Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.canthus.org/llms.txt

Use this file to discover all available pages before exploring further.

​
Daily check-in model

The system supports four check-in tiers so users are never blocked from their tasks on low-capacity days.
Three ratings, one per axis using 1-5 circles.
AxisLow anchorHigh anchorMeasures
Body”Moving gently""Feeling strong”Physical capacity
Mind”Foggy today""Sharp and clear”Cognitive capacity
Mood”A hard one""Really good”Emotional state

​
Data quality across tiers

TierConfidence weightEWMA influencePool accuracy
Full1.0FullBest
Quick0.7ReducedGood
Momentum0.4MinimalApproximate
None0.1NegligibleConservative estimate
Ratings are relative to personal baseline. A 3 means “about my usual,” not an objective severity level.

​
The onboarding sequence

1

Condition selection

Broad chronic illness categories, not diagnosis-level claims.
2

Severity self-assessment

Initializes the coefficient starting range.
3

Optional Visible import

Optional pace-point import for calibration support.
4

Difficulty calibration

User rates 6 common tasks (1-5), producing initial coefficient estimate.
5

Reference day anchors

User marks what they completed yesterday to seed initial pool estimate.
6

Task surfacing preferences

User chooses how strict task visibility should be when over pool.
  • Standard tasks over pool: hide (default), show with caution, or always show
  • Scheduled tasks due today when over pool: show (default) or hide
  • Mid-day recalculation: dynamic hide/reveal as remaining mana changes (default) or keep morning snapshot fixed
7

LLM estimation opt-in

Explains on-device vs cloud estimation; default is off.
Task surfacing behavior is intentionally configurable and should be part of onboarding, not a hidden advanced setting. Users can revise these choices later in Settings.

​
Cold start pool and confidence

For the first week, pool blends reference estimate with observed behavior: 
Day 1:    pool = referenceEstimate x bias
Day 2-7:  pool = blend(referenceEstimate, realSpendingEwma, blendWeight)
Day 8+:   pool = realSpendingEwma x bias
DayConfidenceBiasPool protection
10.000.8020% below reference
10~0.30~0.8515% below spending
27~0.80~0.946% below spending
34~1.000.973% permanent margin
The blend weight increases linearly from day 2 to day 7, avoiding both overconfidence and week-long lock-in to a bad initial guess.