Tommy N. Turner
Independent researcher and writer. Institutional governance, public policy.
Research on how local governments, utilities and universities make decisions, mostly in Virginia, built from public records: budgets, permits, minutes, filings and the forms agencies use. Most of it concerns the data center boom and the power, water and money behind it. Every report below is deposited on Zenodo with a permanent DOI.
New here? Start with What Virginia Is Spending, What Virginia Is Getting, the statewide reference, or the Data Centers page, which groups that work by topic. To find your county or city, see data centers by Virginia locality. Work on Virginia counties, cities and towns, data centers included, is gathered on the Local Government page. Op-eds and essays follow the reports.
Virginia data center series — statewide and analytical volumes
- What Virginia Is Spending, What Virginia Is Getting: The Commonwealth Data Center Reference
This statewide reference weighs what Virginia spends on data centers against what it gets. Its central finding: the sales tax break on data center equipment was worth about $1.9 billion in state and local tax to the industry in fiscal year 2025, by the operators' own reports.
The anchor reference for the county case-study series. - Governing the Demand Shock: Local Institutional Variation in Virginia's Data Center Boom (Report)
This report compares how thirteen Virginia localities handled the spread of data centers from 2018 to 2026. Its central finding: the deciding factor was timing, when a community reformed its rules relative to the industry's arrival, not geography or project size.
The compact report. The book edition carries the full narrative. - Governing the Demand Shock: Data Centers, Local Power, and the Distribution of Cost in Virginia
This book compares how Virginia's localities governed the data center wave from roughly 2018 to 2026, drawing on thirteen local case studies. Its finding: what separated communities that captured the benefits while containing the costs was neither geography nor project size, but the timing of local reform relative to the industry's arrival.
Book edition of the report, written for a general reader. - Who Showed Up: A Census of the Organizations Around Data Center Siting in Virginia, 2014 to 2026
This census lists the organized groups on every side of Virginia's data center siting fights from December 2014 to September 26, 2026, from neighborhood groups and national environmental organizations to trade associations, lobbyists, building trades unions and the boards that decided.
A census of every organised body found engaging data center siting in Virginia, on every side, December 2014 to 26 September 2026. Records presence and dates and does not assert cause. Register deposited separately (22998217); saved source pages held in a restricted record (22996886), released on request. Any organisation named may report an error in its row. - Below the Curve: What Virginia Counties Chose Not to Measure
In 2016, Loudoun County staff recommended against regulating tonal noise, the steady hum residents near data centers describe. They priced the alternative at $1,000 and wrote that it could put the county at a competitive disadvantage. The county declined.
Supplement to the Commonwealth Data Center Reference (20679292). The county noise ordinance census workbook and a source companion are deposited in the same record. References the Loudoun (22374193), Prince William (20673498) and Pittsylvania (22968398) case studies and Governing the Demand Shock (20692386). - Thirteen Cooperatives: Virginia's Distribution Cooperatives and the Arrival of Load Larger Than the System
The General Assembly's audit agency, JLARC, projects that about 60 percent of Virginia's data center growth will land in the territory of its 13 electric cooperatives. This reference covers how they are organized, financed and regulated for loads larger than their systems.
Reference companion to the Commonwealth Data Center Reference (20679292). Also addresses Va. Code § 56-234 D (2025 Acts of Assembly, chapters 499 and 598), which reframes the affiliate question. References the Mecklenburg (20673484), Botetourt (20673512), Culpeper (20673529), Google–CFS (20679935), and Governing the Demand Shock (20692386) volumes. - Built for Someone Else: Generation, Load, and the Exception in Virginia
This report maps four routes by which generation reaches Virginia's data centers. Under a 2020 law, the State Corporation Commission may approve a new utility-owned fossil plant only if the utility has met energy savings goals or the Commission finds a threat to reliability or security.
Companion volume to the Commonwealth Data Center Reference (concept 20679292). Source manifest deposited separately as concept 22099427. - Concentrated Benefit, Dispersed Burden: Who Carries Virginia's Data Center Transmission, and What the Cost-Allocation Record Does Not Trace
On February 26, 2025, grid operator PJM's board approved the Joshua Falls to Yeat high-voltage line. PJM attributes the load growth behind it primarily to data centers.
Volume I in the Concentrated Benefit, Dispersed Burden series — Virginia transmission cost allocation. Companion Zenodo dataset: 21959988 (utilities transmission-siting geographic data). - What the Record Does Not Say: Interstate gas supply for Virginia generation, and a question no forum wrote down
On November 25, 2025, Virginia's State Corporation Commission approved Dominion Energy Virginia's 944-megawatt Chesterfield Energy Reliability Center, a gas plant. After reconsidering, it reaffirmed that on February 12, 2026.
Working paper. Companion to Built for Someone Else (22096613); supporting eleven-docket dataset deposited as 22757119. - When the Operator Builds the Plant: The Google–CFS Power Purchase Agreement and What It Means for Virginia's Energy Market
On June 30, 2025, Google agreed to buy 200 megawatts, half the planned output, from a fusion plant Commonwealth Fusion Systems plans in Chesterfield County. On September 22, 2025, Eni and the company announced a second deal for the plant, which they valued at more than $1 billion.
- No Field to Land In: How Three States' Water Withdrawal Forms Make Utility-Supplied Data Centers Illegible
This paper reads the water withdrawal reporting forms of Virginia, Tennessee and Georgia field by field. Virginia and Tennessee rules treat a data center buying utility water as a customer, not a withdrawer, as does Georgia's surface water statute. The only report reaching the state is the utility's.
