I'm convinced that "completion" models are much more useful (and smart) than "chat" models, being able to provide more nuanced and original outputs. When gpt4 come out, text-davinci-003 would still provide better completions with the correct prompt. Of course this model was later replaced by gpt-3.5-turbo-instruct which is explored in this post.
I believe the reason why such models were later deprecated was "alignment".
I don't believe alignment/safety is the only reason. You also burn through significantly more output tokens in a back-and-forth editing session because by default it keeps repeating the entire code or document just to make one small change, and it also adds useless fluff around the text ("You are absolutely correct, and I apologize for the confusion...")