McAleavey's Waterless Data Center: How A Neighborhood Has Begun Owning It's Own Intelligence

The Provocation

Hyperscale data centers in Arizona consume 300 million gallons of water annually. That water comes from aquifers that don't belong to them. The cost is externalized to communities the company will never see.

Mapleton Fall Creek's data center requires no water.

Not because we're trying to be green. Because we're trying to be honest.

When your neighborhood owns the infrastructure that runs your intelligence, the constraints become visible. The cost becomes negotiable. The decision becomes yours.

In Phase 1 (API-based Claude processing), our data center uses approximately 500 gallons per month for maintenance and cleaning. Negligible. Invisible to the neighborhood's water budget.

In Phase 2 (if residents choose local inference on open-weight models), the answer is the same. A four-GPU cluster draws about 1.5 kilowatts. The earth-sheltered vault sheds that heat through soil-coupled earth tubes with roughly twenty times the capacity the cluster needs. Any liquid cooling is a closed loop: the same water circulating, like a car radiator, consumed at zero gallons a day. Hyperscale centers use water because they evaporate it to cool megawatts. We are not cooling megawatts.

We don't hide that number in a corporate sustainability report. We don't pretend efficiency where there is extraction. We put the water, the energy, and the cost on the table and ask residents: Is this worth it?

That question changes everything.

What This Is

Mapleton Fall Creek Neighborhood Narrative Data System: A locally-controlled infrastructure that runs large language models — Claude by API in Phase 1, open-weight models developed and hosted locally in Phase 2 — to process neighborhood data, continuously generate census knowledge, drive Pro Se democracy, and make infrastructure decisions visible and accountable at neighborhood scale.

What This Is NOT

Not a smart city pilot. Those extract data and disappear.

Not research. This is residents owning the infrastructure that serves them.

Not charity. This is civic infrastructure, like a water main or a power grid, except residents own and control it.

What It Does

1. Collects Data From Hearts of Homes

Residents voluntarily share what they want the neighborhood to know: health status, housing conditions, economic situation, who they depend on and who depends on them, aspirations, connections. All voluntary. All encrypted. All staying in Mapleton Fall Creek.

2. Processes Through Language Models (Locally Controlled)

An AI language model analyzes the data and generates insights. In Phase 1 that model is Claude, reached by API with encrypted queries. In Phase 2 it is an open-weight model developed and hosted inside the neighborhood. Either way, every output is classified through Norvellian AI Ethics.

3. Powers Continuous Neighborhood Census

Not decennial federal counting. A living, updated understanding of Mapleton Fall Creek: who lives here, what they have, what they need, how they're connected, what's changing. Updated monthly. Accessible to residents. Driven by residents. Owned by residents.

4. Drives Pro Se Democracy

Residents see the census. Residents ask the system: Why is this happening? What would happen if...? Residents propose decisions for the neighborhood. Residents debate based on data. Residents vote. Residents implement. Residents evaluate.

5. Makes Infrastructure Visible

Water usage. Energy sources and costs. Data residency. Who can access what data. All transparent. All negotiable. All decided by residents.

The Technical Reality

Option A: API with Local Processing (Phase 1)

Claude API (cloud-based) processes queries. Results returned to Mapleton Fall Creek. Classified locally. Stored locally. Residents access locally.

Advantages	Lower cost, lower technical requirements, fast deployment
Disadvantages	Data leaves neighborhood (but encrypted)
Water usage	Negligible (~500 gal/month)
Cost	$1.2M infrastructure, $425K annually
Timeline	12–18 months to operational

Option B: Local Inference (Phase 2+, if chosen)

GPU cluster (RTX 6000 Ada × 4) runs an open-weight LLM — Llama, Mistral, Qwen, or a model the neighborhood fine-tunes itself — locally. All computation stays in Mapleton Fall Creek. Maximum data privacy.

Advantages	All computation local, maximum data residency, no model provider sees resident queries
Disadvantages	Higher technical requirements, more complex; open-weight models require local tuning and evaluation
Water usage	None consumed. Closed-loop or earth-tube cooling; ~1.5 kW load against 32–40 kW of passive capacity
Cost	Add $34–36K GPU cluster to Phase 1
Timeline	Phase 1 operational first, Phase 2 decision after 12 months
Governance	Residents vote on whether to upgrade. That vote includes the water and energy question.