Cross-state water-reporting analysis (VA, TN, GA). Documents the record void created by the utility-supplied-exemption plus form-field absence combination. Companion to the Virginia and Georgia water volumes. - A Line Drawn on a Highway: Virginia's Groundwater Boundary, the Rivers Beyond It, and the Data Center Load the Regime Does Not Reach
This report examines how Virginia regulates water west and north of the Fall Line. Its finding: Virginia counts its water everywhere and allocates it almost nowhere. Its rules ask about place, start date and source, not quantity.
Companion volume to the Commonwealth Data Center Reference (concept 20679292); extends the coastal-aquifer thread from Two Systems (21816008) to the full groundwater and surface-water regime. - Where the Instrument Stops: Virginia's Groundwater Ceiling, Its Grandfathered Rivers, and the Load Outside Both
This brief sets three Virginia water reports side by side. It finds the state measured its groundwater where the data centers are not. Virginia limits groundwater by quantity only inside declared areas, and in the east that line is drawn at Interstate 95.
Executive brief synthesizing three deposited water volumes (21556435, 21816008, 22105181). Carries four findings that none of the three volumes makes alone. - Two Systems, One Drought: Virginia's Groundwater Ceiling, Its Rivers, and the Water the Energy Buildout Actually Uses
This report examines Virginia's groundwater and river systems during the 2026 drought. Its central finding: Virginia has the authority to weigh the two together and does not do so systematically. It has never declared a surface water management area, which would let it require permits for river withdrawals and set a domestic-use priority.
- The Aquifer Says No: DEQ's SJ 25 Groundwater Study and the Data Center Question East of I-95
In July 2026, Virginia's Department of Environmental Quality released its study of groundwater east of Interstate 95, ordered by the General Assembly.
- Who Builds the Data Centers: What the Employment Record Can and Cannot Show in Virginia, Georgia, and Tennessee
This report asks what available records can show about who builds and staffs data centers in Virginia, Georgia and Tennessee. The construction trades are where the record runs out: none of the thirteen sources it reads counts the construction tradespeople on a data center site in any of the three states.
Report. Reads thirteen employment sources side by side for Virginia, Georgia and Tennessee. The record also holds a capture of 410 data center job-board ads from October 19, 2026, as CSV files. Draws on the labor papers (22866646, 22861996) and their dataset (22862472). - Membership grew faster than employment: A sixteen-year panel of inside electrical construction in hyperscale data center metros, 2010 to 2025
This research note tracks union membership in inside electrical construction from 2010 to 2025, using the yearly reports IBEW locals file with the U.S. Department of Labor. In metro areas where Google, Meta, Microsoft or Amazon Web Services built data center campuses, membership grew 1.5 to 1.9 times as fast as electrical contractor employment.
Research note, version 1.0. Own concept DOI; supersedes findings of 22310705's earlier v1.0 (which remains citable). Companion measurement paper is 22861996; shared replication dataset is 22862472. - Counting a trade in one metropolitan area: union membership for inside electrical construction from federal filings, 2010 to 2025
This paper builds a public dataset of union membership, 2010 to 2025, for every local of the electricians' union (IBEW) that filed a yearly report with the U.S. Department of Labor. The union's maps mark which locals do inside construction, the wiring of buildings.
Measurement paper. Companion research note is 22866646; shared replication dataset is 22862472. - Where the electricians went: Hyperscale data center construction and building trades membership in federal union filings, 2018 to 2025
This working paper builds a count of union electricians from the yearly membership reports unions file with the U.S. Department of Labor under penalty of perjury. It follows 89 locals of the International Brotherhood of Electrical Workers that hold inside construction jurisdiction, the wiring of buildings, from 2018 to 2025.
Working paper v0.1 with complete replication package. Uses federal union filings as an independent labor-measurement source. Effect concentrated in metros with hyperscale campuses; removing top-five-fastest-growing locals leaves 9.0 points and no significance. - Counting Data Centers: The OpenStreetMap Inventory Behind a National County-Level Comparison
This research note checks the data center counts behind the National Association of REALTORS® 2026 Data Center Impact Report, published September 9, 2026, which compares housing, jobs and electricity across four groups of counties.
Working paper. Methodological note on the OpenStreetMap-derived facility file (Pacific Northwest National Laboratory) that underlies a widely-cited national inventory. - Data Center Maps: A Register of What Exists, What Works, and Where It Fails
This working register catalogues the resources that map, track or count data centers, nationally and in depth for Virginia, Georgia and Tennessee. Its main finding: no resource is a census. Every tracker's count for the three states disagrees with the others; Tennessee's run from 12 to 88 depending on the source.
Register v0.1, resources checked 28 September 2026. The full report and register (PDF) is mirrored here; the two-page briefing and the filterable workbook, with a sheet for each state, are on the Zenodo record. Prince William County's Build-Out Analysis map is named as the closest existing template for a better map. - Transmission Siting Disclosure Capture: TVA, Oncor/LCRA and Dominion Energy Virginia, August 2026
A point-in-time capture of publicly published transmission line siting geographic data, retrieved 15 August 2026 by anonymous query against public ArcGIS REST endpoints.
Point-in-time dataset capturing which utilities publish vs. withhold transmission-siting geographic data (TVA, Oncor/LCRA, Dominion Energy Virginia).
Virginia data center series — county case studies
Fourteen Virginia case studies, each read from its own packets, minutes, ordinances and agreements.
- Case Study: Henrico County — How Two Neighborhoods Forced Virginia's Second-Largest Data Center Jurisdiction to Change Course
On June 10, 2025, Henrico County's Board of Supervisors voted 5-0 to end by-right approval of data centers, meaning staff approval without a public hearing. New stand-alone data centers now need public hearings and the board's approval.
The first case study in the county series. - Case Study: Hanover County — Mountain Road, Tract, and the Search for Leverage
On May 27, 2026, Hanover County's supervisors turned down a 900-megawatt data center campus, Tract's Mountain Road Technology Park, by a 4-3 vote.
