Rounded rectangles and round controls are now square across the app; only
status dots, glows, the sun, rings and the breathing circle stay round.
Cards, rows, thumbnails, featured cards, category tiles, the standing card
and the nav are larger. Home gets a quick To-do list kept on the device
(TodoItem in InternalMemory), seeded with sample items in mock mode.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KhQh614DMeYoQEH9SoMRUw
Routines: workout, meditation and must-do routines with timed, rep and
tick steps, sets and rest. A library screen, a builder with templates
and drag-to-reorder steps, and a full-screen player that auto-advances
timed phases, counts backgrounding and logs broken chains as partial.
Goals: a goal editor (deadline, colour, category, default class, archive),
routines linked to goals, and a goal field on tasks and routines. Tasks
added from inside a goal are now actually attached to it via parentId.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KhQh614DMeYoQEH9SoMRUw
A to-do app that doesn't believe you — an enforcement layer rather than a
neutral ledger.
Architecture ported from Autoreceptives/Frontend/Receptive: stacked MVVM with
the mandatory 4-file screen pattern, one ParentViewModel owning the loading /
network / error overlays and the handleError decision tree, one AppDataManager
gateway, dio comms carrying the three identity headers, secure storage with
random-suffixed keys, and a single-chokepoint Navigator. Package root and Dart
package name are both Grounded; org is nya.
The enforcement engine, one unit per formula in utils/:
- DebtEngine w(class) x severity(d) x decay(t), sublinear severity so old
misses cannot swamp the score; abandonment 2x with 30-day
decay immunity; late complete retains 30%
- StandingEngine Good -> Warned -> Grounded -> Lockdown, derived not set;
Grounded replaces home with the overdue queue
- CapacityEngine blocks over-scheduling against p50 of historically completed
minutes, with a learned per-category estimation multiplier
- IntegrityEngine session integrity, weekly volume, plyometric contact ceiling
and enforced recovery gaps
- ExcuseAnalyser on-device excuse clustering plus the confrontation copy
- GuardrailEngine distress detection and rationed amnesty
- ToneEngine all enforcement copy, so the tone cap lives in one place
CommitmentEvent is append-only and is the source of truth rather than the
status field, which is what makes honest history and excuse analysis possible.
Goals contain commitments via parentId, and a task can be run from a
full-screen runner that derives elapsed time from wall-clock so screen-off
cannot lose time. Backgrounding pauses the clock and is counted. The runner is
mirrored into an ongoing notification, with alarm-class full-screen intents
reserved for non-negotiables.
Design language, fonts, icon and native splash are in place; Mason bricks are
retargeted to this project and verified end-to-end.
flutter analyze lib/ reports no errors.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Mfu2gQLSFN21YRBcU2NrTt
