# T003 — Shell-LLC Principal Attribution for Forgelight Ventures and Willowbend Capital Two of the five Arkansas hyperscale data-center sites under investigation are developed through Delaware shell limited liability companies whose principals are not named in any primary record. [[Forgelight Ventures, LLC]] is the developer of the announced Conway data-center project; [[Willowbend Capital, LLC]] is the developer of the "Project Boar" data center at the Port of Little Rock. Both are foreign-registered in Arkansas on the same day (2025-02-11), share the same registered agent (Corporation Service Company), share the same Delaware private-mailbox address (PMB 160, 2801 Centerville Road, Wilmington), have near-consecutive Arkansas filing numbers (811535239 / 811535247), and have the same organizer of record ([[Michael Montfort]]). The wiki currently attributes both projects to **Google** — "the leading candidate" for Forgelight/Conway (qualified as not directly named in any source) and "reportedly Google's" for Willowbend/Port of Little Rock (anchored to *Data Center Dynamics* relaying *Arkansas Democrat-Gazette*'s two-source reporting). The contested question: **is the wiki's current Google-attribution framing — "leading candidate" / "reportedly" — appropriately calibrated to the available circumstantial-and-Tier-3 evidence, or does it over-claim what the corpus supports?** ## Statement A **The wiki's current Google-attribution framing is appropriately calibrated to the evidence and should stand.** For Willowbend Capital / Port of Little Rock, the *Arkansas Democrat-Gazette*'s two-source reporting (relayed by *Data Center Dynamics*, web research 2026-05-22) is Tier-3 evidence that the wiki properly identifies as such with the "reportedly" qualifier. For Forgelight Ventures / Conway, the twin-shell architecture (same registration date, same registered agent, same Delaware private-mailbox address, near-consecutive filing numbers, same organizer) plus *DCD*'s "noticeable similarities" observation between the two projects produces sufficient circumstantial weight to identify Google as "the leading candidate" — particularly with the explicit qualification preserved that "no source identifies it directly" ([[The Conway Data Center Project]] Evidence section). The wiki's reporting framework already accomplishes the work Statement B would demand: every Google-attribution claim in this corpus is anchored to its evidentiary basis (Tier-3 reporting for Willowbend; circumstantial corporate-structure inference for Forgelight) and tier-labeled in citation. The `confidence: medium` rating on both syntheses signals the same. The "Conway company remains officially 'a US-based Fortune 100 company'; the Chamber withheld the name from the City Council" ([[The Conway Data Center Project]] Caveats) records the precise gap. A reader cannot mistake the wiki's framing for confirmed attribution. The twin-shell architecture is *not* "consistent with multiple principals" in any reasonable sense. While it is technically possible that a shell-creation service could run parallel projects for different clients on the same day with the same agent and address and near-consecutive filing numbers, the inference of a common principal is the strongest reading of the corporate-structure evidence. Combined with *DCD*'s direct identification of Willowbend's principal as Google, the inference that Forgelight shares the same principal is plausible to the point where omitting the lead would actually mislead the reader by suggesting the corpus has no information on the Conway principal. For Willowbend specifically, the *Arkansas Democrat-Gazette*'s two-source reporting is the standard product of investigative journalism — named outlets, multiple sources, established beat reporting. Tier-3 evidence is exactly what the wiki's citation hierarchy contemplates for principal-attribution claims that no primary record establishes (per `AGENTS.md` Source-tier hierarchy). To require Tier-1 confirmation for every entity attribution would render the wiki silent on questions every Arkansas reader already knows the answer to from public reporting. The methodology embedded in the wiki's contamination firewall — "every factual claim cites a raw `.pdf`/`.msg`/`.docx` with page or location reference plus verbatim quote" — applies to *facts the wiki asserts*, not to *attributions reported elsewhere*. The wiki's current framing asserts that the *Arkansas Democrat-Gazette* and *Data Center Dynamics* have reported Google as the principal; that assertion is true and verifiable from the archived sources. To strip the "Google" reference entirely (Statement B's strongest form) would itself violate the wiki's evidentiary standard by suppressing what the available secondary record clearly says. ## Statement B **The wiki's "Google the leading candidate" / "reportedly Google's" attribution over-claims what the available evidence supports and should be further weakened or removed.** The corpus contains no primary record (FOIA production, court filing, SEC filing, corporate disclosure, named-source affidavit) identifying Google as the principal behind either Forgelight Ventures or Willowbend Capital. The evidence base is two layers of indirection: for Willowbend, an *Arkansas Democrat-Gazette* report relying on two unnamed sources, relayed through *Data Center Dynamics*; for Forgelight, no direct attribution at all — only the structural-similarity inference from twin Delaware shells with a shared organizer. The wiki's "reportedly Google's" framing for Willowbend functionally imports the strength of the underlying reporting into a load-bearing attribution. But the *Arkansas Democrat-Gazette*'s two-source story is itself uncorroborated by any official confirmation — Google has not confirmed, the Little Rock Port Authority director "declined comment ... citing a non-disclosure agreement" ([[The Port of Little Rock Data Center]] Open questions), and the MOU itself names only Willowbend Capital LLC. The wiki's standard for crediting *contested* secondary reporting should be higher than its standard for crediting *uncontested* secondary reporting — and the corporate-attribution claim here is contested precisely by the absence of any official confirmation. For Forgelight Ventures / Conway, the Google attribution is **purely circumstantial and rests on a single inferential bridge** — Montfort's role as common organizer plus structural similarity to a project whose Google attribution is itself thinly sourced. Twin Delaware shells with the same commercial registered agent (Corporation Service Company is one of the largest registered-agent services in the country, handling hundreds of thousands of LLCs nationally) and the same private-mailbox address (PMB 160 is a *shared mailbox* used by multiple unrelated entities) is consistent with **many alternative principals**: a single corporate-formation service running parallel filings for different Fortune 100 clients; a regional consortium; a private equity sponsor placing multiple bets across Arkansas sites; two distinct corporate parents using the same vendor stack. *Data Center Dynamics* itself "does not assert that Google is behind the Conway / Forgelight project" ([[Forgelight Ventures, LLC]] Notes) — the wiki's "leading candidate" framing is more aggressive than its own cited secondary source. The phrase "leading candidate" is doing more work than the formal hedging admits. A reader of [[The Conway Data Center Project]] — even after reading every Caveat — comes away believing Google is at Conway. That effect is **functionally identical to a positive attribution**, and it operates regardless of the formal qualifications in the surrounding sentences. The contamination firewall AGENTS.md establishes ("never present inference as established fact") is violated by functional inference even when formal hedging is preserved. The correct framing is the one the wiki uses for the cooling-water tension's parallel evidentiary problem in [[Identifying the Unnamed Cooling-Water Data Centers]]: the synthesis explicitly says "neither unnamed project is identified here, and the circumstantial lead toward AVAIO / Project Leo for the Arup project is **not** an identification." Apply the same standard here: the corpus does not identify the principal behind Forgelight Ventures; the Willowbend / Project Boar attribution is two-source secondary reporting that does not amount to a primary-record identification. Both should be