I recently spent a decent amount of time trying to implement a solution for something in my home network, and I was really weirded out when it started respondin

I recently spent a decent amount of time trying to implement a solution for something in my home network, and I was really weirded out when it started responding with: "Now thats real network engineer thinking!" "Now youre thinking like an advanced engineer!" or even a simpler "What an amazing question!" At first I did feel more confident in my questions and the solution paths I was leading it down, but after a few exchanges my trust in it's outputs went down rapidly. Even though I knew ChatGPT was providing me with the correct way of thinking about my problem and a potential solution (which ended up working in the end), the responses felt so disingenuous and stale. And the emojis... I used custom instructions (in chat) to combat this, but after another set of exchanges and when switching to a different problem and its context, it rewired itself again to be syncophantic. Im going to have to try global custom instructions and see if the issue of syncophanty persists