I haven't noticed any improvements, if anything, things have been taken away, or simply stopped working as expected.

IMO, it will only improve if they tie in the features of the Gemini AI and add functionality along with it.

IMHO, most companies are hesitant to tie modern generative AI/LLMs into products that have control over anything. So adding Gemini to Google assistant which can control your home, send messages and interact with you Google account by making calls and sending emails and similar things, is possible but companies are hesitant to deploy that; likely because of the resistance a lot of people appear to have to giving AI any control (or in many cases, any access at all), to their lives. They don't want a machine to decide what to do without being able to know exactly what it is doing and how. AI is seen by many as a kind of "loose cannon" so to speak. I understand that, but I don't necessarily believe that.

I believe that fear is why there hasn't been a larger rollout of an LLM to any major assistant type device.