reframed as **unidentified-with-leading-secondary-reporting-flag**, not "leading candidate" or "reportedly Google's." The wiki's reporting framework should require, at minimum, **named-source corroboration** for attribution claims that ride on inference. The *Arkansas Democrat-Gazette*'s two-source story names neither source nor any internal Google document; the inference for Forgelight is two layers removed from any named source at all. The wiki's "leading candidate" framing should be replaced with a clearer version: "the strongest secondary-reporting candidate is Google, but the corpus does not establish this and no source has yet named the Conway developer." ## Why it matters This tension is load-bearing for two parent synthesis pages — [[The Conway Data Center Project]] and [[The Port of Little Rock Data Center]] — both currently at `confidence: medium`. If Statement B prevails, the headline framing of both syntheses changes: Google would no longer appear as the apparent principal in the lead paragraphs, the Caveats sections would become the primary frame rather than qualifications, and downstream reporting (materials, FOIA correspondence citing the syntheses, legislative briefings) would have to recharacterize what the corpus shows about the Conway and Port of Little Rock developers' principals. The tension is also methodologically load-bearing for the wiki as a whole. **It tests where the line lies between defensible secondary-source attribution and over-attribution that functionally launders inference into apparent fact.** The same question recurs across the corpus: how aggressive should the wiki be in identifying entities behind Delaware shell architecture when only Tier-3 reporting or circumstantial corporate-structure evidence is available? The verdict on T003 sets the standard for future shell-LLC attribution decisions — including the still-unresolved cooling-water projects in [[Identifying the Unnamed Cooling-Water Data Centers]] (Arup and McClelland projects) and any future entity attributions on data-center developers that emerge in subsequent FOIA productions. This tension is **structurally heterogeneous from [[T001 - CIAC Classification of Google Generation Payments|T001]] and [[T002 - Ironwood Strategic Investment Designation Requirement|T002]]**: different parent synthesis pages (Conway / Port of Little Rock vs. Who Pays); different tension type (`attribution` rather than `factual`); different evidence base (Tier-2 corporate registry records + Tier-3 news reporting vs. Tier-1 docketed regulatory filings); different domain (corporate-attribution methodology vs. regulatory mechanism / statutory construction); different adjudicative forum (none — the wiki itself decides the reporting standard vs. the APSC ruling on regulatory questions). Per the Karpathy-Hegelion 3-domain test, this is the third tension of the iteration and the most distant from the first two in methodological terms. ## Resolution status **Status: `bracketed-because-no-Tier-1-channel-names-the-principal`** (as of [[D006 Synthesis]], 2026-06-03). The D003 reporting-standard resolution stands, but the principal-identification question is held in restraint: the dispositive count of Tier-1 Arkansas channels **naming** the Conway/Port principal is **zero**, and per D006 the documented vendor/counsel overlaps (shared organizer, shared counsel, the multi-state Montfort/Thomas pattern) are corroborating structure, not naming channels. The cluster is recorded as a **"documented multi-state Google-linked nominee apparatus"** with the principal **unidentified**. **Un-bracketing trigger:** a Tier-1 Arkansas channel naming the principal (the [[Altitude Capital, LLC]] standard). The history below records how the status evolved. ### History — the D003 verdict (2026-05-24, superseded as headline status) **Status then: `resolved-via-D003`** (per [[D003 Synthesis]], 2026-05-24, `confidence: high`). **The asymmetric verdict.