Agents & autonomy
How agents plan, act in real environments, and recover from their own mistakes without losing the thread of what they were asked to do.
An applied AI research lab
We build machines that reason, remember, and act —
then prove it in the open.
Parallel Universe is a small, independent lab working at the edge of applied AI. We research agents, memory, and human–AI interaction, and validate it by building products people actually use.
- Founded
- 2025
- Base
- Remote
- Live products
- 3
- Focus
- Agents · Memory · Voice
01
Research
Three questions sit underneath everything we make.
Memory & continuity
Systems that remember across days and conversations — so an assistant feels like the same one you spoke to yesterday, not a stranger every time.
Human–AI interaction
The texture of working alongside an AI: when it should speak, how it should ask, and what makes it trustworthy rather than uncanny.
02
Work
Each product is an experiment run in public, with real people.
- ↗
Aria
A proactive voice companion that calls you when it matters and remembers what you said last week. Research into long-horizon memory and warm, naturalistic conversation.
- ↗
Kwality
A decision layer for the physical economy — how much of a thing should exist, where, and when. Starting with food. Forecasting and grounded reasoning over real-world supply.
- ↗
Hired
An agent that scouts roles and prepares applications on your behalf. Goal-directed agents that act reliably on a person's stated intent.
Computer-use agents
Agents that perceive and operate real software interfaces the way a person would — the substrate that makes the products above possible.
03
Principles
How we work.
01
Research first, shipped second.
We start from a hard question, not a feature list. A product ships when the underlying research is sound enough to stand behind.
02
Built to be trusted.
Our systems act on what a person explicitly asks for, within clear limits. We design for transparency and consent, and steer away from anything that deceives or runs without a human in the loop.
03
Small lab, real surface area.
Founder-led, several live products, one shared research core. The breadth is the point — it is how we learn what generalizes.
04
We work with a small number of partners. By introduction.
Access to the lab's work happens through people we already know. If our paths are meant to cross, they will.