Thank you for the kind words! Two main reasons for our approach:

1) JavaScript-based apps these days have complex rendering logic that makes the HTML-parser method to find + translate strings unfeasible. Every company we've worked at has needed to extract each string and wrap them with a special `translate` function in their codebase.

2) We make heavy use of the Babel and TypeScript compilers to work with JavaScript ASTs, and there's been huge progress on those recently.

We've thought about NLP, but quality is a huge concern of ours, and we're not quite ready yet to roll that out to companies. If that's something you're interested in, would love to chat, send me an email at HN username + gmail!