Been diving into AI companion stuff lately — for both fun and research.There are dozens of OSS projects out there trying to build the next waifu simulator with local models, decent UIs, and even memory functions. Technically? Many of them are solid.
But let's be real... 95% of them die in a few weeks. Nobody sticks around. Nobody chats past day three.And I don’t think it’s because of the models. It’s the emotional UX that’s lacking.
If you’re working on a character memory system, emotional state tracker, or even just trying to make an LLM act a little clingy — I’d love to test it.
Let’s stop building bots that just “respond.” Start building ones that make you *want* to come back tomorrow, but it's rather hard.
But let's be real... 95% of them die in a few weeks. Nobody sticks around. Nobody chats past day three.And I don’t think it’s because of the models. It’s the emotional UX that’s lacking.
If you’re working on a character memory system, emotional state tracker, or even just trying to make an LLM act a little clingy — I’d love to test it.
Let’s stop building bots that just “respond.” Start building ones that make you *want* to come back tomorrow, but it's rather hard.
Last edited:
Emotional UX is underrated in OSS. Most focus on features, not feelings. Maybe it's time devs studied attachment theory, not just token limits