Module capacity 

What a day can hold, learned from your own days (docs/capacity.md).

• Heaviness of a task or an event is its highest cost (Demands::level); unrated, its word (level_of).
• Ratings for the plan: said after the task when you said so (the median of the last five of the same item, else of its kind), else forecast (FeltIndex). Ratings given after are more accurate, anxiety is over-predicted beforehand (capacity-budget.md, criteria 6–7; G6).
• Load of an item: rating × hours, at least half an hour each (session-RPE, Foster 2001; criteria 10–11), per cost and in total.
• Time: your estimate corrected by your own record, the log of time spent over the first guess on finished tasks, recent ones weighing more, pulled toward 1.1× until about nine tasks stand behind it, per area pulled toward yours (Ratios; time-estimation.md, TE7–15); the day’s margin held as free time at the 85th percentile of the day’s total, by a seeded simulation with a shared day effect (slack; TE16–20).
• Budgets, per cost and in total, learned from days whose outcome you gave (learn; criteria 13–25), replayed from the record each time so that every device, holding the same record, plans the same.
• The gain minimum: good hours on good days, their 25th percentile, never learned downward from a bad stretch (G8–G10).

Nothing here is ever shown as a number: the pages get words and reasons.

Structs

Change

A change a rule made to a budget.

DayBalance

A day’s balance, in words: each cost and the total against what the day holds, the gain against its minimum, and what the plan does with it.

DayRecord

One day of the record.

Draws 🔒

A small, seeded generator (SplitMix64): the same seed, the same draws, on every device.

FeltIndex

The felt ratings of every task, to plan from (G6): by item, a task’s folded title (a repeating task, or one made again each time), and by kind.

Finished

A finished task, as the ratio learns from it.

GainBalance

What gave back that day, in a word.

Learned

What the record says, as of a day.

Planning

What the plan needs from the record (plan::Settings::capacity). Its default changes nothing: no budget, no correction, no slack, no gain slot.

Proposal

What a task’s form shows faintly for its unsaid costs and gain (FeltIndex::proposal): nothing is kept until “Looks right” or a tap.

Rates

The ratings the plan uses for an item: each cost and the gain, possibly medians of several ratings, and its word for what is unrated.

Ratios

Your ratio of time spent to first guesses, and each kind’s.

Record

Everything the plan and the days’ balance need from your record.

ScaleBalance

A cost of a day, in a word.

Spread

A day’s spread: the day’s own effect, shared by its tasks, and each task’s.

Enums

Outcome

How a day went, as you said it (reviews): too much, about right, too empty.

Start

Where the budgets start, before your days say more (criterion 24).

Constants

AHEAD_DAYS

How far ahead events weigh on the plan’s days (as the plan’s room does).

AROUND_HEAVY

Days before and after a heavy event hold this share (criterion 34). A guess.

BLAME_DAYS

Days looked back to blame a bad day on a full one (criterion 20: 1–3). A guess.

BUDGET_HALF_LIFE

Days for a day’s weight to halve in the budgets’ mean (criterion 17).

CAP_DAYS

Days looked back for the heaviest day well tolerated (criterion 22).

CAP_FROM

A cost needs this many fine days with load on it before its cap applies. A guess.

DAY_PERCENTILE

The day’s total is planned to this percentile (TE17: 80–90). A guess.

DAY_SHARE

The share of a day’s spread that is the day itself: a bad day slows every task (TE18). A guess.

DRAWS

Draws of the day’s simulation.

DROP

By how much a budget falls at once (criterion 20). A guess.

FELT_MEDIAN_OF

How many felt ratings the median looks back over (G6). A guess.

FILL

Each day is planned to this share of each budget (criterion 27). A guess (CB-Q18).

FLARE

Bad days within seven that freeze rises (criterion 21).

FLOOR

A budget never falls below this share of its start. A guess.

FULL

A day this full, followed by a bad one, lowers the budget (criterion 20). A guess.

FULL_WEIGHT_DAYS

Days with an outcome before what is learned has full weight (criterion 24).

GAIN_DEFAULT

The gain minimum, in good hours: before data, and its bounds (G9–G11).

GAIN_FROM

Good days asked before the minimum is learned (G10).

GAIN_HIGHEST

GAIN_HOURS_CAP

A gain counts this long at most (G8). A guess.

GAIN_LOWEST

GAIN_PERCENTILE

The minimum is this percentile of good days’ good hours (G9). A guess.

GAIN_SLOT_MINUTES

Minutes kept each day for the gain slot after its costliest block (G11). A guess.

HEAVY_DAY

HEAVY_EVENT

An event this heavy (its highest cost) lightens the days around it.

HISTORY_DAYS

How far back the record is read to replay the budgets.

INFORMATIVE

A good day informs the budget from this share of it used (criterion 15a). A guess.

KIND_PULL

A kind’s ratio counts as much as yours from this many tasks (TE10: 4–11). A guess.