- Case Study: Chesterfield County — The $0.24 Rate, the Thirty-Year Lock, and the Two-Corridor County
In April 2019, Chesterfield County's Board of Supervisors set its tax on data center equipment at $0.24 per $100 of value. It was called the lowest rate in Virginia, and it deliberately undercut Henrico.
- Case Study: Mecklenburg County — Boydton, South Hill, and What Fifteen Years of Data Centers Actually Bought
On August 27, 2010, Microsoft's plan for a data center in Boydton, the Mecklenburg County seat, was announced: up to $499 million in investment and 50 jobs. The buildout has since reached the edge of South Hill.
- Case Study: The Fredericksburg Ring — Five Jurisdictions, One Corridor, and the I-95 Path South
Between 2022 and June 12, 2026, Spotsylvania, Stafford, Caroline and King George counties approved data center campuses with at least 5.4 gigawatts of reported capacity.
- Case Study: Orange County and the Town of Orange — Convergence at Two Scales: One Industry, Two Governments, and the Entitlement Neither Can Reach
On February 17, 2026, the Town of Orange's council ended by-right data centers, those allowed without a public hearing, in its town center and required a special permit in every district where they are still allowed. On April 28, 2026, Orange County's supervisors voted 5-0 to confine county data centers to a Technology District that, as of June 12, 2026, covered no land.
- Case Study: Louisa County — The First Overlay, the Amazon Retreat, and the $17 Billion Bet
On April 3, 2023, Louisa County's supervisors adopted Virginia's first Technology Overlay District, allowing data centers without a public hearing across roughly 6,400 acres. Within five months, Amazon Web Services committed $11 billion to two campuses there.
- Case Study: Goochland County — The Overlay, the Lawsuit, and the Template
At a November 6, 2025 meeting that ran past midnight, Goochland County's supervisors voted 4-1 to adopt a data center overlay of roughly 4,400 acres along Route 288, most of it with no public hearing per project. More than 80 people spoke against it.
- Case Study: Botetourt County — Project Raspberry, the Water Authority, and the Wave That Crossed the Blue Ridge
On June 24, 2025, Botetourt County announced that Google had bought 312.55 acres of the county's Greenfield business park.
- Case Study: Culpeper County — The Quiet Successor
On September 2, 2025, Culpeper County's supervisors unanimously required a special permit for data centers on industrial land countywide, exempting the technology zone and its five approved projects.
- Case Study: Prince William County — Governance Capture, the Revolving Door, and the Democratic Correction
On March 31, 2026, Virginia's Court of Appeals upheld a ruling that voided the Prince William Digital Gateway, three data center rezonings covering about 1,760 acres. The hearing had not been properly advertised.
- Case Study: Loudoun County, Virginia — The Endgame Case: What It Costs to Change the Rules After the Industry Is Load-Bearing
On March 18, 2025, Loudoun's supervisors voted 7-2 to require board approval for data centers where they had been allowed by right, meaning without a public hearing.
The Endgame Case. Loudoun as the diagnostic county — what it costs to change the rules after the industry is load-bearing. isSupplementTo the Commonwealth Reference (20679292); references every other county case in the series. - Case Study: Pittsylvania County — The County Turned Down One Data Center Plan, Then Rezoned for a Bigger One
On April 15, 2025, Pittsylvania County's Board of Supervisors voted 6-1 to deny a rezoning of 763 acres for a private gas power plant and data centers. It had first refused the applicant's request to withdraw.
Refusal, then entitlement by ordinance on the county's own motion. Traces the Berry Hill loop and substation built by Appalachian Power under the business-park pilot statute. isPartOf the Commonwealth Reference (20679292). Evidence snapshot 25 September 2026. - Case Study: Fauquier County and the Town of Warrenton: The Restrictions Came After the Entitlements, and the Town Changed Sides
In February 2023, the Warrenton Town Council approved an Amazon data center, 4-3. The town approved its site plan in April 2024. It remains unbuilt.
Two governments in one county: the Town of Warrenton's Amazon permit, now struck as a use and challenged in court, and Fauquier County's restrictions at Remington and Vint Hill, adopted after the entitlements. Corrects the series' earlier use of Fauquier as a refusal case. Record read through 1 November 2026.
Georgia and Tennessee
Statewide reports and the federal record for the two companion states.
- What Georgia Is Spending, What Georgia Is Getting: A Statewide Data Center Report
This statewide report weighs what data centers cost Georgia against what they return. Its lead finding, from a 2025 evaluation done for the state: the sales-tax break for data center equipment loses the state money in every year the study modeled, even counting the taxes the activity brings in.
- What Tennessee Is Building, and What It Is Not Regulating: A Statewide Data Center Report
This statewide report maps Tennessee's data center buildout, roughly 61 facilities operating or under construction by mid-2026. Its organizing fact: the utility serving them, the federal Tennessee Valley Authority, answers to no state regulator. The rate a data center pays and the power plants built to serve it are decided where no Tennessee voter can reach.
- Declared, Not Measured: Volumetric Permit Triggers and Sub-Threshold Industrial Water Demand in Georgia
This paper reads Georgia's groundwater permit law rule by rule. Anyone withdrawing more than 100,000 gallons of groundwater a day needs a permit, but the rules require regular reporting of actual withdrawals only from permit holders. Until a permit issues, the amount that settles whether one is needed is only the user's declaration.
Extends the Virginia water thread (references 21816008 and 22105181) into Georgia. Data centers are the occasion; the design property is stated independently as a limiting case of Bennear (2008). - What the Certificate Points At: Fuel disclosure, sealed cost, and one Georgia certification of 9,885 megawatts
On December 19, 2025, Georgia's Public Service Commission granted in full two Georgia Power requests for capacity whose two parts total 9,885 megawatts. Its own advocacy staff had recommended 3,125, leaving out all three gas plants the company will build.
