There is a new thread for Polymath5. It looks like Klas Markstrom is getting close to a proof that for a large enough N any sequence of length greater than one will have discrepancy at least 3. I think there is one last case that is being analyzed. If everything works out I think N will be around 2000.
A new Thread