Weight space programming maybe related to continual learning

Weight space programming maybe related to continual learning in some way? Like self improvement as an extension of test time compute is kind of CL? (in the same way as updated ICL is kind of continual learning)