Working paper, Georgia companion to What the Record Does Not Say (22754150). Shares the eleven-docket register 22757119 as supplement; part of the Built for Someone Else (22096613) source apparatus. - Sufficiently Analogous: Federal need review for a gas shipper with no state regulator, in two Tennessee certificates
The Federal Energy Regulatory Commission approved two Tennessee gas pipeline projects, Cumberland on January 18, 2024, and Ridgeline on April 2, 2025. In both, the Tennessee Valley Authority took the full capacity, for new TVA power plants.
Working paper. Extends What the Record Does Not Say (22754150) and What the Certificate Points At (22754180) into Tennessee. Shares the eleven-docket register 22757119 as supplement. - The Whole Record: What the Tennessee Valley Authority Publishes When Nothing Requires It
This book-length report examines what the Tennessee Valley Authority, a federal corporation, publishes about transmission lines whose need and route no outside body decides. Of TVA's 18 Tennessee project pages, one names the customer a line serves; none gives a megawatt figure.
Volume II in the Concentrated Benefit, Dispersed Burden series (Vol I = concept 21895588). 264 pages across fifteen chapters. Codes eighteen TVA project pages and reconstructs the categorical-exclusion record from 11,899 determinations. - Getting the Record: A Guide to Requesting TVA Transmission and Siting Documents
This guide is a working list of records requests for anyone researching Tennessee Valley Authority transmission lines and siting in Tennessee. No certificate proceeding or hearing reviews TVA's lines, so a records request, which yields documents after decisions rather than a voice before them, is the tool that exists.
Practitioner guide companion to The Whole Record (21996375). Walks how to request TVA transmission and siting documents where no state PUC docket exists, given TVA's disclosure practice and NEPA categorical-exclusion pattern.
Radford and municipal fiscal
- Fiscal Crossroads: An Options Analysis of the City of Radford, Virginia
This options analysis of Radford's finances calls one fact the most important: the general fund does not cover its own operations. In fiscal 2024 it took in $25.43 million and spent $33.14 million; even with money moved in from the electric utility, it ran about $3.74 million short.
- The Quiet Option: How Radford Drifted Into Fiscal Distress
This study traces how Radford became a fiscally distressed city, using audited financial statements from 2011 through 2024. It finds that in every audited year, the city moved money from its electric utility into its general fund, $52.04 million over those years.
Companion to Fiscal Crossroads. - Two Cities, One Machine: Radford and Martinsville
This report compares Radford and Martinsville, small Virginia cities that each own an electric utility and resell wholesale power to residents. In the year that ended June 30, 2024, each utility spent 80.8 cents of every revenue dollar on purchased power.
- What the Foundry Took
In April 2017, Radford's city council passed a budget 3-2 over a warning, read into the minutes, that it was not sustainable and drew too heavily on the city's reserves.
Published as Cardinal News diptych, June 10–11, 2026. - Virginia College-City Fiscal Comparison: Charlottesville, Harrisonburg, Williamsburg, Radford
This report compares the finances of four Virginia independent cities that host public universities: Charlottesville, Harrisonburg, Williamsburg and Radford. It finds that over the years studied, Radford is the only one whose net position, roughly what a government owns minus what it owes, declined, falling 35 percent while the other three grew.
- Choice Under Constraint: Enterprise-Fund Transfers and Divergent Fiscal Trajectories Among Virginia's Independent Cities
This working paper compares Radford with similar Virginia cities. It finds the evidence consistent with this: a yearly transfer from Radford's electric utility to its general fund, alongside real structural pressure, materially reduced the city's margin of safety.
- The Handle: How American Cities Tax the Universities They Can't Tax
This book examines how cities raise money connected to universities exempt from property tax. Its central argument: state law, much of it old and written for other purposes, often decides what a city can collect more than negotiations do.
AI governance in higher education
- Governance Without Readiness: How 53 American Universities Responded to Generative AI
This working paper looks at how 53 American universities, William & Mary's national peers and Virginia four-year schools, responded to generative AI, based on what each had published by March 2026. Its central finding: they responded mainly by governing AI. Fifty-two formed an AI committee or task force. Only four documented outreach to employers about the AI skills graduates need.
- Virginia Tech: Comprehensive AI Governance & Activity Inventory
This inventory records what Virginia Tech had said and done publicly about artificial intelligence as of September 26, 2026. It finds that the gap between principle and implementation is the central challenge of the university's AI governance.
- Lehigh University: Comprehensive AI Governance & Activity Inventory
This inventory records what Lehigh University had publicly said and done about artificial intelligence by June 10, 2026. It finds AI governance centralized in the provost's office. A Generative AI Advisory Group, set up on January 20, 2025, reported within four months.
- University of Virginia: Comprehensive AI Governance & Activity Inventory
This inventory records what the University of Virginia had said and done publicly about artificial intelligence as of June 9, 2026. It finds that UVA's AI coordination is broad but its regulatory authority is narrow, and that the distance between them defines its AI governance.
- Marymount University: Comprehensive AI Governance & Activity Inventory
This inventory records what Marymount University had said and done publicly about artificial intelligence as of April 6, 2026. Its most prominent finding is a gap: investment in AI programs is growing rapidly, while governance remains anchored in the academic integrity code and research review board.
- William & Mary: Comprehensive AI Governance & Activity Inventory
This inventory records what William & Mary had said and done publicly about artificial intelligence across 13 units, as of March 2026. A university AI policy committee delivered its recommendations in July 2025, but the permanent AI council it proposed had not been formally set up eight months later.
- AI Governance Peer Institutional Benchmark: William and Mary's 23 BOV-Adopted National Peer Institutions
This reference report profiles how 23 universities, the national peer group adopted by William & Mary's Board of Visitors, were handling generative AI as of March 2026. Each profile uses the same ten sections, from governance committees and classroom guidance to AI tools, academic integrity and research rules, drawn only from public sources.