LEAST_MINUTES

Every item counts for at least this long, so a dreaded ten-minute call is not free (criterion 11). A guess to tune.

LIGHT_DAY

A load this share of the budget or less is a light day; above the budget, a heavy one. Guesses.

PRIOR_RATIO

Before your own record: tasks take about 1.1× the first guess (TE9: 1.0–1.2×).

PRIOR_SIGMA

The spread of the log ratio before your own record (TE9).

RATIO_HALF_LIFE

Days for a task’s weight to halve (TE8: 4–5).

RATIO_HORIZON

Finished tasks older than this are not looked at.

RISE

By how much a budget may rise in a week (criterion 19). A guess.

SCALES

The four costs, then their total, in this order everywhere.

SLACK_SHARE

The free time a day keeps for steps running long is at most this share of its room for tasks, so that it never empties a day, thin data included. A guess to tune (docs/capacity.md, “The day’s free time”).

STABLE_DAYS

Stable days asked before a rise, and between two changes (criterion 19).

STANDS_FROM

Your ratio stands on its own from this many tasks (TE9: 8–10). A guess.

TOTAL

The total’s place in a Load.

TYPED_WEIGHT

What a typed or unknown minute weighs against a timed one (TE15: little). A guess.

WINDOW_DEFAULT

The learning window, in days: its default and bounds (the person’s setting).

WINDOW_LEAST

WINDOW_MOST

WINSOR

One day’s load counts at most this much of the budget (criterion 23).

Z_DAY

The standard normal’s quantile at DAY_PERCENTILE (0.85).

Functions

add

Adds b into a.

around_heavy

The days before and after each heavy event (its highest cost 7 or more), from today on.

balance

The balance of record’s day against learned (as of that day), and what learning from it would change (after: as of the next day; None when not known).

day_records

The days from from to to, each with what it held and how it went:

days_between 🔒

finished

Each finished task with an estimate and time spent: its log ratio, all its sessions summed (TE14); the first estimate, else the current one (a task estimated before the first was kept). Dropped tasks never count.

gather

The record read and weighed as of now: the ratings said after tasks, the ratio of time spent to first guesses, the days before (their tasks, sessions, events and outcomes) with the budgets replayed over them, the events of the coming weeks; and what the plan is given (Planning). Events of the past are read only once an outcome was given; events reads those between two instants (agenda::occurrences, or a cache of it).

held_by_events

What each day’s events hold, by day: those that hold time (Occurrence::holds_time).

item_key 🔒

learn

The budgets and the gain minimum as of today, replayed day by day over records (oldest first; days before today only), from start, with window days of memory. The rules, in order, on each day with an outcome: a drop when it was bad after a full day; a freeze after a flare; the target from informative good days; a rise after a stable week (docs/capacity.md, “Budgets”).

level_of

A task’s heaviness: from its costs when any is rated, else its word.

median_of 🔒

midnight_of 🔒

outcomes_of

The days’ outcomes, as the reviews say them: the night’s mix, else the work review’s.

ratio_kind

What a task is for, as the ratio sorts tasks: its area, three kinds, always known (TE10: few kinds; docs/capacity.md, “Corrected estimates”).

ratios

The ratios from finished tasks, as of today.

reasons

Why today holds what it holds, in at most two lines: a budget lowered yesterday after a bad day that followed a full one; a heavy event the day before or after; a change against two weeks ago once it passes 30 % (criterion 48). Nothing when nothing changed the day.

rounded 🔒

seed

A seed from what the day holds: its date and its steps (FNV-1a).

slack

The free time a day keeps for its steps taking longer than their median: the 85th percentile of the day’s total less the sum of the medians, by a simulation in which a shared effect moves every step of the day together (TE16–18). effect and spread.day are today’s as known so far.

slack_estimate

The same free time as slack, by the Fenton–Wilkinson approximation of the same model (Fenton 1960; TE18): a lognormal matched to the mean and variance of the day’s total, the shared day effect included. Cheap, from two sums of the steps’ medians (sum = Σ m, squares = Σ m²): the plan checks with it, step by step, that a day holds its steps and their free time; the free time it keeps is then slack’s, on what it laid.

start_budget

The budgets to start from: a day of room minutes (the fullest day of your week) full of usual work (4 an hour), and two heavy hours (7 instead of 4), fitting under the 85 % the plan fills, times the start chosen. Each cost starts at the total: a cost of an item never passes its highest, so a cost’s budget binds only once learning has lowered it.

suggestion

An item you rated as giving back well (felt gain 6 or more, no felt cost above 3), rotated by day: what a gain slot may suggest (G19). Never one you did not rate yourself, never a social one by default (G21).

weighted_quantile 🔒

The weighted median of (value, weight).

winsorised 🔒

Far values pulled in to three scaled deviations of the median: real overruns stay large, a timer left running all night stops ruling the mean (TE13).

Type Aliases

Load

Rating × hours, per cost and in total.