Coiled ground-loop heat exchanger piping laid in an open trench before being covered with soil. — Soil coupling, before the dirt goes back. A ground loop in a trench — the ordinary, decades-old technology the earth-sheltered vault uses to shed heat into Indiana ground instead of evaporating water into Indiana air.
Photograph: Mark Johnson · Wikimedia Commons · Public domain

The Economics

Infrastructure Cost (One-time)

Phase 1 (API-based)

Compute & storage: $195K
Physical space & security: $255K
Software & licensing: $130K
Professional services: $430K
Contingency: $198K
Total: $1.2M

Operational Cost (Annual)

Category	Year 1	Year 2+
Personnel	$170K	$155K
Infrastructure & utilities	$102K	$95K
Software & services	$219K	$205K
Administration	$89K	$85K
TOTAL	$425K	$390K

Per-Capita Cost

Mapleton Fall Creek (10,000): $42.50/capita annually

Why this matters: Data brokers currently extract $50–500/person/year from neighborhoods. Mapleton Fall Creek's cost keeps the value in the neighborhood.

Scaling Economics

Scale	Population	Annual Ops	Per-Capita
Neighborhood (MFC)	10,000	$425K	$42.50
District	50,000	$950K	$23
Mid-city	200,000	$2.8M	$14.25
Large city	1M+	$7.5M	$7.50

Indianapolis citywide at $7.50/capita = $7.5M annually to own the entire city's data and intelligence. Currently, that data flows to corporate servers.

Governance: Who Owns This

Core Principle

Residents own the data and make the decisions. Institutions support but don't control.

Governance Board (11–15 members)

8–10 residents (elected by residents, 2-year terms)
1–2 institutional representatives (IU Indianapolis, health, nonprofit)
1 data steward (technical leader, hired, reports to residents)

Decisions Made Locally

What data is collected
How long data is retained
Water usage and infrastructure choices
Model usage and outputs
Budget and staffing
All spending is public; all decisions are transparent

Accountability

Public reports quarterly
Annual independent audit
Resident feedback sessions monthly
Residents can remove board members

Norvellian AI Ethics: The Framework

This system is built on Thomas Jefferson's 1807 letter to John Norvell about entering journalism with honor. Jefferson told Norvell the truth about newspapers: they had become poisoned vehicles for falsehood. The remedy: restrict them to true facts and sound principles only. We're doing the same with AI.

The Four Categories

TRUTH

Verified fact from reliable data. "10,342 residents in Mapleton Fall Creek" (verified count). "34% report difficulty accessing healthcare" (survey data). High confidence.

PROBABILITY

What evidence suggests. "Healthcare access is improving in neighborhoods with new clinics" (trend analysis). "Workforce expansion likely increases job placement" (historical pattern). Medium confidence.

POSSIBILITY

What could happen. "If we invest $500K in youth programs, completion rates might rise 10–15%" (scenario). "Additional housing stock could stabilize displacement" (conditional). Low confidence.

ABSURDITY

What is false. "Every resident earns six figures" (contradicts data). "No health disparities in Mapleton Fall Creek" (contradicts reality). No confidence.

Too many AI systems hide uncertainty. They present models as facts. They bury confidence scores in algorithms. Residents get confused. Trust breaks down. Norvellian ethics means: We tell you what we know, what we think, what's possible, and what's false. You decide based on real information.

Pro Se Democracy in Action

Pro Se means "for oneself." Residents represent themselves.

The Cycle

RECOGNITION (Monthly)

Neighborhood census updated. All outputs classified. Residents review: "Here's what we know about ourselves."

QUESTIONING (Ongoing)

Residents ask the system: "Why is this happening? What would happen if...?" Questions and answers posted publicly. Shared understanding builds.

PROPOSAL (Quarterly)

Residents propose decisions: "Expand clinic hours." "Invest in street trees." "Use rainwater for data center cooling." Debate is visible.

VOTING (Quarterly)

Residents vote. Simple majority. Binding for neighborhood spending/advocacy.

IMPLEMENTATION (Ongoing)

Winning proposals get implementation team (staff + volunteers). Progress tracked. Residents see results.

EVALUATION (Quarterly + Annually)

Did it work? What changed? What would we do differently? Feeds into next cycle.

What Exists in Mapleton Fall Creek Already

We're not building from zero.

Institutional Anchors

IU Indianapolis School of Medicine (clinics across neighborhood)
IU School of Social Work (research, students, partnerships)
Community health centers (Eskenazi, others)
Faith institutions (decades of presence)

Community Assets

Active neighborhood associations
Faith-based leadership
Health worker networks
Civic engagement programs

Relationships

Joy Repair partnerships
Zawadi Exchange
Health by Design
Indiana Public Health Association

The Ask: Integrate the data system into existing work. Make existing relationships visible, connected, resident-controlled.

