Vega & Enso — Rafe Johnson

Between client work I run a personal research thread on embodied AI — what it takes for an artificial companion to feel alive, and what it’s actually for. Vega came first: a spatial agent. Enso is its evolution — the same questions, reshaped into a product. Both remain living works in progress.

Vega — the spatial agent

Vega is a spatial companion prototype exploring the design possibilities and interaction dynamics of spatial agents — their interaction, their aesthetics, and the relationships people can build with them. It sits across 3D design, technical art, prototyping, and product design.

How human should the interaction feel — where’s the line between lifelike response and ethical boundary?
Does a consistent appearance strengthen the relationship, or does variability make the agent more useful?
Is the agent a tool or a collaborator — and what are the implications of displaying artificial emotion?
How does an agent’s identity personalise and evolve through interaction?
What are the real use cases — daily assistance, companionship, productivity — and do agents work alone or together?

I chose particles as Vega’s primary form. I wanted it to feel fluid and adaptive — a liquid form that can shape itself based on conversational context. Particles also give the agent interesting, consistent ways to express its states: talking, processing, listening. Voice and reasoning run on Eleven Labs and OpenAI APIs.

Fig. 01 — Vega, seaside intro

Fig. 02 — Voice commands

Early Vega concept sketches — Fig. 03 — Early sketches

Vega particle form exploration — Fig. 04 — Form exploration

Screenshot of the Vega prototype — Fig. 05 — Prototype capture

Enso — from agent to product

Enso carries Vega’s questions into an embodied AI companion built to help manage goals and focuses. It grew out of an interest in AI companions and the very real challenge of balancing multiple projects and life aspects at once.

At its core, Enso features:

A flexible goal and focus management system, visualised through interactive bubbles
An AI companion with awareness of all the content in the app
Continuous learning through conversations, building a deeper understanding of you
Personalised guidance that improves as your relationship with Enso develops
The ability to execute functions around the app based on dialogue

The concept is still at the early prototype stage, but the core features are built out and a first round of user testing is complete.

I wanted an interface that could hold both the high-level overview and the small reminders — something built on motion and adaptability rather than fixed lists and folders. The fluid bubble system won out over the structures of typical notes and goal apps.

The form of Enso matters just as much. Having Enso feel alive is important for interesting interaction — but avoiding the uncanny valley is crucial, and I deliberately avoid mimicking human behaviours to prevent unhealthy over-reliance. Users can customise Enso’s colour today; longer term, its appearance will evolve to reflect its understanding of you and your goals.

Fig. 06 — Enso

Fig. 07 — Day one

Fig. 08 — Form study

Fig. 09 — Session capture

Enso app interface — Fig. 10 — Interface still

Credits

Everything — Rafe Johnson