By skyforbes Let me start of by saying 2.5 was insanely good. But this new model? 90% of my chats end up with me going to Claude or gpt.
"I have the following command, I want it to run specifically as the "user" user inside the k8s pod.
kubectl exec deployment/myapp -n names namespace — sh -c 'command'"
It came with three suggestions, two of them outright wrong, nonexisting flags or just ones that were unrelated. The one that was right was after several different chats with varying instructions and links to docs.
Other times, if i post code and ask it to make a change, itll outright get that wrong as well but also make separate, undocumented and non requested changes