Three tiers of validation.

Does every output field exist? Are all sections populated? Is the model output structurally complete? This is the baseline - the model must produce all expected outputs before deeper validation is meaningful.

Does the reasoning contradict itself? Do conclusions follow from inputs? This tier verifies that every number in the model is internally coherent, the weighted sum actually produces the OWI, sub-index scores follow from their components, trajectory matches the computed ratio, and no formula is silently violated.

Is the model correct relative to reality? Does it track against ground truth? Does it outperform a naive baseline? Are the feedback loops working? This is the hardest tier - it requires external reference points and cannot be satisfied by internal checks alone.

Each sub-index score = live data components average + milestone bonus + news signal. Every live component is sourced from a named public data series. If a component has no live data, it contributes 0 rather than a fabricated value.

The OWI is a model-backed system with an AI interpretation layer. Real economic data forms the base score. Claude scores each headline −5 to +5 for its relevance to the optional work prediction, nudging each sub-index by up to ±5 points. This section shows exact prompt sent to Claude.

You are the OWI scoring engine for optionalwork.com, tracking Elon Musk's prediction that AI/robotics will make work optional within 10-20 years.

TASK 1 - Score each headline -5 to +5 for OWI impact:
+5 = Major acceleration (AGI achieved, mass robot deployment, UBI enacted)
+3 = Moderate positive signal
+1 = Weak positive
 0 = Neutral
-1 = Minor setback
-3 = Moderate setback
-5 = Strong negative (ban, major failure)

Categories: ai_capability | robotics | economic_abundance | labor_displacement | wealth_distribution

TASK 2 - Detect if any headline CONFIRMS a watched target (high confidence only):
WATCHED TARGETS:
[live target list injected at runtime - 241 targets across 5 categories]

Return ONLY valid JSON, no markdown:
{
  "scored_headlines": [{"headline":"...","score":3,"category":"ai_capability","reasoning":"...","url":"..."}],
  "composite_signal": 1.4,
  "signal_summary": "2-3 sentence summary of today's OWI signal",
  "milestone_detections": [{"target_label":"...","category":"...","source_name":"...","source_url":"...","date":"YYYY-MM-DD","confidence":"high"}]
}

Confirmed milestones contribute a log-scaled bonus to each sub-index. Every confirmed event requires a source URL - no unverified claims affect the score.

Multiple output analyses to characterize whether the OWI behaves as a constrained model (bounded, deterministic, rule-enforced) or as structured reasoning with checks (logic gates, consistency enforcement, interpretive layers). A well-designed index should exhibit both properties at different layers.

Four deterministic test scenarios with hand-calculable expected outputs. Each scenario uses fixed, simple inputs so the result can be verified with a calculator before running the model. If the model produces the correct output for all four scenarios, the formula implementation is confirmed correct.

Enter raw data values directly and run the OWI formula manually. Every field maps to a real data series. Use actual published values from the sources linked below each field to verify the model produces the expected output.

The transition to optional work is a complex adaptive system - not a linear progression. This simulation models the five OWI stocks with their reinforcing (R) and balancing (B) feedback loops using Euler integration over a 20-year horizon. Adjust the parameters to test how sensitive the prediction is to initial conditions and loop strengths.