How CONTINUUM generates intelligence

The current engine: deterministic rules

Every intelligence output in CONTINUUM — the Demand Index, Capacity State, Adaptive Narrative, Agenda Intelligence, Weekly Intelligence, and Evening Debrief analysis — is produced by a deterministic rules engine. No machine-learning model, large language model (LLM), or probabilistic AI system is involved at any stage of the current intelligence pipeline.

Deterministic means: given identical inputs, the system produces identical outputs. Every output is fully traceable to a documented rule. There is no hidden learning, no opaque weighting, and no dependence on external AI APIs.

What the engine reads

• Calendar event metadata — timing, duration, domain classification, demand signals
• User profile fields — declared work window, capacity, peak energy pattern, timezone
• Self-reported debrief inputs — subjective energy, demand, and reflection entries

It reads no biometric data, no wearable data, and no external data source beyond Google Calendar.

Explainability

Every index and state CONTINUUM surfaces can be explained:

• The Demand Index is a weighted average of five rule-computed dimensions: Load, Intensity, Fragmentation, Variety, and Recovery Gap.
• Capacity State is a deterministic state machine with five outcomes: Open, Protected, Committed, Constrained, Recovering.
• The Adaptive Narrative is a template-based assembly — not generated text. Each sentence maps to a specific rule and input condition.
• Every output carries a confidence indicator derived from data completeness and sync freshness.

Explainability collapsibles are present in the Today cockpit for the Demand Index and Capacity State panels.

Confidence and uncertainty

Every intelligence output includes a confidence level (Low / Medium / High) computed from:

• Whether a calendar connection is active
• How recently the calendar was synced
• The fraction of events with confirmed domain classifications
• The volume of user classification corrections

When confidence is Low, the system states this explicitly. No output fabricates certainty from insufficient data.

Planned extension points

The intelligence architecture includes defined extension points for future AI-assisted modules. These are not yet active. If and when AI assistance is introduced, it will be subject to the following constraints — documented here in advance:

• Labelling: any AI-generated output will be labelled as AI-assisted and visually separated from deterministic index/state outputs.
• Isolation: AI modules may not write to or replace the Demand Index, Capacity State, or any deterministic output. They may only surface alongside them, in clearly labelled advisory panels.
• Explainability: AI outputs must include a source attribution, confidence indicator, and input summary.
• User override: advisory AI outputs may always be dismissed with no effect on core functionality.
• Audit: AI advisory requests will be logged in audit records without PII in the payload.

No AI extension is active at the time of this disclosure.

Not medical advice

Intelligence outputs are informational. They are not medical advice, clinical recommendations, or diagnostic assessments. See our Medical Disclaimer.