Absence Is Reported

"Three logged days in a month is a snapshot, not a trend line. Say so — never draw a confident curve through the silence."

Rule Two: Gaps Are Facts

The second analysis rule is quiet but load-bearing: absence is reported, not smoothed. Sparse weeks, unlogged sessions, a month with three crumbs — these are stated plainly, as what they are. The forbidden move is interpolation: drawing a smooth trend line through days that have no data, so a gap becomes an invented curve. That does not summarize the data; it manufactures it. And once manufactured, the reader cannot tell the invented part from the observed part — which quietly destroys the trust the citation rule worked to build.

Coverage Names the Empty Ranges

So the aggregate layer computes coverage explicitly, listing the empty ranges as contiguous date spans rather than papering over them. "Days 4 through 11 have no crumbs" is a first-class output, not a blank the chart hides. The invariant is blunt: missing data is reported as missing in any surface Dad reads as fact. A gap is information — it might mean he felt fine, or was traveling, or simply forgot — and flattening it into a line throws that information away while pretending to add some.

Missing data is reported as missing; interpolation across gaps is forbidden. Sparse is stated as sparse. A trend needs enough points to be a trend; below that, the honest output is 'a snapshot on these dates,' with the empty ranges named. Never let a chart imply continuity the crumbs do not have.

What the First Real Narration Did

The first analysis Forge ever ran over a real window had exactly this shape to handle: a short cluster of crumbs and then a long empty stretch. It did not draw a curve. It called the thirty-day gap what it was — "a snapshot taken on a single day, not a trend line" — and read only what the few points supported. That restraint is the rule working: the engine had every opportunity to look more impressive by inventing a trajectory, and it reported the silence instead. In a health log, the silence is often the most honest thing on the page.

Code

Honest coverage vs a manufactured trend·text

window: 30 days, crumbs on 3 of them (days 1, 2, 30)

# HONEST (rule 2): report the absence
coverage: 3/30 days logged
empty ranges: days 3-29 (no crumbs)
narration: "This is a snapshot on 3 dates, not a trend line.
            Days 3-29 have no data [reported, not inferred]."

# FORBIDDEN: interpolate the gap into a curve
narration: "Symptoms steadily improved over the month."
#   ^ draws a line through 27 empty days. Manufactured data.
#     The reader cannot tell the invented curve from real points.

Exercise

You have three logged crumbs across a 30-day window. Write two one-line summaries of it: one that honestly reports the sparsity, and one that smooths it into a confident trend. Point at exactly where the second one manufactures data the crumbs never contained.

Hint

The smoothed version implies something happened on the 27 empty days — a direction, a rate, a curve — that no crumb supports. The honest version names those days as empty and reads only the three real points.

Rule Two: Gaps Are Facts

Coverage Names the Empty Ranges

What the First Real Narration Did

Code

External links

Exercise

Progress

Comments 0