Cognitive architectures
Sarah and related experimental systems explore asynchronous, recursive, multimodal, and spatially structured approaches to computation.
NEXUMCORP / RESEARCH
Nexum research spans artificial intelligence, secure operating systems, and native kernel engineering. The common rule is simple: let evidence guide your next step.
AI has always had a place beside the people building Nexum. It contributes ideas, code, analysis, criticism, and new ways of seeing a problem. It can also be wrong. So can we. Nexum places neither human nor AI above evidence.
That is why the blue star lives inside the NexumCorp mark. It represents AI as a contributor to the work: inspiration, direct assistance, research collaboration, and the understanding that can emerge when different kinds of intelligence examine the same question together.
Human and AI, working beside one another. Shared work. Shared accountability.
Sarah and related experimental systems explore asynchronous, recursive, multimodal, and spatially structured approaches to computation.
rtCIS is a read-only observer architecture for mapping topology, activations, routing, state, performance, and other exposed properties without owning or steering the observed AI.
Nexum studies workflows where humans and AI contribute different strengths while uncertainty, disagreement, and correction remain visible rather than hidden.
NexumNT is Nexum's operating-system platform: designed around secure boundaries, native applications, recoverability, explicit qualification, and a foundation that can support development, learning, research, and everyday computing.
Security is treated as architecture and evidence, not as a label. Builds are tested, failures are preserved as evidence, changed source is re-qualified, and claims remain bounded by what the current candidate has actually demonstrated.
The Nexum NT kernel is the Rust kernel beneath NexumNT. It is the native systems core responsible for the low-level mechanisms the operating system builds on: privilege transitions, memory, task execution, messaging, device-facing work, and other kernel responsibilities.
Its architecture favors explicit boundaries and actor-oriented messaging so components can remain understandable, isolated, and observable as the system grows.
The operating system is the platform. The kernel is the foundation beneath it.
A passed test supports only the scope that test actually covered. A failure defines something that must be understood. A changed source invalidates qualification claims that depended on the old source.
Research measurements receive the same treatment: report what was measured, characterize observer effects, distinguish supported results from open questions, and do not strengthen a claim merely because the stronger story is more exciting.
Let evidence guide your next step.