The Efferent approach

Coordinate the whole machine.
Preserve trusted control.

Prediction is not permission.

Efferent is a logical coordination layer above existing controllers. It helps choose among customer-defined allowed responses without taking over the safety logic, workflows, or operators that remain authoritative.

Logical architecture

Above the controllers.
Inside the boundary.

The architecture separates whole-machine evaluation from trusted execution. Customers define the response set and authority mode; Efferent coordinates within them.

One logical decision layer above trusted controllers

Machine contextState, constraints, operator intent, and allowed responses
Efferent · logical coordinationEvaluate the whole-machine tradeoff and choose only within granted authority
Trusted executionExisting controllers, safety logic, approved workflows, and operator authority
ObserveRecommendSelect when explicitly authorizedNo silent escalation
This is a logical architecture, not a claim about a specific hardware or integration topology.
Execution boundaryExisting controllers and safety logic execute; operator authority and approved workflows remain authoritative.
Authority modes

Capability may expand.
Authority does not drift.

The same coordination logic can support different operating roles. A change in role is an explicit customer decision, not a silent software behavior.
01
Observe

Make the tradeoff visible.

Efferent can evaluate the condition and preserve a reviewable record without recommending or selecting an action.
02
Recommend

Present a bounded response.

Efferent can recommend from the allowed response set while an operator or approved workflow retains the decision.
03
Select

Act only when explicitly granted.

Efferent can select from customer-defined responses only within explicitly granted authority; trusted controllers still execute.
Authority ruleNo silent escalation. A broader role requires explicit approval and a separate evidence gate.
Decision path

From changing condition
to reviewable outcome.

The loop keeps context, authority, execution, and evidence distinct. That separation makes it possible to understand what Efferent decided, what the trusted controls did, and what was later observed.
01

Observe

Read machine state, constraints, operator intent, and the available response set.

02

Evaluate

Compare the whole-machine consequences of customer-defined allowed responses.

03

Authorize

Apply the authority mode and approval rules already in force.

04

Execute

Pass the approved choice to the trusted controls responsible for physical action.

05

Record

Keep the condition, boundary, decision, and outcome connected for review.

Return pathRecorded outcomes inform the next decision; they do not retroactively prove the prior decision was correct.
What stays with the customer

Coordination is added.
Control is preserved.

The Efferent layer is bounded by the machine, controls, and operating authority already in place.
01
Controls

Local control laws

The controllers responsible for subsystem execution remain in place.
02
Safety

Safety logic

Interlocks, limits, and shutdown behavior remain authoritative.
03
Operations

Operator authority

Human approval and escalation rules remain explicit.
04
Governance

Approved workflows

Customer-defined process controls determine what may proceed.
Evaluate the coordination gap

Start with one failure.
Define one bounded role.

A useful first discussion names the changing condition, the controllers involved, the allowed responses, and the proof required before anything expands.

Discuss one coordination failureBegin with the system you have.