I am currently building Deepreed. I do a lot of writing in my day to day work and I work with LLMs, but I face two issues: 1. The LLMs generate a lot of text, so reading it all is hectic, and you can easily pass on mistakes. 2. LLMs don’t preserve your style, so the text doesn’t sound like you. I am solving both by constraining the LLM to do targeted edits on the document, this way you most of the writing stays in your style, you understand it but also use the strengths of LLMs in your work. See a demo at https://deepreed.com/demo-v1. Reach out to me at [email protected] (Lam)