Biggest thing for me is that the new 3.5 apparently plays competent chess - at a high amateur level - iff you prompt it just right. I would not have expected that, considering how Anarchy Chess ChatGPT's normal play is. Once again demonstrates that you can never prove the absence of a skill.