Project-Labs-x
ProjectLabs-x
Project-Labs-x — Ideas & Vision

CONCEPTS

Not features. Not products. The ideas that drive everything we build — published so the world can challenge, steal, and improve them.

SCROLL TO READ
01
Zero Interface
02
Distributed Mind
03
Temporal Canvas
04
Liquid Memory
05
Open Frontier
01 / 05

Zero Interface

"The best UI is no UI."

We are building toward a world where software disappears. AI infers intent from context — past behaviour, current state, ambient signals — and acts before the user reaches for a button. Not voice commands. Not prompts. Pure ambient intelligence.

Intent ModellingAmbient AIHCI
0%interaction reduction in early tests
01
02 / 05

Distributed Mind

"No single model should hold all the power."

Intelligence should be a mesh, not a monolith. Specialised micro-models that communicate via structured contracts, each owning a narrow domain, composing at runtime into something greater. No single point of failure. No single point of control.

Distributed SystemsFederated AIMicro-Models
0×faster inference via parallel mesh routing
02
03 / 05

Temporal Canvas

"Time is the dimension no tool has mastered."

Video, audio, and narrative are fundamentally temporal — yet every editor treats them as a flat timeline. We are designing tools that understand causality, pacing, rhythm, and story arc. The editor that thinks in scenes, not frames.

Video AITemporal ReasoningCreative Tools
0%narrative cut accuracy in Flux beta
03
04 / 05

Liquid Memory

"Context should flow, not reset."

Every AI conversation starting from zero is a design failure. We are working on persistent, privacy-preserving user context that flows across sessions, devices, and applications — remembering what matters, forgetting what doesn't, owned entirely by the user.

Memory SystemsPrivacyPersonalisation
0secaverage time lost re-contextualising per session
04
05 / 05

Open Frontier

"The most important research should be free."

Frontier AI locked behind paywalls and corporate APIs is a civilisational risk. Every model weight, every training run, every benchmark we produce is released openly. Not because it is easy, but because the alternative — concentrated AI power — is far worse.

Open SourceAI SafetyOpen Weights
0%of our research published openly
05

Our operating principle

Build in public.

Think in public.

Fail in public.

Read the researchChallenge an idea