I'm building an AI companion around one simple idea: being remembered
I've always felt that sometimes you just want someone to talk to.
Not necessarily about something important. It could be something that happened at work, a random thought, a small frustration, or something you don't really feel comfortable sharing with anyone else.
I started trying AI companion products because of that.
The conversations could be surprisingly good. But I kept running into the same problem:
I'd tell the AI something important today, and the next day it felt like the conversation had never happened.
I'd mention an upcoming interview, something I was worried about, or a small habit of mine — and then have to explain it all over again.
That made me realize something:
Maybe the value of an AI companion isn't just how well it can talk to you. Maybe it's how well it remembers you.
So I started building Mylovian.
The idea is simple:
To remember you. To stay with you.
Instead of treating memory as just a way to make an AI answer questions more accurately, I'm trying to make memory part of the relationship itself.
For example:
Day 1
“I have a big interview tomorrow. I'm really nervous.”
The AI remembers the event and the emotion behind it.
Day 2
You come back and simply say:
“Hey”
Instead of starting from zero:
“Hey ❤️ Good to see you again. Wait... how did your interview go yesterday? I remember you were really nervous about it.”
That small interaction is what I'm trying to build around.
Not an AI that simply remembers facts.
An AI companion that remembers what matters.
What I'm building right now
Mylovian currently has:
· 3 official AI companion personalities
· Long-term memory
· Relationship event memory
· Follow-ups for important events
· Proactive interactions
· AI-generated companion photos
· Subscription plans
Each companion has their own personality and independent memory, so the goal is for the relationship with each character to develop differently over time.
The hypothesis I'm trying to validate
My biggest question isn't:
“Can people chat with an AI?”
We already know they can.
It's:
Does remembering someone's story make them more likely to come back — and eventually pay to keep that relationship going?
That's why I'm particularly interested in 30-day subscription retention, rather than just downloads or daily chat volume.
If people use Mylovian for a few days because it's novel, that's not enough.
I want to know if the feeling of:
“It remembers me.”
can become a reason to return.
I'm still early in the build, so I'm sharing this here because I'd love to learn from other founders.
If you've built a subscription product, AI product, or anything around retention, I'd especially love to hear your thoughts.
The retention hypothesis is really interesting.
I’m curious how you’re planning to get the first users though. With AI companion products being pretty crowded, I feel like positioning and the first acquisition channel could make a huge difference.
Are you planning to start with a specific niche of users, or go broad?
The interview example is the part that makes the memory idea click. Remembering the emotion and then asking about it the next day is very different from just storing facts. Mylovian feels like it could be pulling in a lot more organic traffic than it probably is right now, especially while you’re testing whether that feeling actually drives 30-day retention. Have you seen any early signs that the memory itself is what brings people back?
That’s a great question! To be honest, I’ve just finished the build and haven’t onboarded my first batch of users yet, so I don’t have the data to confirm that just yet.
Since you haven’t onboarded the first batch yet, I’d think about distribution before trying to judge the 30-day retention signal.
One thing worth testing alongside organic growth is targeted media placement around the “AI that remembers you” angle. The right features could put Mylovian in front of a relevant audience, while also building early authority, brand awareness and a Featured In footprint. Unlike running ads, a good placement can remain a long-term asset that keeps generating visibility, referral traffic and credibility after the initial launch.
If you’re interested, I have a few ideas for how I’d approach this specifically for Mylovian. Happy to share if you want to explore it.
The 30-day retention metric makes this much more interesting than simply measuring chat volume. How are you planning to determine whether memory itself is driving the return behavior rather than just being one feature inside a novel AI companion?
It’s definitely the biggest challenge in my current validation phase.
My plan to isolate "memory" as the key driver is two-fold:
Qualitative Feedback: I’m tracking specific "memory-triggered" interactions (like when a user reacts to the AI following up on a past event). If users explicitly mention these moments in feedback or if these specific sessions have higher engagement scores, it’s a strong signal.
Retention Cohorts: I’m looking at the correlation between "memory depth" (how much personal data the AI has successfully retained and reused) and the 30-day retention rate. If users with higher "memory density" stay significantly longer than those who just have casual, generic chats, it validates the hypothesis.
It’s still early, but I’m trying to move away from "novelty" and toward "utility through familiarity." Would love to hear if you’ve seen other products successfully measure a single-feature impact like this!
That’s a thoughtful way to approach it, especially separating memory-triggered moments from general usage. I’ve enjoyed this discussion — what’s the best email to reach you at?
I've really enjoyed our exchange as well!
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.
This comment was deleted a month ago