What does a row on the desk tell me?
The dataset's name, where its data comes from, how many rows it holds and over what dates, and a cluster of pills counting the studies it has produced.
The row reads left to right, from what the dataset is to what it has found.
On the left: the name, the source word (computed, manual, imported, journal), and the instrument. Under them, the coverage line — how many rows, and the first and last date they span. If the dataset has fallen behind the sessions it should be holding, an amber note appears at the end of that line saying so; opening the dataset is what brings it current.
On the right, a cluster of pills. They all look the same on purpose — same shape, same size, colour doing the work — and every one of them explains itself if you hover:
- n keptStudies that are part of this dataset's report.
- n to judgeCandidates the edge scan filed that you have not yet kept or discarded.
- A · n and B · n — how many studies carry each grade. Only A and B appear; a C is a lead, not a finding, and does not earn a badge.
- n reversesThe one pill in red. Studies whose edge points the other way under a different but equally defensible definition of the outcome. It is red because it is the only count on the row that should change which study you open first.
- n weakensThe softer version: the edge survives but drops below the guard floor under another reading.
Past a hairline sit the two actions, kept visually separate so a count is never mistaken for a button: Log today and Delete.
