Source manifest and datasets for 'Built for Someone Else: Generation, Load, and the Exception in Virginia'

The data apparatus for Built for Someone Else: Generation, Load, and the Exception in Virginia (Virginia Data Center Series — Analytical Volumes, Version 1.0), concept DOI 10.5281/zenodo.22096613. Start with README.md, which says what this record holds and how the files relate. MANIFEST_sha256.csv is the index to everything: 125 rows, giving for each file a byte count, a sha256, and — for 101 of the 108 primary source documents — the URL it came from. Re-fetch the URL, hash the bytes, compare. A match means you are reading what the volume read; a mismatch means the publisher revised the document after 24 August 2026. RETRIEVAL_NOTES.md covers the documents whose links will not last, by issuing body. The Virginia SCC serves docket documents under opaque internal tokens carrying no case number; the 2020 legislative record sits on a system the Commonwealth itself calls legacy; FERC's eLibrary serves downloads only to a browser session. The notes give the case, docket or report number to search on when a URL dies. Two Department of Environmental Quality guidance documents have no recorded URL at all, and the manifest says so rather than guessing. Five datasets, each with a blind-recode second-coder sheet: • DEQ data center air-permit register — 201 issued permits across 190 facilities, plus 37 pending applications, built from DEQ's public air permit search and verified against DEQ's own permit sequence.• Virginia generation fleet inventory — 1,003 generator records, every unit physically in Virginia that EIA-860 (2024 vintage) reports as operable, proposed, or retired.• ECHO–DEQ crosswalk — 134 EPA Enforcement and Compliance History Online records joined to the permit register, carrying join method and confidence per row.• NEI 2023 extract — 604 emissions records for Virginia data center air sources.• Integrated crosswalk — the joined product, 201 facilities, 30 columns. Two READMEs document the coding decisions, the join key and its limits, and the sixteen ECHO records that could not be joined. build_va_fleet_inventory.py is deposited with the data. It is deterministic: rerun it against EIA-860 archive eia8602024.zip, sha256 0aaae04812cd4ab87a3e346bdf93848a3cc15053fd4dc2a4cf82d2aeac95f12b, and it reproduces the fleet inventory byte for byte. EIA reissues vintage files without changing the filename, so verify that hash before citing. What is not here. The 108 primary source documents themselves are not redistributed. Eight are copyrighted by PJM Interconnection, Monitoring Analytics and the Piedmont Environmental Council and are quoted in the volume under fair use; the rest are public records retrievable from the URLs the manifest carries. The hashes let a reader confirm they hold the same bytes; they do not license a copy from this author. Evidence snapshot 24 August 2026. URLs re-verified 25 August 2026.