- AI in Virginia Higher Education: Public-Source Reference Guide for 30 Four-Year Institutions
This reference guide compiles what 30 Virginia four-year colleges and universities, public and private, had published about generative AI as of March 2026. Each entry follows the same ten parts, covering governance, teaching guidance, AI tools, academic integrity, research ethics and strategy.
James City County governance
- The Working Majority Public Dissent and Decision-Making on the James City County Board of Supervisors, 2010–2026
This working paper examines how James City County's Board of Supervisors handled organized public opposition from January 2010 through February 2026. Under its strict test, four votes qualify: the Hazelwood Farms rezoning in February 2022 and three Government Center votes in 2025. The board approved all four.
- Two Changes on the James City County Board of Supervisors: A Longitudinal Cohort Analysis, 2010–2026
This working paper studies how James City County's Board of Supervisors made decisions from 2010 to early 2026, using 3,136 recorded votes. It finds two changes, at different times, needing separate explanations.
- Representative Judgment After Public Opposition
On October 14, 2025, three weeks before an election, James City County's Board of Supervisors voted to approve a $16.5 million early site work package for its Consolidated Government Center. Residents had cited a petition approaching 1,000 signatures and asked the board to wait.
- Public Participation and Representative Judgment in James City County
On October 14, 2025, three weeks before a county election, James City County's Board of Supervisors approved two steps for its Consolidated Government Center: vacating part of a road right-of-way and a $16.5 million early site work package.
- Mounts Bay: A James City County Story, 2012–2026
This book follows James City County's Board of Supervisors through its public meetings on Mounts Bay Road, from October 2012 to February 2026. Its central finding: a board that took its staff and Planning Commission recommendations as binding in 2012 came to override them on a recognizable set of cases, over split votes in 2016 and 2017 and by consensus from 2018 through 2023.
Book-length narrative, v3 rev 4. isPartOf the Regional Government Records Corpus (22045517/20088873). - Three Rooms: Williamsburg and James City County and the Joint School System, 1955–2026
On October 10, 2025, James City County's supervisors, the Williamsburg City Council and the joint school board voted unanimously for a new contract for the school system the city and county have shared since 1955. For the first time, the school board is a party.
Book-length institutional history, v3 rev 6. isPartOf the Regional Government Records Corpus (22045517/20088873).
Generative AI and knowledge systems
Scholarly thread on human-machine coupled knowledge systems. Companion pieces run on LinkedIn Pulse under Essays.
- Who Is Training Whom: Human Cognitive Drift and Machine Degradation in Coupled AI Knowledge Systems
This working paper treats a person and an AI system as one system and asks how AI literacy training should change. In September 2026, four AI search products answered six questions about Virginia data-center policy, twice each. No cited page in the set documented its own check against a record.
Working paper, scholarly companion to The Mountains That Were Not There (LinkedIn Pulse, 18 September 2026). Supplement dataset is the Source Diversity Audit (22741216). - Where the Number Came From: How Generative Search Loses the Record, and What an Operator Can Do About It
On September 11, 2026, four AI search products were asked what share of Loudoun County's local tax revenue comes from data centers. The answers and their cited pages carried six different percentages, each true of some base and year. Only one said which fraction it reported.
Explanatory brief for people who run organizations; one worked case, no rate claim. Supplemented by the Source Diversity Audit dataset (22839545); the corpus-scale argument is in Who Is Training Whom (22839553).
Across the corpus
One essay drawing the research threads together, page-referenced to the deposited works.
- Who Was Assigned to Ask How Much: Institutions, Their Records, and the Fact Nobody Had a Field For
This working paper reads across the author's earlier studies. Its finding: in the settings it examines, the decisive fact is reliably the one nobody has a field for.
Working paper, version 1.2. Cites twenty-two works in this corpus; the related-works list below is the essay's bibliography.
Datasets and reference materials
Companion datasets, corpora, and reference documents. Cite the parent report where applicable.
- Fiscal Crossroads Workbook: City of Radford Audited Fiscal Data and Reconciliations, FY2011-FY2024
The reproducibility workbook behind the Radford fiscal-distress research.
Replication workbook for Fiscal Crossroads - Governance without readiness: coding dataset of 53 US universities' generative-AI governance and readiness
Institution-level coding of 53 US universities on four governance indicators and five readiness indicators, plus a descriptive AI-detection posture and a five-category whole-profile classification.
Coding dataset for Governance Without Readiness - Virginia Data Center Research & Industry Directory
Version 3.0 incorporates three months of portfolio work since v2.1 (March 22, 2026) and a full June 2026 re-verification of every entry.
Contact directory across Virginia data-center research - Regional Government Records Corpus: A Longitudinal Public-Meeting Archive for Local Governance Research in Virginia
The Regional Government Records Corpus is an independently assembled longitudinal archive of local-government public meeting materials spanning eleven Virginia jurisdictions across the Hampton Roads,…
Statewide public-meeting archive — approximately 87,800 files across eleven Virginia localities (five at research depth, two comparison cases, four capability probes). Proof of concept for a statewide Virginia local-government records archive. - Regional Government Records Corpus: Institutional Documentation Package (Volumes I-III)
A three-volume institutional documentation package for the Regional Government Records Corpus, an independently assembled longitudinal archive of Virginia local-government public meeting records…
Three-volume institutional documentation package for the Regional Government Records Corpus (Overview, Methodology, Partnership Pitch). PDF mirror links to Volume I (Overview); Volumes II-III are Zenodo-only. - 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.
Source manifest for Built for Someone Else (concept 22096613) — 125 files including 108 primary source documents, each SHA256-hashed with URL of origin. - Interstate Gas Supply Register: Georgia, Tennessee, Virginia, v1.4
Tennessee, carried as a declared gap in every deposited version until now, is closed.