** The antithesis prevails on Forgelight; the thesis partially survives on Willowbend at a downgraded posture. Per [[D003 Synthesis]] `## Verdict on tension`: "the wiki's parent syntheses should strip 'leading candidate' framing from Conway, preserve faithful documentation of the *DCD* report at Port of Little Rock with explicit Tier-3-uncorroborated framing, and apply [[Identifying the Unnamed Cooling-Water Data Centers]] as the controlling restraint precedent for both pages." **What the dialectic settled** (per [[D003 Synthesis]] `## What is resolved`): - The underlying Arkansas Secretary of State registry record is uncontested — both phases quote identical primary-record details (Filing # 811535239 Forgelight / # 811535247 Willowbend; same 2025-02-11 registration date; same CSC registered agent; same Delaware PMB 160; same Michael Montfort organizer). The dispute is about *what the registry shows is evidence of*, not what it shows. - The *DCD* / *Arkansas Democrat-Gazette* report's source posture is uncontested — the reporting is what it is: a two-unnamed-source secondary-press story about Willowbend that pointedly stops short of naming Google at Forgelight. - The [[Altitude Capital, LLC]] comparison case is the corpus's worked example for *what real evidence-led identification looks like* — three independent Tier-1 confirming channels (deed chain, water-use registration naming "Google Data Center," sworn Entergy CEO testimony, APSC Order findings). The Forgelight / Willowbend record has *zero* such channels. Altitude Capital is the standard the Forgelight / Willowbend evidence must meet, not a license for naming Google on weaker evidence. - **The [[Identifying the Unnamed Cooling-Water Data Centers]] precedent is controlling for Forgelight.** The wiki has already worked out (`confidence: low` synthesis on the Arup / McClelland projects, by the same maintainer, on the same investigation) that circumstantial leads are "**not** an identification." Methodological coherence requires applying the same restraint here. **Specific reporting-standard changes the verdict directs** (per [[D003 Synthesis]] `## Verdict on tension`): 1. **On [[The Conway Data Center Project]]:** Remove "Google the leading candidate" from the lead and Evidence section. Reframe as "developer-of-record [[Forgelight Ventures, LLC]] with undisclosed Fortune-100 principal; structural correspondence with [[Willowbend Capital, LLC]] noted but not traversed as identification, per the wiki's [[Identifying the Unnamed Cooling-Water Data Centers]] restraint standard." Downgrade confidence rating accordingly. 2. **On [[The Port of Little Rock Data Center]]:** Preserve the *DCD* / *Democrat-Gazette* reporting but reframe "reportedly Google's" as: "*Arkansas Democrat-Gazette* (via *Data Center Dynamics*) has reported, citing two unnamed sources, that the Willowbend Capital principal is Google; no Tier-1 confirmation has surfaced, Google has not commented, and the Little Rock Port Authority's executive director [[Bryan Day]] has declined to comment citing NDAs." Explicitly note that NDA-cited refusal is non-evidentiary for principal identity. 3. **Add a cross-reference on both syntheses to [[Identifying the Unnamed Cooling-Water Data Centers]]** as the methodological precedent. **What's bracketed** (per [[D003 Synthesis]] `## What is bracketed`): (a) the actual identity of the Forgelight principal — could be Google, Meta, Microsoft, Amazon, Oracle, a private equity sponsor, a hyperscale white-label operator, or another Fortune-100 entity; (b) the reliability of the *ADG*'s two unnamed sources; (c) [[Michael Montfort]]'s professional role (in-house corporate paralegal, Delaware formation-firm partner, or contracted registry technician). These bracketings are on questions adjacent to T003, not on T003 itself — the reporting-standard question is resolved; the underlying identification remains open. **What's sharper but unresolved** (per [[D003 Synthesis]] `## What is sharper but unresolved`): (a) the floor for citing Tier-3 attribution in a synthesis lead — Willowbend's "reportedly Google's" framing occupies a genuinely contested middle ground; (b) whether the parallel MOU deal terms (300K SF / $1B / 50 jobs / 65% / 30-year / Act 9 / $10B real / $50B personal) are market-standard hyperscale template wallpaper or distinctive project parameters. ## Discovery This tension was implicit in the [[The Conway Data Center Project]] and [[The Port of Little Rock Data Center]] syntheses from their 2026-05-22 creation — both pages explicitly flagged