Progress

Weekly Recap and Weekly Contract

Overview keeps “This week's contract” beside “Last week's recap”. A Weekly Contract is three targets you choose yourself: Protected days, repetitions, and focus time. Protected days are the union of days with real learning activity and declared Break days, with overlaps counted only once.

If the current week has no contract yet, use “Set this week's contract” once. After it is saved, this week's targets are fixed, so they cannot be reduced later to turn a missed target into a completed one. Put changes into “Next week's contract”; when the week changes, that stored row becomes the current contract automatically.

Declared Break days protect only the Protected days target. They do not add repetitions or focus time. If real learning also happened on a Break day, the day still counts once rather than being doubled.

“Last week's recap” is non-AI. It directly aggregates recorded first-study time, repetition time, repetition count, active days, new root records, and declared Break periods, then shows Active days, Break days, repetitions, focus time, records, and recall rate. “Change from the week before” shows only measured differences and does not generate reasons or causal claims.

Overview, Focus, and Memory

Overview runs top to bottom through this week's contract and last week's recap, continuity with Lifetime / Last 30 days totals, Milestones, and Journey. Focus contains the day / week / month / year focus-log views and source breakdowns. Memory contains Memory Lab's measured retention curve and stability distribution.

The Lifetime and Last 30 days switch inside Overview changes the summary range for focus time, active days, records, repetitions, recall rate, and lapses. It is separate from the day / week / month / year controls in Focus.

Current and longest streak

“Current” is the review-day run that is still connected now. “Longest” is the longest run in your stored history. Losing the current run does not erase the longest one.

A declared Break does not add a day to the run and does not end it. If today simply has not had a repetition yet, the current run is not treated as lost at that point.

Streaks are based on days with repetitions. The “Active days” statistic in the range cards also includes days that only have recorded first-study time.

Six accumulating numbers

Milestones and Town rewards

Milestones are derived from history CurveMora already stores, such as streaks, focus time, records, repetitions, and durable memories. Once a persistent Milestone is recorded, it does not disappear when the current streak later ends.

When a Milestone has a Town reward, the object is added to Memory Town automatically the first time the server confirms the achievement. There is no claim button and no random draw.

Journey

Journey is not another journal you need to maintain. It keeps a sparse timeline of durable events derived from your learning history, starting with Milestones.

The first version records Milestones only. It does not copy source titles, notes, or individual repetition content into Journey.

Build a Memory Town

In Memory Town, start with “Create city” to make a 16 × 16 city. If you own more than one, switch with “City” and use “Add another city” to create another.

Drag an Inventory thumbnail onto the city and a translucent ghost preview snaps to the nearest tile. Objects already on the map can be dragged the same way to move them.

You can still use the tap/click flow: select an Inventory or placed object, then choose its destination tile. A ghost position that overlaps another placement is not committed, and the server remains the final authority for placement validation.

Use “Rotate 90°” for rotatable objects and the visible Variant choices for objects with alternatives. “Store” returns an object to Inventory. “Undo” and “Redo” reverse or reapply placement, movement, rotation, Variant, and Store edits from the current editor session.

Memory Tree, Central Library, Clock Tower, Grand Observatory, and Infinite Library keep changing through four visual stages after you own them. Their stage is derived from existing repetitions, durable memories, records, focus time, and active days. There are no separate Town chores and no wall-clock waiting timer. If a stage advanced since the last time you opened Town, a quiet growth notice appears on your next visit.

The “DISTRICT COLLECTION” control at the lower left treats combinations of Town objects as districts. Owning the required objects changes a set to “Ready to arrange”; placing them close together in the same city forms Memory Grove, Study Quarter, Focus Square, Observatory Campus, or Endless Archive. Districts are derived from the current arrangement, so moving required objects apart dissolves the district. They do not add multipliers or Town maintenance tasks.

All five formed districts have quiet ambient identity: leaves and memory lights, warm lamps and book-like marks, clock rings and path lights, sparse stars and orbital arcs, or shelf lines and knowledge lights. Those details become slightly denser across the four growth stages of Memory Tree, Central Library, Clock Tower, Grand Observatory, and Infinite Library respectively. They disappear when the district dissolves, and continuous motion stops when the device requests reduced motion.

Memory Lab

Memory Lab builds measurements from your own repetition history. It begins with sample count, memories tracked, median stability, and recall rate.

With little data the screen says “Waiting for repetitions” or “Early signal”. As history grows it becomes “Established history”. This is a reading aid for the amount of evidence on the screen, not a statistical-significance test.

The first two charts

“Empirical retention curve” groups each repetition by the interval immediately before it: under 1 day, 1–3, 3–7, 7–14, 14–30, 30–60, and 60+ days. It shows the recall rate in each range and keeps n=sample count visible below every point. A range with no observations is left blank rather than drawn as 0%.

“Stability distribution” shows how many currently tracked memories fall into each estimated stability range. Stability is the scheduler's estimate of how long a memory is likely to hold.