Domain bridge / lesson 04
Map a lifecycle without importing its domain
Observe, measure, evaluate, and decide has the same system shape as a neural feedback loop.
Know what the primitive owns.
Problem
Domain-specific lifecycles often appear to require a new framework even when their underlying transitions are generic.
System boundary
CMG provides the generic transform/trace/evaluation/feedback skeleton. It does not ship a harness, scheduler, goal, agent, or workflow.
State changes become visible.
Inspectable: Raw records, extracted totals and timings, measured success rate, threshold result, and feedback action.
Change the number of successful records and inspect the threshold response.
Static specimen / JavaScript adds controls without changing the lesson.
Same idea, honest surfaces.
const graph = createModelGraph({
id: "lifecycle-bridge",
name: "Lifecycle Bridge",
transforms: [signalExtraction, measuredState],
evaluator: thresholdEvaluator({ threshold: 0.8 }),
feedbackResolver: defaultFeedbackResolver(),
});
const run = await graph.run(rawData);graph = create_model_graph(
"lifecycle-bridge",
"Lifecycle Bridge",
[signal_extraction, measured_state],
evaluator=threshold_evaluator(threshold=0.8),
feedback_resolver=default_feedback_resolver(),
)
run = graph.run(raw_data)What this does not imply.
The analogy proves a reusable shape; it does not make a private harness part of the open-source library.