I came across this tweet: https://x.com/guohao_li/status/1727641559259029921?s=46. It seems this guy came up with the same idea, did experiments, published the paper on NeurlIPS 2023, all of which were way ahead of AutoGen. There's also hints on plagiarism/academic misconduct. If so, huge shame on MSFT!! What do y'all think?