Kernel Space
Linux PREEMPT_RT, XTOS scheduling helpers, the BCI framework, and device drivers establish real-time boundaries.
Five modular layers connect devices, the real-time core, the data plane, and Super Intelligence research applications.
BCI analysis & feedback · medical research · edge AI inference
/dev/xtos0read / ioctl · Character-device accesssched_* / mlockScheduling & memory setupxtos_bci_core / xtos_bci_bufferopenbci_cytonPREEMPT_RT / sched_* / mlockxtos_schedApplications implement inference and feedback, not an in-kernel model or a verified hardware control loop.
Sample format and unit adaptation remain incomplete: the driver emits int32 nanovolts, but the SDK reads into float without conversion.
The real-time core, user space, and SDK each have a clear responsibility. Components can be built, verified, and evolved independently.
Linux PREEMPT_RT, XTOS scheduling helpers, the BCI framework, and device drivers establish real-time boundaries.
BusyBox, device-data pipelines, and signal tools provide a buildable, verifiable user-space environment.
Three independent libraries expose scheduling, signal processing, and device sessions to applications.
The two ends of the five-layer system: hardware provides input and applications produce inference and feedback. The layers between them keep interfaces open.
x86_64, ARM64, and RISC-V64 share the source and build-validation path. Real-device behavior needs separate measurements.
Applications use the SDK to read signals and process events; inference, feedback, and control remain application responsibilities.
Data follows one path with inspectable boundaries for input, processing, and response.
Receive input from devices, sensors, or existing data streams.
Keep data traceable through buffering, parsing, and signal processing.
Use the SDK to deliver data to applications for feedback and action.