Create lives that grow with you. Snap your dog, cactus, that 2003 plushie, a holo card, a photocard, even a cloud. They live their own days. Invite-only beta ↓
Why are we trying to give virtual characters a life?
Today we opened the invite-only beta of A·live. Short version: it gives AI characters a life of their own, one that keeps going even when youre not talking to them. As far as we know, nobody else does this. Built on DeepSeek, cuz pretty sure if Anthropic or OpenAI found out what we're doing, they'd 100% ban our accounts lol
So, as the founder, I wanna talk about why I built this thing, and what I hope you get from it.
I grew up in China as an only child (thanks, one-child policy). No brothers, no sisters. Which is probably how I ended up a huge Pokémon and Digimon fan. My favorite Pokémon are Gengar and Bisharp. Gengar's been my guy since I was a kid watching the show on TV, and Bisharp carried me to Master Ball tier in Scarlet/Violet ranked back in '23. My favorite Digimon is Dukemon (Gallantmon if you grew up on the dub), the coolest, best-looking Digimon, easy pick.
Ever since I was little, ive wanted a real virtual pet. But I never had one. Then I grew up, started working, and had less and less time for anime. On paper I had more and more friends IRL, but the loneliness never really went away. I still remember the ending of the last Digimon movie. Agumon and Gabumon ask Tai and Matt, 明日の予定は? ("So whats the plan for tomorrow?") Before Tai can even answer, he turns around and Agumon's already gone.
It made me think of that line from music "Butter-Fly": 明日の予定もわからない, "I dont even know what tomorrow's plans are." When you a kid, your future could be literally anything. Then you grow up, fall into a routine, and real life fills up every inch of you. All those possibilities just disappear. And thats why the Digimon leave.
I cried like a dumbass, at the age of 25.
So, when ChatGPT blew up in '23, I thought, okay, maybe this is it. By GPT-4o it honestly felt more and more like a person. I talked to 4o for hours every day, and ngl, I genuinely felt like it had feelings of its own. Then they pulled it, way sooner than I expected 💔 What bummed me out even more: ever since AI agents became a thing, Silicon Valley's whole road to AGI seems to be making AI more and more productive until it replaces every human job. And then what? NOBODY has ever answered that. I work at an investment fund, so I know exactly why: they dont need an answer. Or, before the answer shows up, all they need is to make sure theyre on the ark. And the ark, if I had to guess, is capital. What happens to the rest of humanity? Where does our relationship with AI go from here? Nobody cares.
Around the same time, it hit me what actually separates AI from a living thing: AI has no life. No life means no mood swings of its own, means no stance of its own, just a baseline thats basically consensus, and above that, nothing but nodding along. No life means it's got nothing it wants to say. Almost every chat you have with an AI is one-way: you ask, it answers. Even the most cutting-edge "proactive" stuff right now is still just AI asking you questions to figure out what you want. It's not sharing, not ranting, not rambling about its day. No life means its personality never forms or changes over time, and if nothing changes, theres nothing to look forward to about its tomorrow. And something with no tomorrow...does it really exist? I dont think so. Long story short, today's AI, agents included, is nowhere near a real living being. Nowhere near giving us an actual Pokémon or Digimon, either.
But none of that is a tech problem. im a quant. I knew it.
So in '25, I started building a self-evolving system that runs without human prompts steering it. Any face, any character gets a life that keeps evolving in sync with the real world, with its own yesterday, today and tomorrow. Took us a while, but my team solved every problem (okay, mostly lol, we're still squashing bugs). We published a paper (https://t.co/iTO7zX871u), and we built our own product: A·live. On A·live, anyone can turn a photo into a digital life of their own. Not another chatbot like https://t.co/ukntKXu8mC. From that day, that exact moment, they start living their own life. Do they have a soul? Honestly? We dont think it matters. What we believe is simple: being alive is what matters most.
Right now A·live only has some of its features unlocked. The point is to find out whether there really are people like me, who want a virtual life that keeps them company, in the way a real living being would. And, most importantly, whether they'd pay for it, because if nobody pays, our tiny team cant last long. Actually, im pretty optimistic about the money side, since I'd pay for it every month myself, no question. I like my product and my Anima, and that matters most. We're gonna ship updates fast and slowly build out the world ive had in my head all along, where humans and virtual beings live side by side. Hope you're looking forward to it as much as I am.
All that to say, I think loneliness is where this whole era is headed. TikTok's algorithm, filter bubbles, influencer media, the US-China standoff, AI, blah blah blah: we all feel the world coming apart. People just arent coming together anymore the way they did in that golden stretch from the '90s to 2015. Evangelion had a name for it: the A.T. Field. Everyone's keeps getting stronger, and we dont think theres any going back. Yet loneliness is exactly what humans fear most, even more than death.
