Noah,
after I posted the pdf, I did indeed type up this piece of code as a python script. And tested it on a set of random numbers generated by python.
The code worked fine. Havn't tested it again to see if it returns the same split values as ArcMap symbol editor though.
However, I overlooked your original intention, classify an "image", and get those values into a reclassify routine.
I found that the script worked well with, say, several 100's of records. But it doesn't scale very well. All those mat1, mat2 lists which it has to shuffle through. Once these get large, it gets very large.
So what you might have to consider is to resample your slope raster using 1000 random points then use those values as an input into the Jenks classifier routine. Obviously converting the entire image (via raster to point or something), you will end up with too much data.
Cheers,
Neil