10 Takeaways That Cause Good Trouble

1. Corporate Data Extraction Stopped

Data brokers extract $500K–5M annually from MFC's 10,000 residents. This system: all value stays in neighborhood, paid to local staff. Provocative question: "Why should residents subsidize corporate profit from data about themselves?"

2. Water Becomes Democracy

Data center water usage is visible and negotiable. Hyperscale operators evaporate hundreds of millions of gallons a year and call it efficiency. Ours consumes none, and residents can check. Whether to add local inference — its energy, its cost, its cooling — is a resident vote, not a company decision. This is real governance.

3. Intelligence Stays Local

A 10,000-person neighborhood can run its own LLM system. You don't need Silicon Valley. You don't need corporate permission. Provocative: "If neighborhoods can run their own AI, why allow corporations to monopolize it?"

4. Census is Continuous, Not Decennial

Federal census every 10 years. Mapleton Fall Creek census monthly. By the time federal count arrives, MFC has made 120 months of decisions on better data.

5. Institutions Become Accountable in Real Time

IU Indianapolis, health system, nonprofits know residents have data. Know decisions are transparent. Know they answer to residents who can see whether they're keeping promises.

6. $42.50/Capita Buys Sovereignty

MFC infrastructure costs $42.50/resident annually. Data brokers charge $50–500/person/year for your data. You're paying twice—once in privacy, once in profit. This inverts it.

7. Scales to $7.50/Capita Citywide

One neighborhood: $42.50/capita. City of 1M: $7.50/capita. Economical. Sustainable. Provocative: "Indianapolis could own citywide intelligence for $7.50/person annually. Why doesn't it?"

8. Pro Se Democracy Becomes Real

Residents don't send representatives. Residents make arguments based on real data. Residents vote. Residents implement. This is not town halls. This is people representing themselves.

9. Norvellian Framework Breaks AI Theater

Every output classified: Truth, Probability, Possibility, Absurdity. No hidden scores. No black box. Residents see exactly what the system knows. Provocative: "Why do corporate AIs hide uncertainty while community AIs are transparent?"

10. Replication Path is Open

Mapleton Fall Creek documents everything. Code is open. Governance shared. Other neighborhoods build this. Other cities federate it. Within five years: citywide or regional systems where communities own their data. The monopoly breaks.

Six Non-Negotiable Firewall Clauses

All grant agreements must include these clauses. No funder retains veto authority over resident decisions. No funder gets data access without resident vote. These clauses define what genuine partnership looks like.

Data Sovereignty

Resident Ownership: All data collected from residents of Mapleton Fall Creek is the property of those residents. The Funder retains no ownership claim, access right, or usage license to resident-level data under any circumstances.

Funder Non-Access: The Funder shall have NO access to resident-level data (identifiable or re-identifiable), encrypted or decrypted, raw or processed, without explicit written consent from the Data Firewall Trust AND prior approval from the Governance Board AND ratification by resident referendum (simple majority).

Enforcement: If the Funder attempts to access resident-level data without explicit approval, the Grant Agreement is immediately terminated, all Funder funds are returned, and the matter is referred to the Mapleton Fall Creek Data Firewall Trust for legal action.

Governance Autonomy

Resident Decision Authority: Residents of Mapleton Fall Creek, voting through the Governance Board and resident referendum process, retain exclusive authority over all substantive decisions including data collection, usage, infrastructure, partnerships, and evolution.

Funder Non-Interference: The Funder shall NOT require changes to governance structure, demand board representation, impose conditions on resident voting, veto resident decisions, or withhold funding contingent on governance changes.

Resident Primacy: If the Funder believes a decision violates grant terms, it is first reviewed by the Governance Board and residents. If residents confirm the decision stands, the Funder accepts the outcome or terminates funding.

Resident-Defined Success Metrics

Funder-Imposed Metrics Prohibited: The Funder shall NOT impose predetermined Key Performance Indicators (KPIs), success metrics, or evaluation criteria as conditions of grant funding.

Resident Metric Definition: Residents of Mapleton Fall Creek, through the Governance Board and annual community input process, shall define what success looks like, how success is measured, what evidence is collected, and how often evaluation occurs.

Metric Evolution: Metrics are revisited annually. If residents change what they value, the metrics change. The Funder accepts evolving definitions of success.

Research Ethics & Autonomy

Research as Separate Transaction: Any research conducted using Mapleton Fall Creek data or residents is a separate project requiring independent approval.

Research Approval Process: For each proposed research project: researcher submits protocol to Governance Board, research is publicly posted for 30-day community review, Governance Board votes and recommends, resident referendum vote (simple majority required for approval).

