Explore how attention changes in English and Russian Wikipedia, with inspectable counts, dated comparisons and explicit limits rather than an opaque popularity score.
For Chess in English Wikipedia, the week 8–14 September 2026 recorded 19,804 pageviews, compared with a baseline of 18,369.5: +1,434.5 views (+7.81%). The baseline is the median of the four preceding complete weekly totals, not simply the previous week. These are request counts, not unique readers or causes.
The retained evidence contains all 210 daily observations for six preselected EN/RU series across science, books and chess, covering 11 August–14 September 2026. Retrieved 18 October 2026; the later retrieval does not make those observation dates current. The sample is selected, historical and not representative Wikipedia coverage.
From the repository root with Python 3.12 or newer; no third-party package required:
python -m tools.verify_d02_artifacts --retained evidence/d02-20261018
python -m unittest discover -s tests -p 'test_*.py' -v
python -m compileall -q tools testsThe first command reads the two checked-in JSON members in memory, verifies their hashes and recomputes all six comparisons. It does not extract files, contact Wikimedia, publish a site or modify inputs. The tests distinguish the actual historical fixture from deliberately synthetic/mutated failure controls. The source record distinguishes the original Actions artifact from the deterministic data-only repack.
The merged D02 collector collects the same fixed historical window on a runtime with Wikimedia access:
python -m tools.d02_probe
python -m tools.verify_d02_artifactsRead artifacts/d02/receipt.json and artifacts/d02/sample.json. A failed command is
not a valid dataset. Run the collector as a module, not python tools/d02_probe.py.
Execution details explain the six sequential requests,
30-second request timeout, five-minute job budget and seven-day Actions retention.
The checked-in capsule does not expire with that Actions artifact.
D02 bounds successful/error bodies to 1 MiB plus a detection byte; oversized, interrupted or inconsistent Content-Length responses do not receive a false complete response hash. Inexact daily/weekly/median/delta values and non-finite results fail explicitly. Source observations can survive a calculation error for investigation, but metrics remain unavailable and the run unsuccessful.
The all-PR offline workflow checks exact heads on Python 3.12/3.13, without live collection, retained credentials or write permissions. Only actual checks establish hosted success. The D02 runtime workflow is separate and does not deploy a website.
The interactive atlas is not yet integrated and there is no verified public site. The remaining pipeline, visitor interface and scoped readiness are tracked in #7, #8 and #9. Local previews are not releases.
Volume, absolute change and baseline-relative change are distinct. Relative change
is unavailable for a zero baseline; incomplete windows are not comparable. No validated
rising/returning/fading classifications are published. agent=user is not verified
human traffic; attention is not importance, approval, truth or causation. Compare
Wikipedia editions, not countries or populations. Top-N absence is not zero.
The historical evidence has no stable page-ID observations and retains reported raw HTTP hashes, not raw response bodies. It cannot prove cross-snapshot identity continuity or independently rehash the upstream bodies. The D03 identity contract remains binding: no title-only stitching, silent redirect aggregation or invented mapping. Wikimedia Analytics API data is CC0; article text and images are not included or relicensed.
The single roadmap orders collector acceptance, retained source adoption, the accessible M0 journey with browser CI, then verified publication #12. Rolling refresh #13 must preserve version-bound shared views and the last valid publication; broader coverage #14 follows with explicit identities and measured actual coverage. No live URL is advertised before verification.
Product anchor · Research registry · Source register · Project manifest