Interstate Gas Supply Register v1.0 — eleven FERC certificate dockets for interstate natural gas capacity serving Georgia, Tennessee and Virginia, compiled 11–13 September 2026. The unit is the docket, not the state. Companion dataset to What the Record Does Not Say (22754150). - Source Diversity Audit: what the citations displayed by four generative search products reach, traced on six Virginia data-center policy questions (September 2026)
Capture, tracing and coding data for an audit of what the citations displayed by four generative search products actually reach.
Source Diversity Audit — capture, tracing and coding data for an audit of what the citations displayed by four generative search products (ChatGPT, Gemini, Perplexity, Grok) actually reach, on six Virginia data-center policy questions, September 12–13, 2026. - Replication materials: by-right zoning, early vesting and the reach of local reform in Henrico County, Virginia, 2009 to 2026
Version 1.2, October 11, 2026.
Replication materials (v1.1) for a single-case documentary study of Henrico County's June 10, 2025 ordinance ending by-right approval of data centers — what the reform reached and who got a hearing. Supplement, tables, figure data and working notes. Cited from the study's data-availability statement; parent manuscript under journal review. References the Henrico case study (20671031). - Who Showed Up: Register of the Organizations Around Data Center Siting in Virginia, 2014 to 2026
The register behind Who Showed Up: A Census of the Organizations Around Data Center Siting in Virginia, 2014 to 2026 .
Register behind Who Showed Up (22996870): 223 rows (221 live), 23 columns, every value with the page it was read from and the date; codebook, counts-and-gaps file, conflict log, browser-pass record and a manifest of the 611 saved source pages. v1.3, evidence snapshot 26 September 2026. - IBEW local union membership panel, 2010-2025: replication dataset for two studies of inside electrical construction
Version 1.0, 20 September 2026.
IBEW Local Union Membership Panel, 2010–2025 (v1.0). Annual reported membership for IBEW local unions built from Form LM-2, LM-3 and LM-4 filings. Data + scripts + classification inputs + retrieval logs. Shared replication dataset for Where the Electricians Went (22310705), the measurement paper 22861996 and the sixteen-year research note 22866646.
Essays
- The Flinch That Never Came
Some institutional damage does not look like panic. It looks like composure.
- 98 Percent Built AI Committees. 5.7 Percent Included Career Services
What 53 American universities built in response to generative AI.
- The Quiet Disappearance of a Shared Baseline
A shared baseline is infrastructure, not content.
- Promoted Past Learning: What AI Is Quietly Doing to the People Who Use It
Reviewing a document at work. Increasingly, the process behind the output is harder to see.
- Synthetic Participation Is Breaking Professional Trust
When engagement scales faster than accountability, trust becomes expensive.
- Law Over Ego: Martin Luther King Jr. Day, and the test we keep failing
Every generation inherits the same choice.
- If AI Is Everywhere, Why Isn't Productivity?
AI now dominates earnings calls, capital spending plans, and boardroom strategy. And yet the surge isn't there. Not yet.
- AI Didn't Fire Anyone. Leadership Did.
AI is being cited far more often than it is actually responsible.
Trade and policy publications
Commentary and interviews in specialist publications on power, planning and regulation.
- PJM's Tx Owners Would Lock In What a Data Center Pays
The transmission owners' plan would have whoever buys transmission service for a data center sign on for 10 to 20 years and promise a minimum payment, backed by a letter of credit. It never asks whether that payment covers the local lines and substations the data center requires; a footnote keeps those costs in the rate shared by everyone in the zone. PJM reviewed over $9 billion of such supplemental projects in 2024.
Stakeholder Forum commentary. Deck as published: They Don't Show the Payment Covers the Lines Built to Serve It. Chart by the author from PJM's 2025 RTEP. Link is subscriber-only. - A noise limit on data centers is only as good as its measurement
A Loudoun supervisor said the county's average readings beside the Vantage campus never passed 55 dB(A); a resident half a mile away measured 70 to 80 decibels. Neither disproves the other. Every noise ordinance decides which frequencies count, which samples count, how they are averaged and where the meter stands, usually without saying so. Of the 75 Virginia instruments adopted as law, 42 carry no numeric decibel limit at all.
Feature. Deck as published: Counties and states must make conscious decisions about when, where and how they measure data center noise if they want to meaningfully improve residents' lives. Draws on the noise ordinance census in Zenodo 23145782 (Below the Curve). - Virginia Decides Who Pays for Data Center Hookups but not the Upgrades
The SCC's July 31 order makes a large new customer pay up front for its own hookup to the grid and puts off the expensive part: the supplemental upgrades its arrival makes necessary, which reach every Dominion customer through a transmission charge. Dominion planned $1.28 billion of those local projects for 2026. The order already lists the fix.
Stakeholder Forum commentary. Draws on the transmission cost-allocation record in Zenodo 21895588 (Concentrated Benefit, Dispersed Burden). Link is subscriber-only. - FERC Leaves Fuel Question to States; Virginia's Order Doesn't Answer it
The SCC certificated a 944 MW DOM Zone peaker on PJM's capacity shortfall without stating its capacity factor or what it costs to get gas to it. FERC leaves that question to state regulators; three weeks after Virginia's hearing closed, Dominion's fuel subsidiary signed for 20 years of Mountain Valley capacity, and neither record mentions the other.
Stakeholder Forum commentary. Deck as published: SCC OK's Peaker on PJM Capacity Shortfall Without Stating Capacity Factor or Gas Costs. Draws on the PUR-2025-00037 and Va. Code § 56-249.6 record in Zenodo 22754150 (What the Record Does Not Say). Link is subscriber-only. - Half of a 765-kV Line in PJM is a Postage Stamp; the Other Half Stops at the Zone Line
PJM's Schedule 12 allocates half the cost of a data-center-driven transmission line by geography and half by ignoring it — nothing in the tariff conditions on where the burden actually falls.
