Multimodal perception
Models that read images, video, audio and structured data as a single field — the foundation of any system that must act in the physical world.
X38 AI Labs builds general-purpose world models — systems that perceive, reason, and act beyond language. Our work spans applied research, infrastructure, and the products built upon them.
The dominant paradigm of the last decade treated intelligence as a function over text. It produced extraordinary instruments — assistants that converse, generate, and summarise. But text is a compression of reality, not reality itself. To act in the world a system must also perceive it.
X38 AI Labs is engineering the substrate for what comes next: foundation models grounded not only in language, but in vision, audio, structure, code, and the latent physics of the systems they interact with.
We treat intelligence as a research programme, not a product feature. Each of our five product surfaces is, in effect, a deployed experiment — a place where a hypothesis meets a paying user and is held to account.
The output of the laboratory is not a model. It is a generation of systems that perceive, reason, and act.
Every system we ship is a public claim about how intelligence should be built. The claim must be defensible.
General intelligence is an orchestration of specialised systems, not a single monolith. Architecture before scale.
A model only finishes training in production. Our products are the instrumentation by which the laboratory hears the world.
Models that read images, video, audio and structured data as a single field — the foundation of any system that must act in the physical world.
Specialised agents that plan, delegate, and self-correct in collaboration. The unit of computation is no longer a token; it is a decision.
Controllable models for image and video synthesis, with an emphasis on compositional editing rather than open-ended prompting.
Bringing foundation-scale reasoning to the systems organisations already run — databases, dashboards, customer flows, internal tools.
“The next foundation of AI will not be a larger language model. It will be a world model — a system that has been taught what reality looks like, how it behaves, and what it feels like to act inside it. That is the laboratory we are building.”