2026Now
Releasing psychosis-guard
Released open-source safety middleware that tracks how risk builds over long LLM conversations.
Technology for people
Building technology for people, close to everyday life.
Vision
Make something people want
We believe technology begins with everyday life. Building products for Deaf and hard-of-hearing users taught us to understand and address real problems. We then researched psychosis-guard, open-source middleware for LLMs that gradually reinforce a user's delusional beliefs over long conversations, and are now extending that concern to safety for long-running AI Agents, from coding to everyday tasks. Hands-on experience building and operating products informs the search for trust that lasts beyond a single answer.
History
A journey that begins with everyday questions
From hands-on experience to new possibilities. Exploring a new product direction.
2026Now
A pivot toward a new product
Building on previous product experience to research safety for long-running AI Agents and explore a new product direction.
Founder
Namwoo Jang
Finding overlooked problems and building products people can actually use. Experience developing and operating accessibility products now informs research into safety for long-running AI Agents.
Explore the research