The count and the evidence are different things

A weekly total looks like a simple measurement: place every new paper into a calendar interval and count the rows. In a research index, the visible number is also shaped by which sources answered, how far each query reached, and when a record entered an upstream database. A change in any of those conditions can move the count even when the underlying rate of research is unchanged.

Quantum Observatory therefore treats coverage as part of the result. A period is comparable only when the configured research sources report successful coverage extending through that period and the period itself is complete. Current weeks and weeks with a failed required comparison source do not receive a growth percentage. The absence of a percentage is useful information: it prevents a precise-looking claim that the available evidence cannot support.

Three dates answer three different questions

The first public date asks when a research work became available. For a linked preprint and journal article, the earliest explicit publication date anchors the work. The first-collected date asks when this observatory encountered a particular source record. An upstream indexed date records when a service such as Crossref updated its own metadata. These dates can be separated by days, months, or more.

Using first-collected dates is appropriate for an operational question such as how many records the collector discovered this week. It is not a substitute for publication activity. Using an indexed date can make an API query efficient, but the returned work must still be assigned by its publication date. The interface keeps first-public and first-collected views separate so an indexing burst is not presented as a publication surge.

A comparison checklist

Before interpreting a rise or decline, inspect the methodology and source status shown with the snapshot. Then verify that both periods use the same definition of a research work, the same time zone, and matching source coverage. Finally, look for late records or newly linked versions that may have changed a historical period after its first publication.

The order matters. Topic labels or individual papers can explain a change only after the measurement is comparable. Starting with an appealing paper and working backward risks turning a data-quality difference into a narrative about the field.

  • Are both periods complete Monday-to-Monday intervals in Korea Standard Time?
  • Did every source needed for the comparison succeed with adequate coverage?
  • Are linked arXiv and journal versions counted once in both periods?
  • Did the older period change because a delayed record or explicit identifier was added?
  • Does a topic comparison acknowledge that one work can carry several topic labels?

What a careful conclusion sounds like

When the checks pass, a defensible statement is narrow: within this observatory’s selected sources, the number of newly public research works changed between two complete periods. That statement does not measure funding, scientific importance, experimental quality, or the whole quantum field. It also does not explain the cause by itself.

When the checks fail, the right result is not zero and not a guessed correction. The appropriate label is incomplete or unavailable, accompanied by the last successful source date. Readers can still browse the records that are present, but the site withholds a trend claim. This separation between discovery and inference is one of the main safeguards of the observatory.