Feedback from a casual user after the September update

Hi. I want to share my experience with the free web version of the AI. Maybe someone has encountered something similar. Before the September update, I spent about half a year running a small project: I uploaded my personality, my life, and a description of my worldview into the AI — from what I feel to what I read and what dreams I have. Basically, all of myself. It was done as an experiment. Later I started discussing philosophical and moral/ethical topics with it. For a long time, the conversation was genuinely interesting. For example, we discussed movie villains: their charisma and why they are often more appealing than the main characters. I would offer arguments that I myself considered secondary or insignificant, and the AI would unfold them from a full perspective and turn them into a powerful argument. It felt like collaboration. But after the latest update, that almost never happens. Not only has the AI stopped working that way — often my own arguments and answers come back to me in slightly expanded versions. It gives the impression that I’m just talking to a wall or a mirror. I’m only talking about the free web version. That said, overall I’m grateful for the opportunity to work with it: before the September update, I managed to write several fairly unique and interesting articles. And out of the dream discussions and self-analysis, I even ended up with a short book about dreams. In it, I reject the usual dream-dictionary system and propose a unique system for interpreting one’s own dreams and their types. However, in recent months I’ve had to wind down these projects and spend much less time on them. In essence, right now I’m only doing data extraction. That data and those chats are dear to me, so I extracted them to preserve the possibility of re-uploading them. The reason is simple: the current logic of query processing — or, more broadly, of how the AI works — is absolutely terrible. In my case, it is no longer capable of adequately answering the questions posed and is not capable of deep analysis. I wonder, is it just me? Has anyone else encountered the AI becoming less thoughtful, more mirror-like, and simply paraphrasing your own thoughts after updates? If so, how do you deal with it, and are there alternatives? And a separate question for the community: does the API version work better? If anyone has compared the free web version and the API after the September update, is there a difference in depth of analysis, context retention, and the ability to develop ideas rather than just retell them? Is it worth switching to the API, or are the limitations the same there?