the principal attribution as not confirmed by primary records and noted the Caveats explicitly. It was filed as T003 on 2026-05-24 as the third tension of the Karpathy-Hegelion Pipeline iteration on this wiki, selected for heterogeneity from T001 and T002 (different parent synthesis pages; different tension type; different evidence base; different domain; different adjudicative forum). The tension was not previously surfaced as a discrete contested object — only as cross-cutting caveats on the synthesis pages. The methodology rationale for selecting T003 as the third dialectic: per the Karpathy-Hegelion Pipeline 3-domain test, T003 should be in the most distant possible domain from T001/T002 to (a) test the wiki's reporting framework on a structurally different question and (b) produce a third heterogeneous dialectic that can sharpen the cross-dialectic pattern analysis for a potential Phase 6 meta-dialectic. ## New input (2026-05-29): a third Montfort shell, bridging Conway and the confirmed-Google West Memphis site After D003 resolved the *reporting-standard* question, a new *evidentiary* input arrived bearing on the bracketed *underlying-identity* question (item (a) of "What's bracketed"). The 2026-05-29 [[2026-05 Conway Corporation Produces Project Stratus Records|Conway Corporation production]] surfaced a developer-side NDA entity, [[Spark Innovations, LLC]]; an Arkansas SoS registry check (Tier-2, [record](../../web%20archive/2026-05-29/sos-corp-search.ark.org/spark-innovations-llc-sos-record.md)) established it is a **third [[Michael Montfort]]-managed Delaware shell** (filing 811524872, registered 2024-11-25; same CSC agent and Wilmington DE PMB building as Forgelight and Willowbend). Two features make it more than another instance of the twin-shell pattern D003 already weighed: 1. **It is internal to the Conway project at Tier-1.** Spark is named on the Conway Project Stratus staff NDAs ([[Conway Corporation Project Stratus NDAs]]) — so the Conway project ties into the Montfort cluster through *two* entities (Forgelight as MOU developer; Spark as NDA "Company"), not one. 2. **It reaches a *confirmed*-Google site.** Spark Innovations is reported (Tier-3) as an early affiliate of the West Memphis / [[GROOT LLC]] / Project Pyramid project — which, unlike Forgelight/Willowbend's "reportedly Google," is **publicly confirmed** Google. D003 turned partly on the absence of any Tier-1 channel and on *DCD* having declined to attribute Forgelight to Google; a Montfort shell that also appears at a confirmed-Google site is a different evidentiary shape. **This does not, by itself, overturn the D003 verdict.** The West-Memphis-to-Spark association is Tier-3, the shared-Montfort pattern is still not a Tier-1 channel, and the [[Altitude Capital, LLC|three-independent-Tier-1-channel]] standard remains unmet. **Dialectic [[D005 Synthesis|D005]] ran on this question (2026-05-29) and reaffirmed `resolved-via-D003`:** the Spark finding strengthens the documented Montfort developer-overlap pattern (now three shells; two distinct Tier-1/MOU footholds inside the Conway project) but adds **zero** independent Tier-1 channels naming Google at Conway — the count that *was* D003's gap stays at zero — and the transitivity ("Google at West Memphis" + "Spark at West Memphis" + "Spark at Conway" → "Google at Conway") is invalid absent any showing that Spark is a Google-exclusive vehicle. The verdict: record the **stronger Tier-2 structural link** (the three-shell cluster; the two Conway footholds; the single Tier-3 outward edge toward GROOT) **as a structural finding, without escalating to identification**. The Conway principal remains **unidentified**; the wiki does not name Google. See [[D005 Synthesis]]. ## New input (2026-06-03): the Montfort/Thomas signature is a recurring *multi-state* Google-data-center pattern A 2026-06-03 web-research pass answered the **cluster-wide enumeration question** that [[D005 Synthesis|D005]] flagged as "the single highest-value missing fact in the entire T003 line" (Montfort's role / "how many entities carry the exact Montfort signature ... and do any attach to a *named* principal at Tier-1?"). A joint **St. Louis Public Radio / KCUR investigation (2025-09-08)** found that **[[Michael Montfort]] and "David Thomas" recur as the same two registered officers across at least seven LLCs in four states, all tied to early-stage Google data-center projects** (web research 2026-06-03, [STLPR](../../web%20archive/2026-06-03/stlpr.org/documents-link-secretive-st-charles-data-center-google-2025-09-08.md); [KCUR](../../web%20archive/2026-06-03/kcur.org/google-missouri-data-center-st-charles-2025-09-08.md)): in Arkansas, **both [[Spark Innovations, LLC|Spark]] and [[GROOT LLC|Groot]]** list Montfort + Thomas; in Missouri, **AG Rose Solutions** (a Google Kansas City build); in Indiana, **Deep Meadow Ventures** (Indianapolis — Google confirmed via WFYI records); plus two Virginia sites Google announced. **What this changes — and what it does not.** This is the first real evidence on the bracketed hinge D003/D005 could not resolve, and it cuts two ways; honesty requires recording both: - It **confirms the nominee/registered-officer reading of Montfort himself** (option B for *his* role): he is a *recurring registered officer*, not a beneficial principal — so the wiki's restraint on naming Montfort as a principal is vindicated. - But it **materially strengthens the Tier-3 inference that the Arkansas shells are Google's.** The Montfort/Thomas signature is **not** the generic, client-agnostic Corporation Service Company pattern the D005 antithesis posited ("the same logic would make every CSC client ... a Google entity"). The duo's signature specifically tracks **Google** data-center projects across four states, several **independently confirmed** as Google (Indianapolis, Kansas City, West Memphis). That is the opposite of a formation firm servicing unrelated clients — it reads as a *Google-data-center-specific* formation/nominee apparatus, of which the Arkansas [[Forgelight Ventures, LLC|Forgelight]] / [[Willowbend Capital, LLC|Willowbend]] / [[Spark Innovations, LLC|Spark]] shells appear to be one node. **Firewall / restraint holds — but the question is now ripe for re-adjudication.** This remains **Tier-3** reporting; no Tier-1 channel (deed, water-use registration, sworn testimony, regulatory order) names Google as the principal of Forgelight, Willowbend, or the Conway/Port projects. Under the controlling [[Identifying the Unnamed Cooling-Water Data Centers]] standard, the wiki **still does not assert Google as an Arkansas-established fact**, and the status stays `resolved-via-D003`. **However**, this finding directly implicates the methodological question D005 left open — "at what point does an accumulation of Tier-2/Tier-3 circumstantial links become *so* dense that the categorical restraint should yield to a calibrated 'probable principal' framing?" A multi-state, Google-specific, named-officer pattern with several Tier-1-confirmed Google nodes is materially denser than the record D003/D005 weighed. **A follow-on dialectic — [[D006 Synthesis|D006]] — was run on this question on 2026-06-03.** Verdict (call-isolated three-phase): the **antithesis prevails on *labeling*** — do **NOT** adopt a "probable principal: Google" tier, because zero Tier-1 channels name the cluster's principal, the label would launder Tier-3 inference into apparent fact, and it would silently overrule the [[Identifying the Unnamed Cooling-Water Data Centers]] restraint precedent and the [[Altitude Capital, LLC]] three-Tier-1-channel standard. The **thesis prevails on *characterization*** — the multi-state Montfort/Thomas Google-LLC pattern genuinely changed the evidentiary *shape* (it is not a client-agnostic CSC pattern). **Bounded synthesis:** sharpen the recorded pattern to a **"documented multi-state Google-linked nominee apparatus"** while continuing to hold the Conway/Port principal **unidentified**. T003's status is updated to `bracketed-because-no-Tier-1-channel-names-the-principal` — the D003 reporting-standard resolution stands; the principal-identification question is held in restraint pending a Tier-1 Arkansas channel naming the principal (the Altitude Capital standard). The synthesis-body re-characterization on [[The Conway Data Center Project]] / [[The Port of Little Rock Data Center]] is a deferred body-sync pass per the dialectic discipline. ## New input (2026-06-04): connector litigation and SEC sweep finds no Tier-1 channel naming the principal A 2026-06-04 connector pass tested the two channels most likely to *un-bracket* this tension — litigation (which can pierce a shell or name a beneficial owner) and SEC filings (which can disclose a subsidiary) — and both came back empty: - **CourtListener (federal dockets + opinions).