No idea, but this makes me wonder how we'd even go about getting training data. And it reminds me one time I built a poc editor[0] which saves deleted content s

No idea, but this makes me wonder how we'd even go about getting training data. And it reminds me one time I built a poc editor[0] which saves deleted content so you can get a similar view to manuscripts in which the author strikes out text. Maybe that would be an interesting starting point… [0]: https://github.com/trosh/nohide https://github.com/trosh/nohide