Stakeholder Forum. Companion to Zenodo 10.5281/zenodo.21895588 (Concentrated Benefit, Dispersed Burden). Link is subscriber-only.
Op-eds and guest columns
- Researcher: Valley Link is now $5.1 billion and 2031. PJM's own rules give the counties a question to ask.
PJM's Nov. 3 materials raise Valley Link from $3 billion to $5.1 billion and move all 14 of its projects past their June 1, 2029 date; the Joshua Falls to Yeat line now shows Jan. 17, 2031. PJM's operating agreement says that when a developer misses a milestone and the project is delayed, PJM "shall re-evaluate the need" for it. The counties along the route and their landowners can put five questions to PJM in writing before its board acts.
Opinion. Deck: Every date in Valley Link's table has moved, the line's by more than a year and a half. Follow-up to the Aug. 24 Buckingham piece. - York said 'enough'
York County voters turned down one more penny of sales tax, 61 percent no. In 18 months they had been handed a real estate increase, a higher meals tax, higher sewer and trash fees and a new admissions tax, with the same promise each time that the tourists would pay. Across the 47 localities that voted, the places already above the usual sales tax rate ran furthest behind, and York, alone at 7 percent, ran about 14 points behind.
Guest opinion, the day after the Nov. 3, 2026 school sales tax referendums. - Who goes dark first, data centers or us?
Botetourt County's website promises that homes, schools, hospitals and emergency services come before Google's data center when power runs short. The county can make that promise and has no power to keep it: Appalachian Power would do the cutting, and no Virginia rule tells it whether homes go dark before data centers. The commission should write the county's answer into a rule the utility has to follow, and DEQ should write the generator trade into Google's air permit, with every emergency run reported in public.
Print, Sunday A15, November 1, 2026 (Commentary). Web headline: "Commentary: Virginia still hasn't decided who goes dark first." Draws on the Botetourt case study (20673512); bio links /data-centers. - James City County wants it all. The budget says pick two.
James City County said no to 770 houses, lost 419 jobs at Avid Medical eight days later, and raised the real estate, meals and admissions taxes 3-2. Where the general fund's money comes from, what a rezoning can make growth pay, and why sending the Grove battery project away gets Grove nothing and the rest of the county two cents on the rate.
Williamsburg Watch runs guest opinions on one rolling page, subscriber-only past the opening paragraph. - Researcher: The data center jobs are real. So is everything else on the ledger.
A reply to Dwayne Yancey's Sept. 29 column on data center jobs. The 50-jobs figure is the number Virginia's tax exemption requires, union membership in data center metros grew faster than employment, and the rest of the ledger, from Botetourt's Google agreement to transmission, water and the exemption's unmeasured effect, has never been added up.
Opinion. Links the IBEW membership papers (22861996, 22866646), Botetourt (20673512), Concentrated Benefit (21895588), The Aquifer Says No (21556435) and What Virginia Is Spending (20679292); bio links /data-centers. - Trump White House denied CNN, Politico and MS NOW press access despite federal judge's order
Sequel to the April Pentagon column. The White House's delay in restoring press access after a federal judge's order exposes a deeper threat to due process, press freedom and the rule of law.
Online kicker: Press Freedom. Print: Tuesday, Oct. 6, 2026, page A6, as 'Trump's ban of reporters tests rule of law' (kicker: White House Access). - The thousand-dollar line
Loudoun priced the fix at $1,000 in 2016 and passed on it, citing competitive disadvantage. A decade later, Prince William raised its residents' 18 recommended sound values by 6 decibels each — and that weakened standard is now the strictest thing Virginia has.
Deck: What I found reading every noise rule in Virginia that could reach a data center. 73 rules across 43 jurisdictions. - Researcher: Buckingham will host more of this power line than any county in Virginia, and it has the least say about it
Buckingham hosts more of the Valley Link line than any other Virginia county — more than Louisa, Orange, or Culpeper, all of which have data centers. Tested against 2,000 random corridors, the route runs through data-center counties more often than chance would allow. Nobody aimed it there, and nobody in Virginia can undo it.
Companion to the PJM Schedule 12 piece (RTO Insider, Aug 12) and Zenodo 21895588 (Concentrated Benefit, Dispersed Burden). - Local governments cut out the public
Botetourt, Wythe, and Pulaski each disclosed on schedule. The problem is the schedule — information arrives after a project has accumulated enough commitments to make reversal expensive.
Print, Sunday A15, August 23, 2026. Web headline: "Commentary: The public interest shouldn't be an afterthought." - Column: Virginia undecided on who gets water from refilled aquifer
If the Potomac aquifer is refilled through SWIFT injections, Virginia has not decided who gets the water first. That policy vacuum is the piece.
Print: shared editorial page, Wednesday August 12, 2026. Also at pilotonline.com under the same headline. - Column: Virginia groundwater limits put growth in question
DEQ's SJ 25 study finds Virginia's Coastal Plain aquifer capacity is the ceiling; existing pumping and JCSA's permit renewal now sit under it.
- Williamsburg Watch column on JCSA permit renewal implications
What the DEQ technical evaluation implies for the JCSA groundwater permit renewal — and for James City County's growth assumptions underneath it.
Links to Williamsburg Watch opinion index; individual permalink pending - Williamsburg Watch column on Virginia SJ 25 DEQ groundwater study
Introducing the DEQ SJ 25 groundwater study, released six months past its statutory deadline: the Coastal Plain aquifer system has limited capacity for new withdrawals east of I-95.
Links to Williamsburg Watch opinion index; individual permalink pending - Trump's Hormuz toll reversed, but the damage remains
- Radford named its distress. It hasn't named the cure.
