As the solo founder of Dusk AI, an AI companion app, I set out to solve a common complaint: AI companions that forget interactions after a short period. To address this, I built Dusk AI around persistent memory, allowing the app to track users’ preferences, mood patterns, and inside jokes over time.
The feedback has been overwhelmingly positive, with users often expressing surprise that the app actually remembers their interactions. However, I’ve noticed that the users who stay longest aren’t necessarily the ones who care most about memory as a feature. Instead, they’re the ones who have stopped noticing it, and for whom the conversation has become continuous rather than episodic.
This has led me to wonder if I’ve been marketing the wrong thing. Perhaps what users actually want isn’t an AI that remembers, but rather the feeling that someone is paying attention, with memory being the underlying infrastructure that enables this. Another unexpected discovery is that a significant portion of my users are non-English speakers who switch between languages mid-conversation, and the app’s ability to follow their language pattern has been a key factor in their engagement.
As a solo developer, I’m constantly building and iterating, and I’m likely getting many things wrong. However, one thing that keeps coming back to me is the gap between what users request on feature lists and what actually makes them stay. So, I ask: what’s the thing that made you stay with a companion app? Was it a feature you could name, or something you only noticed in its absence?