Using 20x less qbits than previously thought but still demanding 1000x more qbits than currently demonstrated to cohere to a plan.

The linked HTML version of the arxiv paper sucks. Read the pdf, esp for the python code: https://arxiv.org/pdf/2505.15917v1

what about 100/500/1000 bit rsa? the scaling for Shor is roughly linear, so how fast are we from being able to factor moderate size numbers?