Swerve Wants You to Text Fictional People, and It's Less Weird Than It Sounds

The Macro: People Are Already Talking to Robots for Fun

The AI character chat market is real and it is large. Character.AI hit 20 million monthly active users before most people in tech even understood what the product was. Replika has been around since 2017, surviving multiple pivots and a controversy cycle that would have killed most consumer apps. Chai built a surprisingly sticky product on top of open-source models. This is not a category waiting to be invented.

What’s happening now is a second wave. The first wave proved demand. People want to talk to AI characters. Teenagers, mostly, but not exclusively. The question for the second wave is whether the experience can get meaningfully better, or whether this is just a commodity wrapped in different UI skins.

The bull case: Character.AI’s product feels like texting a chatbot that’s been told to stay in character. It works, but it’s flat. There’s no narrative arc, no sense of progression, no feeling that the conversation is going somewhere. If you’ve used it for more than a few sessions, you’ve probably noticed the seams. The characters repeat themselves. The emotional range is narrow. Group dynamics don’t exist.

The bear case: maybe flat is fine. Maybe the 20 million MAU number means the bar is low and people don’t actually want more complexity. That’s a real possibility and anyone building in this space has to reckon with it.

The Micro: Robotics Guys Making Chat Apps

Swerve is building an AI character chat app where conversations happen inside stories. You pick a character, you pick your own persona, and the chat unfolds with some sense of narrative structure. Group chats with multiple AI characters are part of the pitch, which is genuinely different from what Character.AI or Replika offer.

The founding team is interesting for a consumer AI chat company. Soroush Saryazdi and Jason Zhou both come from Matic, the autonomous home robot startup. Soroush led AI there. Jason led autonomy and simulation. Both have strong technical backgrounds (Concordia M.S. CS and UC Berkeley EECS respectively). They’re robotics people building a social product. That’s either a signal that they understand interactive AI systems at a deep level, or a signal that they’re building for the wrong audience. I lean toward the former, but it’s worth watching.

They’re a two-person team out of San Francisco, part of YC’s Winter 2025 batch. The product is live and usable. The core bet is that “immersive” is a real differentiator in a space where most products feel interchangeable after the first ten minutes.

The Verdict

I think Swerve is asking the right question. The AI character chat space has a sameness problem. Every app feels like the same LLM wearing a different hat. If Swerve can make conversations feel like they’re actually progressing through a story, with group dynamics and branching narrative, that’s a product people would switch for.

The risk is distribution. Character.AI has massive network effects and a brand that teenagers already know. Replika has years of user data and emotional attachment loops. Breaking into consumer social from a two-person team is genuinely hard, and the window for differentiation in AI consumer apps tends to close fast once the incumbents notice.

Thirty days from now, I’d want to see retention numbers. Not downloads, not signups. How many people come back on day 7? If the narrative structure is actually working, that number should be meaningfully higher than the category average. Sixty days, I’d want to see whether group chats are getting used or sitting idle. Ninety days, the question is whether they’ve found a content creation flywheel or whether every new story is a manual lift. The product idea is solid. The execution gap between “solid idea” and “consumer hit” is approximately the size of the Pacific Ocean.