In basically not more than 2 years, AI coding assistance has gone from "....here's a junior developer, give him simple stuff to do", to "....here's a senior developer, try to reign him in and make sure he doesn't get unnecessarily complicated".
This week I keep having instances where chatGPT is delivering correct results to difficult dev problems but Grok is not, where I had the OPPOSITE experience a few weeks ago. I have a theory: both of these services, after you do "n" number of queries, quietly downgrade you to smaller & less expensive models, so your results aren't as good.
