Good question! That experiment did not pan out like we’d hoped. I’m afraid there were a number of issues with that implementation that would have been a significant burden for our small team to solve.
Instead of trying to build it all ourselves (an expensive undertaking in terms of precious developer resources), we want to start with an existing, well-supported library that gets us most of the way there and lets our team focus on those finishing touches to make sure it feels great and fits in well with Discourse.