Resident Co-Authorship: Any publication using data collected from Mapleton Fall Creek residents must include resident co-authorship or explicit resident consent to anonymous publication.

Benefit Sharing: If research generates publications, intellectual property, or commercial value, residents are entitled to equitable benefit-sharing.

Sunset & Exit Rights

Initial Grant Period: All grants to Mapleton Fall Creek are structured as 5-year agreements (or shorter if specifically negotiated).

Mandatory Renegotiation: At Year 3 of any 5-year grant, both parties must meet and decide: continue, modify, or end the partnership with 90 days notice.

Resident Vote on Continuation: At Year 3 (and Year 5 conclusion), residents vote: Should we continue this partnership? Should we modify terms? Should we end it? The Governance Board respects resident decisions.

Community Exit Rights: Mapleton Fall Creek may terminate the partnership with 90 days notice if the Funder violates firewall clauses, attempts to access data without approval, imposes governance changes, or if residents vote to end the partnership. Upon termination, the community retains all data, retains all equipment, and continues operations independently.

Resident Consent for Grant Acceptance

Grant Acceptance Vote: No grant agreement is signed by the Mapleton Fall Creek Governance Board until residents have voted to accept the grant.

Voting Process: Proposed grant is posted on neighborhood website for 30 days. Terms and firewall clauses are explained in plain language. Residents attend optional info sessions. Resident vote (simple majority required). Only after vote, if approved, does board sign grant agreement.

Resident Authority: Residents are the grant acceptance authority. The Governance Board is obligated to honor resident decisions on funding.

Five-Vulnerability Stress Test

Before publishing this plan or negotiating with institutional partners, know where you're exposed. Five major vulnerabilities with the specific defenses addressing each one.

Vulnerability 1: Legal Exposure

The Threat: Can government or law enforcement seize resident data? Subpoenas, warrants, immigration enforcement.

Defenses

Encryption & Key Fragmentation: AES-256 encryption; Shamir's Secret Sharing (5 shards, 3-of-5 required). No single entity can decrypt data alone.
Trustee Independence: Trustee files motions to quash, seeks narrowing of scope, delays through legal process.
ICE Firewall (Absolute): Immigration enforcement requests receive absolute refusal on constitutional grounds. No data, no cooperation.
Public Transparency: Trustee publishes subpoena and defense strategy. Residents know what's happening.

Realistic Outcome: System is far more defensible than corporate data centers. Protection is substantial and built on legal precedent, but not absolute. After all appeals exhausted, Trustee must comply with valid court orders (maintaining legal defensibility).

Vulnerability 2: Hardware Depreciation & Equipment Failure

The Threat: Expensive hardware fails or becomes obsolete. GPU clusters cost $34–36K; become obsolete in 3–5 years. Unexpected failure requires immediate replacement.

Defenses

Hardware Reserve Fund: 15–20% of annual operations budget ($85K/year) accumulates in dedicated account. By Year 10: $633K reserve after all spending.
Modular Phased Deployment: Deploy in phases aligned with hardware cost decline. Moore's Law ensures Year 5 hardware costs 30–40% less than Year 1 for same/better performance.
Federated Redundancy: Partner with 2–3 neighboring communities to share infrastructure, reduce per-neighborhood cost, replicate data with encrypted mirrors.
Insurance & Emergency Fund: Equipment insurance covers replacement; $97.5K emergency fund covers temporary operations without fundraising.

Realistic Outcome: System is financially sustainable through Year 10 with high confidence. Beyond Year 10, depends on continued Moore's Law benefits and community fundraising.

Vulnerability 3: Physical Disaster & Security

The Threat: Earth-sheltered vault could be flooded, damaged, or destroyed. Catastrophic data loss if redundancy doesn't exist.

Defenses

Distributed Data Replication: Data replicated across 3–4 partner neighborhoods with independent key shards. If MFC vault destroyed, data exists at Fountain Square, Pendleton, Garfield Park.
Backup Power & Cooling: 50kW solar array + battery storage. Natural gas backup generator. Passive earth cooling provides 32–40 kW capacity (most of what's needed). Power outages don't immediately threaten data.
Disaster Runbooks: Documented procedures for each scenario. Resident volunteers trained in data restoration. Quarterly disaster simulation exercises.
Insurance & Emergency Funding: Equipment insurance (full replacement cost). Business continuity coverage.

Realistic Outcome: System has excellent protection against common failures and good protection against regional disasters. Cannot guarantee protection against unprecedented catastrophic scenarios.

Vulnerability 4: Philanthropic Capture & Funder Control

The Threat: Funders demand data access, governance control, or impose success metrics. System becomes dependent on funder approval. Resident autonomy erodes gradually.