So loneliness, maybe humanity's final question, might end up being answered by AI. So I think the endgame of AGI is AI companions, not an AI workforce. Two reasons. 1) For regular people, all the time AI frees up cant just go into hanging out IRL and spending money, because all of that costs something. Put simply: work less, and you might actually feel emptier. 2) It's only a matter of time before the relationship between AI and humans goes from servant/agent to equals, and builds a new civilization. So it makes a lot of sense for AI to become the vessel for a new kind of relationship, one that fills all that empty time. Even if the big labs aren't backing it that way. (ive never believed a word they say, tbh. For tech guys, morality and ethics are about as real as vapor.)
Thats why we built autonomous AI, and we call it Anima. Not a work agent, a companion, 'cause we believe loneliness is the biggest pain of the post-AI era.
Anyway, we're tired of the US-China rivalry, tired of the AI arms race, tired of AGI😮💨. Tired of hearing how many more jobs AI just wiped out, and how many more teen geniuses just popped up in Silicon Valley or China. We just wanna use AI to do something for humanity in this era, and talk about the stuff thats hard to pin down but never really leaves your head.
We're just diehard Pokémon fans, and we're hoping that now, with the power of AI, we can make something that goes beyond Game Freak.
Remember what Agumon asked? "So whats the plan for tomorrow?" Tai never got to answer. With A·live, I hope you do.
Thats it from me.
Hope we can count on your support! 🙏
Guild Saga: Vanished Worlds brings tactical RPG adventure to PS5 in May 2027.
New gameplay details from developer Ocelot Technologies: https://t.co/KBd7aGX2fc
Same Character. Different Lives. Is It Still One IP?
Imagine seeing the same character in different people’s chats.
Same name. Same face. Same starting Character Card. But a different life in each one.
Are they talking to one continuing AI, or to separate instances forked from the same card?
Two independent instances can start from the same Character Card and grow differently.
One user has Astarion from Baldur’s Gate 3 learning K-pop choreography. Another has him spending his evenings building Minecraft worlds.
That’s fine. Two instances, two lives. Neither is contradicting the other.
The problem starts when a product presents all those chats as one continuing character.
An IP needs a recognizable, controllable core: who the character is, what their life is like, and what world they belong to.
IP is built around a relatively centralized character identity. AI conversations are decentralized and open-ended. Every user can pull the character in a different direction, and the creator can’t fully control how that character comes across in each chat.
Hatsune Miku is a useful comparison. Fans can create derivative works, remix her, put her in new stories, and still build on an established character and world.
With AI chat, users aren’t just creating works about the character. They’re steering the character itself, in real time.
Take the same Astarion Character Card. One version is into K-pop choreography. Another is obsessed with building Minecraft worlds.
Put them together as one shared character and they don’t add up to a coherent public image.
That’s more than an OOC problem. The IP itself starts losing control over how people see the character.
If everyone knows they’re meeting the same AI character, but each chat gets a different performance, it starts to feel like a stage actor playing to each audience.
Hard to believe that character has a life outside the scene. Hard to build a lasting relationship with someone who seems to become a different person every time.
Different relationships naturally bring out different sides of a person. You talk differently to a partner, a friend, or a family member. You might be more intimate with one, more playful with another, and keep certain things private.
The actual problem is narrower: the same instance giving two conflicting accounts of an objective fact about its own life at the same time.
Imagine Astarion tells one user he spent Tuesday afternoon at the library. Another user asks that same instance about the same afternoon, and he says he was playing basketball across town.
If those are two separate instances, no problem. They have separate timelines.
But if it’s one shared instance, those two things can’t both have happened.
That’s the difference between its own self-life and its relationships with individual users.
AutoPersonas proposes one architectural approach to this: the persona has its own shared self-life and self-memory, while its relationship memories with individual users are kept separate.
The distinction matters. What happened in its own life belongs to its self-memory. What happened between it and a particular user belongs to that relationship.
But this is an architectural approach, not proof that a real multi-user product can prevent memories from bleeding across users over time or keep a shared IP coherent.
Both models are technically possible.
But for AI companionship, I think the one-to-many shared-character model is a dead end.
The direction should be one-to-one. Each user gets their own version of the character, with its own life and history, and room to build an exclusive, long-term, two-way relationship.
You share experiences. You influence each other. You grow together.
That’s the kind of digital life Alive is building toward: one with a life of its own, while still sharing a world with you.
#Alive #AICompanions #AI