Does anyone really like and enjoy LLM products with memory at this point? To me this seems to be a case where the technical ability to do memory vastly exceeds its actual usefulness (for most people).
Absolutely, to the point where I don’t know if I’d use AI tools at all without memory. It is critical for working on long-term projects whose complexity I don’t want to repeat every time I ask a question or explore an idea.
But I am using ChatGpt mostly as a way to flesh out ideas, question my assumptions, and do similar things, so YMMV.
In my experience, if I create a ChatGPT project or a Gemini Gem, it seems to remember everything I put in the instructions, which includes my tech stack, specific packages I'm using, preferences, and so on. ChatGPT remembers some of these things across chats that aren't within a project, but it's hit or miss. Gemini remembers absolutely nothing from one chat to another unless you're working in a Gem, but inside a Gem it's been very good at remembering quite a lot of stuff, particularly when using Gem 2.5. YMMV.
But I am using ChatGpt mostly as a way to flesh out ideas, question my assumptions, and do similar things, so YMMV.