
The autonomy system
Autonomy is a system.
A machine that works on its own is five layers working as one. Machine Autonomy is building them to work together.
01 / Machines and hardware
Layer 1 of 5The body.
Airframes, actuators, grippers, on-board compute.

02 / Sensors and perception
Layer 2 of 5One picture of the world.
Cameras, lidar, radar and inertial units, fused and filtered into one picture of the world: where the machine is, what is around it, what is moving.

03 / Models and planning
Layer 3 of 5The reasoning a machine brings to the job.
Vision-language-action policies, planners, task logic.

04 / Control
Layer 4 of 5Turning a decision into motion.
Thousands of times a second.

05 / Runtime intelligence
Layer 5 of 5The layer that works while the job is running.
It learns a new case from one demonstration, lets confident actions through and holds the rest, keeps every rule on every action, and remembers what worked. It is where autonomy is finished.

06 / Where MA Reflex sits
Layer 5 of 5On top of what the machine already runs.
MA Reflex sits on top of the policy or controller the machine already runs. Your hardware, your sensors, your perception and your models stay as they are. MA Reflex adds what none of them carries: learning on the job, a gate on every action, lasting memory and a signed record.


07 / Machines made of software
The same system for machines made of software.
An AI agent has the same layers: a model, tools, memory, and a runtime. MA Duplex is that runtime. It sits in front of any frontier model, learns from every answer and governs every turn.

08 / The whole system
The whole system.
Autonomy is finished at runtime. Around it, we are building every layer of the system to work as one.


Start with the fifth layer.
The runtime goes on top of the machine, the sensors and the model you already have.