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.
What users see and what stays internal
Visible
- Daily mana value
- Human transparency sentence
- Task cost on cards
- Essential-task behavior
- Pattern and stability framing
Hidden
- Personal coefficient value
- Confidence score
- EWMA internals
- Axis factor internals
- Raw system parameters
Internal metrics are intentionally hidden to reduce anxiety and false precision.
Design principles
Why Mana, Not Energy
Why Mana, Not Energy
“Energy” is clinical. “Mana” gives users a lighter, less stigmatizing way to talk about limits while still operating as a strict budgeting unit.
Why Relative Cost, Not METs
Why Relative Cost, Not METs
Relative effort framing is more honest than implying clinical precision from population MET tables that do not represent altered chronic-illness physiology.
Why Tasks Disappear Instead of Greying Out
Why Tasks Disappear Instead of Greying Out
Greyed tasks repeatedly present what cannot be done. Hidden non-doable tasks reduce cognitive burden and learned helplessness signals. Essential tasks are the safety exception.
Why the Pool Is Conservative
Why the Pool Is Conservative
Conservative bias (0.80-0.97) protects against over-allocation and good-day optimism error.
Why No Streaks
Why No Streaks
Streak mechanics punish normal variability and create failure framing in chronic conditions.
Why No Absence Penalty
Why No Absence Penalty
Absence is neutral. Return is handled with brief confidence moderation instead of retroactive decay.
Why the Orchard Matters
Why the Orchard Matters
Sustainable pacing requires protected space for meaningful activities, not only maintenance tasks.
Why Patterns, Not Trends
Why Patterns, Not Trends
Pattern framing is useful without implying moralized progress/decline narratives.
The One Exception to Softening
The One Exception to Softening
Bad states are not hidden. Honest visibility is required for trust and safety.
Why the User Is Never Locked Out
Why the User Is Never Locked Out
Tiered check-ins preserve autonomy: lower capacity yields lower data quality, not denied access.
Transparency sentence policy
| State | Sentence style |
|---|---|
| Low confidence | learning + conservative framing |
| Medium confidence | history-based with safety margin |
| High confidence | stable-fit framing |
| Post-PEM | recovery framing |
| Return after absence | gentle catch-up framing |
The product contract is not cosmetic. These rules constrain algorithm outputs, copy behavior, and UI states so user trust remains stable as the model evolves.
Capacity display rules (ENG-31)
Default: qualitative
By default the Home screen displays capacity as a qualitative tier viaManaOrbCluster. Tiers are derived from the remaining / allocated ratio:
| Tier | Ratio | Meaning |
|---|---|---|
| 1 | < 0.25 | Very low |
| 2 | 0.25 - 0.50 | Low-medium |
| 3 | 0.50 - 0.75 | Medium-high |
| 4 | >= 0.75 | Plenty |
Opt-in numeric display
Numeric mana (e.g. “34 mana remaining”) is available via feature flagff_show_numeric_mana. Controlled by numericManaEnabledProvider. When on, the hero card shows the numeric value alongside the orbs.
Low-capacity mode (ENG-33)
Threshold:remaining / allocated < 0.15 OR allocated <= 0.
Behaviour:
- Actionable Today collapses to at most 1 item. Priority: smallest-cost essential, else smallest-cost any.
- Gentle Upcoming hidden entirely.
- Section header copy: calm phrase (e.g. “Rest is on the list today.”).
- “Show full list” expander available if user wants to see more.
- No urgency indicators, no “blocked” language, no exclamation marks.
task_copy.dart.