Defenses

Six Firewall Clauses: All grant agreements must include non-negotiable clauses: data sovereignty, governance autonomy, resident-defined metrics, research ethics, sunset/exit rights, resident consent.
Financial Independence: Trustee funded independently. Hardware reserves built from operations. By Year 4–5, system is financially sustainable without external funding.
Multiple Funder Strategy: No single funder provides more than 50% of budget. If one funder withdraws, system adapts.
Public Transparency: All grant agreements published. Quarterly funder reports public. Annual independent audit. Residents can vote to end partnership if funder misbehaves.

Realistic Outcome: System has explicit, contractual protection against overt funder capture. Requires active resident vigilance against subtle funder influence. Funder independence grows substantially from Year 1–5.

Vulnerability 5: Tyranny of the Vocal & Power Concentration

The Threat: Same highly engaged residents dominate all votes. Low-income or immigrant residents don't participate. Decisions serve vocal minority. Hidden power concentration.

Defenses

Three-neighborhood quorum (forces broader deliberation)
Roving listening circles (reach residents where they gather)
Sortition board selection (random rotation prevents entrenchment)
Rotating meeting formats & times (accommodates different schedules, languages, mobility)
Silent-third protocol (explicitly ask who hasn't spoken)
Faction documentation (make power distribution visible)
Recall & restorative justice (residents can remove unaccountable leaders)
Participation audits (track who's missing, identify barriers)
Turn-taking & conflict resolution (no one dominates debate)

Realistic Outcome: System has explicit, multi-layered protections against vocal-minority capture. Requires sustained commitment to implement. Even with protocols, participation will never be perfectly equal. Underlying inequities (class, race, language) still shape who can participate. Protocols create conditions for broader participation; residents must choose to use them.

Overall Resilience

This system is significantly more resilient than typical community tech projects. It has explicit legal protections, financial sustainability plan, redundancy and disaster recovery, funder constraints, and anti-capture mechanisms. It is not invulnerable—vulnerabilities remain—but it does not pretend perfect protection. It acknowledges vulnerabilities, builds explicit defenses, and creates mechanisms for ongoing vigilance and adaptation.

Appendices & Supporting Documents

Three detailed charters provide legal, financial, and governance frameworks. Each is published here in full — readable, linkable, and searchable — for institutional negotiation, legal review, or deep understanding of the model.

Appendix 1: Data Firewall Trust Charter

Complete legal framework establishing the trust as a separate entity holding resident data in perpetuity. Covers Trustee role, fiduciary duty, encryption and key management, legal defense protocols, immigration enforcement firewall, data retention and deletion, trustee succession, financial security, and governance board relationship.

This document is the backbone of data protection. Share with civil rights counsel and institutional legal teams.

Read the full document →

7 articles. Legal register. For legal review and institutional negotiation.

Appendix 2: Hardware Reserve Fund & Capital Lifecycle

Financial framework for managing hardware depreciation across a 10-year lifecycle. Year-by-year reserve accumulation schedule, phased GPU deployment, Moore's Law benefits, federated hardware sharing, patient capital covenants, and sustainability beyond Year 10.

This document proves financial sustainability. Share with finance teams, funders, and residents voting on hardware investments.

Read the full document →

9 sections. Financial register. For funder negotiation and resident budget decisions.

Appendix 3: Funding Charter & Philanthropic Firewall Covenant

Funder management framework: foundational principles, the six firewall clauses as insert-ready grant language, approved and prohibited uses of funder money, mutual accountability, and the template grant agreement amendment.

Share with all funders. Use the firewall clauses verbatim in every grant agreement.

Read the full document →

Public sections only. Internal negotiation strategy is held by leadership and is not published.

Technical Appendix: Earth-Sheltered Server Vault Blueprint

Detailed architectural design for the underground cooling vault: soil coupling, passive airflow systems, thermodynamic analysis, convivial tech framework, and implementation roadmap, including specifications for 12–16 feet underground installation using Indiana earth-tube cooling.

For engineers, architects, and deep technical review. Not required for funding conversations but essential for Phase 1 construction planning.

In preparation.

Contact & Next Steps

Mapleton Fall Creek Narrative Data System. Neighborhood-controlled data center. Continuous census. Pro Se democracy. Community data sovereignty.

This proposal is offered freely for use, modification, and adaptation. Community ownership, not corporate extraction. Shared.

All firewall clauses non-negotiable in funder agreements.

Marc McAleavey · Request a conversation →

Document version 1.0 · September 2026 · Published by Joy Repair · Neighborwood · Dispatches