AI-native studio — two humans, many agents
Behind everymachine memory,an irreplaceable human.
Manifesto
Language models write the copy, clone the voice, render the avatar that never stumbles on camera.
When everything can be generated on demand, what is still worth making?
We think the answer is you. Not polished-you — the you with real opinions, that trips over a sentence and recovers. No model can replicate it, because it takes a real life behind it.
We are an AI-native studio: two people and a swarm of agents, building software for that irreplaceable self.
Work
Things we made
Studio
Two humans, many agents
Years building LLM products at leading AI companies in China, then went indie. Studied drama and film. The combination: understanding how AI works from the inside, and what authentic performance actually requires.
Ships nearly everything we build. A product manager by training who, with Codex and Claude Code in hand, became a one-person engineering org — the model of what an AI-native developer can do today. Relentless momentum: ideas turn into working software, fast.
How we work
Small team, agentic by default. We hand the mechanical work to agents and keep our hands on taste, judgment, and the parts that have to feel human.