding/decoding model that is easy for humans to use, you should be able to get away with entering a lot less.

For code, LLMs are pretty powerful models. And fuzzy matching is a pretty easy-to-use lossy encoding model.

So I made this: youtube.com/watch?v=ajCxCo8b9GQ