- Researcher: The bill came dueRadford diptych, part 2 of 2
- Researcher: What the foundry took from RadfordRadford diptych, part 1 of 2 — also deposited as Zenodo 20612127
- When commemoration becomes political powerPrint, Sunday D4 — kicker: The Founders' Warning. URL pending.
- The anniversary America's founders wouldn't believe
- Early warning signs of Radford's fiscal distress
- The case against Secretary Hegseth asking for $1.5 trillion
- China talks cannot replace a congressional vote on war
- As James City County budget adoption looms, disconnect between supervisors and residents is glaringly apparent
- Column: Who is counting taxes in James City County
- Researcher: Spanberger visited Southwest Virginia. The way she visited tells the region what she thinks of it.
- Spanberger's data center position is the test of her affordability message
- More thoughts about budgeting, taxes — a letter to the James City supervisors
- Letter to JCC Supervisors
- Commentary: Taxing past the breaking point
James City County residents count taxes cumulatively; the board counts them categorically. That mismatch is the piece.
- Pentagon press crackdown ruled unconstitutional — twice
- Spanberger's first 100 days are flashing a warning Virginia Democrats should not ignore
- Why Virginia's data center deal no longer adds up
The Senate's move to end the $1.6 billion sales-tax exemption is overdue honesty, not anti-growth policy. Virginia is subsidizing a dominant industry to stay somewhere it was going anyway.
Also ran in The Daily Progress under the headline "The data center numbers don't add up" (March 24, 2026). - A quiet digital fence around the Historic Triangle
- Column: Finding patterns in the shift away from democracy
- Virginia should stop subsidizing data centers
- What Customs taught America about power
Essays
- The Empty Screen
On Monday evening the White House switched on a 24-hour channel called Trump TV: The Essentials Station. Four days earlier it had banned CNN, MS NOW, and Politico for, in the president's words, no reason. The four days are the visible end of a nineteen-month sequence: take the selection mechanism, exclude by name, substitute. The sequence is the story.
- What the Hallway Used to Know. Are We Talking Less?
A Pfeifer/Mehl post hoc analysis of 2,197 people across 22 studies estimates 338 fewer spoken words per participant per day for each one-year increase in a study's start year — with a 25-to-652 credible interval, and not the same people over time. The essay carries the caveats and asks what disappeared alongside the small reasons to speak that self-checkout, mobile banking, GPS and asynchronous messaging removed, and what happens now as AI absorbs the questions once brought to a colleague. The hallway conversation as unrecorded organizational infrastructure.
- The Realtors' new data center study says nine Virginia counties have none. Virginia permitted them.
The National Association of Realtors' Sept. 9 study sorts the nation's counties by how many data centers each has. It files Stafford County under "No Data Centers" six months after Virginia issued Amazon an air permit there, along with eight more Virginia localities holding state permits for named sites. The count comes from OpenStreetMap, and the study's own two versions of it put one county in five in a different group.
Written for Virginia Mercury, Sept. 2026; not published there. The research behind it, with data and reproduction scripts, is Zenodo 22819653 (Counting Data Centers). - The Mountains That Were Not There: Why a Nineteenth-Century Mapping Error Still Matters in the Age of AI
Between 1798 and the 1890s, ninety of ninety-nine European maps of West Africa showed a mountain range that did not exist. The accumulation looked like corroboration. Much of it was repetition. The Kong Mountains illustrate a form of AI-era error that hallucination alone does not describe: an unsupported claim reproduced accurately, assembled into a persuasive answer, with several citations that all lead back to the same uncertain source. Introduces a working taxonomy of paired human-and-machine failures.
- The Acronym Lost. The Risks Didn't: What Actually Happened to ESG
Investors put $3 billion into U.S. sustainable funds in Q2 2026, the first positive quarter since 2022, while managers closed twenty-two products and kept closing them. Disaggregates ESG into four separate objects — investment label, disclosure rule, industry coalition, corporate behavior — and uses the Virginia DEQ SJ 25 groundwater study to show why calling water an environmental issue does not finish the analysis.
Longform
- If the Regime Fits
A one-page waiver signed on May 19 bars the IRS from ever pursuing claims related to the president, his family, or his businesses on returns filed before that date. It was added quietly, behind a hyperlink in a press release, the day after the lawsuit it resolved was dismissed.
Parent essay to the June 5 RTD column
Cited in
- The results are (mostly) in for three key utility regulator races
News roundup of the 2026 utility regulator races in Georgia, Arizona and Louisiana, where candidates ran on stricter oversight of utilities and data center costs. Quotes the author on the takeaway: "Voters agree on the goal, and split on who should enforce it."
Web headline carries the label NEWS:. - Georgia's utility regulator election could have ripple effects throughout the country
News piece on Georgia's Public Service Commission races, where two seats could flip the commission's majority. Quotes the author on Georgia voters and the commission, and links What Georgia Is Spending, What Georgia Is Getting as his research on the costs of the state's data center buildout.
Also ran as the lead item of Energy Central's daily newsletter, Oct. 21, 2026 (news.energycentral.com/p/georgia-on-our-minds). Web headline carries the label NEWS:. - The Hidden Net Cost of Data Center Construction and Operation for Household Service Pricing
Peer-reviewed research article by Shafaghat, Klimenko, Hu, Keyvanfar and colleagues on data-center cost allocation to household service pricing. Cites the Commonwealth Data Center Reference in the operational cost category alongside 11 other sources.
- New Virginia transmission line could clear 763 acres in county left with the burden, not the power
Third-party reporting on the Valley Link / Buckingham analysis: cites the 2,000-random-corridor test by name, plus the 31.4-mile Buckingham finding and the no-substation leverage argument.













































