Guth
NewsDocs
Page actions
View as Markdown ↗
Explore MCP tools ↗

instrument mechanism / verified-clean

Menstrual Cycle Length Tracker/Predictor mechanism

The inspectable method record for Menstrual Cycle Length Tracker/Predictor, kept separate from implementation and qualification.

Status: VERIFIED_CLEANLifecycle: verified-clean

Model scope

NEEDS_REVIEW because the exact ambiguity is: cycle length variability is named as an output but has no formula, and fractional average lengths have no specified date conversion or rounding rule. The honest-label requirement is present and should remain visible on every prediction panel.

  • The formula for cycle length variability is missing, so that output cannot be verified or computed.
  • The average cycle length may be fractional, but no rule specifies how to convert a fractional number of days into a date-only predicted start date.
  • The prediction formula is a simple arithmetic estimate, not a generally reliable physiological prediction; irregular cycles, missing logs, and nonrepresentative history are not addressed.
  • No minimum history length or validation rule is specified. At least two chronologically ordered, distinct period-start dates are needed to calculate a cycle length.

Units and conventions

  • Input dates are valid ISO 8601 calendar dates in the proleptic Gregorian calendar.
  • Cycle length is the date difference in calendar days between consecutive period starts; the start day is not counted inclusively.
  • The average cycle length is measured in days.
  • Date arithmetic is timezone-independent and ignores time-of-day and daylight-saving transitions.
  • A predicted start is expected to be a calendar date; the specification does not define rounding or timestamp behavior for fractional-day predictions.
  • Cycle-length variability has no defined unit or statistical method because no formula is supplied.

Inspectable vector

One of 4 source-bound review vectors:

{
  "inputs": {
    "history_of_period_start_dates": [
      "2025-01-01",
      "2025-01-29",
      "2025-02-26",
      "2025-03-26"
    ]
  },
  "expected_outputs": {
    "predicted_next_start": "2025-04-23",
    "cycle_length_variability": null
  },
  "working": "Differences: 2025-01-29 − 2025-01-01 = 28 days; 2025-02-26 − 2025-01-29 = 28 days; 2025-03-26 − 2025-02-26 = 28 days. Average = (28 + 28 + 28) / 3 = 84 / 3 = 28 days. Next start = 2025-03-26 + 28 days = 2025-04-23. Variability = undefined because no variability formula is provided."
}

Failure boundary

BREAK_CLAIMED · SELF_FLAGGED

The specification permits or fails to define behavior for invalid and undefined inputs: a zero-length cycle produces a raw prediction unless explicitly rejected, one start date causes a 0/0 average, and cycle-length variability has no formula. Date-counting convention is also unspecified and can shift the prediction by one day.

Other records

Sources and evidence

These pointers bind the page to evidence. They do not grant authority or collapse verification into deployment.

  • VERIFIED_LEDGER.json
  • evidence-tranche4/tranche4-results.json
  • BOUNDARY_HAZARD_CATALOG.json