# Taking their land — second_pass_v7_676 Activated as the qualified statistical release on 8 September 2026. Read PUBLICATION_STATUS.json, the active-release pointer and the companion site PUBLICATION.json for release and deployment identity. Copied dated review records are historical evidence, not current instructions or claims of later publication. ## What to use - `data/scores_v2.csv`, `data/na_types_v2.csv`, `data/corpus_v2.json` and `data/analysis_metadata.json` are the exact numerical inputs to the completed diagnostic cycle. Do not substitute corrected metadata when reproducing it. - `events_enriched.json` is current publication metadata for 676 active events. It includes five earlier approved corrections documented, with before/after values, in `editorial/publication_carry_through.json`. No raw score changed. - `editorial/primary_family_labels.csv` and its ledger are the reviewed editorial taxonomy: 672 classified events and four family holds. These are not KMeans assignments or a claim that the factor solution recovered six natural kinds. - `editorial/eight_trait_cards.csv` gives the numeric public-formula projection for all 676 events. It is NOT the final display for every authored case. `editorial/atlas_profiles.json` gives the actual 370-case display: 2,960 traits, including 340 source-qualified rows with no fabricated numeric bar. - `editorial/atlas_display_resolution.json` and its decision/provenance records preserve the earlier qualitative restrictions and eleven resolved conflicts. The six earlier F8 raw-coding follow-ups remain open, separately listed. - `results/` contains all 565 scenario summaries and lossless full results. There are 536 unique preprocessed matrix cases, not 565 independent studies. Both canonical fits retain nine factors; named axes and some sensitivities differ substantially. This is not a fairness, universality or census verdict. - `reproduction/numerical_reproduction_20260907a.zip` includes the explicit input/code subset and a no-fit preflight. Its historical acceptance files identify the numerical inputs, not later public family corrections. Runtime versions are recorded; cross-platform bit identity has not been established. - `comparison/` and dated `provenance/` files preserve superseded formulas, labels and decisions. Consult the current outputs above for present labels. The roster arithmetic is 675 minus 13 retired rows plus 14 bounded successors: 676 active rows. The historical input corpus additionally preserves 95 inactive audit rows (771 total). The 370 mapped cases and 380 playback stops do not change in this publication pass; this does not authorize authoring the remaining 306. Four family holds are unrendered. Prehistoric post-Lapita Vanuatu is unrendered and is not the colonial case named Vanuatu already visible in the atlas. ## Rights and reuse Original data/coding: CC BY 4.0. Original prose/graphics: CC BY-SA 4.0. Original software: MIT. Third-party material retains its own terms. See the license and third-party notices; neither a file extension nor a reference grants rights to the underlying source publication. Source PDFs, browser captures, caches, credentials, raw coder journals and Git history are not public payloads. PUBLIC_EXPORT.json enumerates each selected file and the exact members/hashes inside the numerical ZIP. This selection is separate from release activation, website deployment, private backup, private GitHub and the unpublished Zenodo draft. No DOI or archival deposit is claimed. ## Final metadata and chronology supplement (8 September 2026) The33 accepted older metadata requests are now applied. Current names and date/era labels are synchronized across events_enriched.json and the two editorial tables. The original data/*.json model inputs and results/565-scenario outputs stay unchanged; use data/date_metadata_addendum/ and results/date_metadata_addendum/ together for the amended chronology. The latter is an explicit518-patch overlay,511 amended date records and two new unique era-dependent fits shared by14 aliases. Canonical fit unchanged;629 eligible candidate onsets,0 measured durations. Do not silently replace historical inputs when reproducing the original run. The new reproduction/date_metadata_reproduction_20260908a.zip includes the original99 numerical-export members unchanged and10 supplementary members. Default input checks need no private cache. Full consumer reproduction explicitly recreates the original fits, then executes the two-fit addendum; all-platform bit identity is not established. The detailed README is inside. Original statistical result/verification records describe their dated scope; the prior review does not cover the later addendum. The current public release does not resolve the six raw-F8 follow-ups, four family holds, or remaining authoring holds. ## Activation scope Only the reviewed 370-case display is enabled. Earlier projection-level `atlas_rendering_authorized: false` and `release_authorized: false` values describe the authority of those historical producer stages, not a prohibition on this later accepted release. They do not grant authority to render all676 events. The original prepared HTML is retained as an excluded historical artifact. The original inputs/results and predecessor release are unchanged.