Not necessarily: transactional, impersonal directions to a machine to complete a task don't automatically imply, in my mind, the sorts of feedback loops necessary to induce AI psychosis.

All CASE tools, however, displace human skills, and all unused skills atrophy. I struggle to read code without syntax highlighting after decades of using it to replace my own ability to parse syntactic elements.

Perhaps the slow shift risk is to one of poor comprehension. Using LLMs for language comprehension tasks - summarising, producing boilerplate (text or code), and the like - I think shifts one's mindset to avoiding such tasks, eventually eroding the skills needed to do them. Not something one would notice per interaction, but that might result in a major change in behaviour.

Acceptance of vibe coding prompt-response answers from chatbots without understanding the underlying mechanisms comes to mind as akin to accepting the advice of a chatbot therapist without critically thinking about the response.