You can use your own writing to create a font set for OCR to recognize off of. Will doing a better job off a data set trained off of your own writing.

I've done some minor OCR to scan my bank PDF statements. I use tesseract iirc, there are different training sets to use. You can build a training set from your own writing.

Nothing has ever beaten the Newton Messagepad for me for this (certainly the last models made, the 2100).

I really wish a modern Newton existed. Evernote started off quite Newton'y (Newton team members were on board) but I fell off quite early.

I found that using an iPad with the Pencil has been a great middle ground. The OCR in the various Notes apps (including stock) is able to recognize my horrendous writing (it’s really bad).