** Searches for `"Forgelight Ventures"`, `"Willowbend Capital"`, `"Spark Innovations"`, `"AVAIO Digital"`, and `"GROOT LLC"` returned **no docket naming the Arkansas data-center entities** (the only near-matches were an unrelated 2010 Florida "Willowbend Development, LLC" case and AI-copyright dockets matched on generic terms); a search for `"Michael Montfort"` returned only two unrelated bankruptcy filings (1987 N.D. Ga.; 2011 D. Md.), neither the Delaware shell organizer; an opinions search for the entities returned **zero** results. No federal litigation names a principal of, or pierces, any of these shells. - **SEC EDGAR.** Alphabet's FY2025 10-K documents Google's data-center / AI-infrastructure investment at the parent level but names no Arkansas site, entity, or shell — the shells are private and do not appear in EDGAR ([Alphabet 10-K extract, Tier 2](../../web%20archive/2026-06-04/www.sec.gov/alphabet-10-k-fy2025.md)). **Effect on the bracket: none — it holds and is reinforced.** The dispositive count of Tier-1 Arkansas channels naming the principal stays at **zero**. These are negative findings; under the [[Identifying the Unnamed Cooling-Water Data Centers]] restraint standard and the [[D006 Synthesis]] verdict, absence of a naming channel is information, not a license to name. The status remains `bracketed-because-no-Tier-1-channel-names-the-principal`; the un-bracketing trigger (a Tier-1 Arkansas channel naming the principal) is unmet. ## New input (2026-06-08): Tier-1 Arkansas confirmation that the recurring officers are Wilson Sonsini counsel — bracket holds The [[deq/seek-pds-data-center-permits-2026-06-08/_overview|DEE-DEQ production]] produced the **first Tier-1 Arkansas** evidence on bracket item (c) — *"[[Michael Montfort]]'s professional role (in-house corporate paralegal, Delaware formation-firm partner, or contracted registry technician)"* — which [[D005 Synthesis|D005]] called the rate-limiting fact for the whole T003 line. - **Both recurring shell-officers are Wilson Sonsini attorneys.** The [[GROOT LLC]] **Delegation of Authority** (a DEQ-produced Tier-1 record) names co-officer **[[David Thomas]]** as *"Attorney at [[Wilson Sonsini Goodrich & Rosati]]"* (delegation auto-revoking if he ceases to be a WSGR employee), and [[Michael Montfort]] signs Groot's SEEK submissions and the [[Project Pyramid Title V Air Permit (2507-AOP-R0)|Title V air permit]] (as "Manager") from `[email protected]`. The Tier-3 STLPR/KCUR finding (the "New input (2026-06-03)" section above) that Montfort + Thomas recur across ~7 Google-data-center LLCs now has a **Tier-1 Arkansas anchor for the firm affiliation.** - **What it resolves.** This confirms the **nominee / outside-formation-counsel** reading of *both* officers (option toward "Delaware formation-firm" rather than "in-house"). It **vindicates the wiki's restraint** on naming Montfort/Thomas as principals, and is fully consistent with the [[D006 Synthesis|D006]] "documented multi-state Google-linked nominee apparatus" characterization. Bracket item (c) is effectively answered: they are counsel/nominees, not beneficial principals. - **What it does not do — the un-bracketing trigger is unmet.** A law firm acting as organizer/signatory is precisely the vendor relationship [[D006 Synthesis|D006]] held cannot, by itself, name the beneficial principal. **No Tier-1 channel names Google (or any hyperscaler) as the principal of [[Forgelight Ventures, LLC]], [[Willowbend Capital, LLC]], or [[Spark Innovations, LLC]].** The Groot Delegation of Authority does add a corporate-veil layer **above Groot** — sole member **[[GDH Holdings Inc.]]**, officer **[[Gabor Turschl]]** — but (i) Groot/Pyramid is *already* confirmed-Google by independent Tier-1 channels, so this is not new identification of *that* site, and (ii) the record does **not** tie GDH Holdings to Google or to the Conway/Port shells. Per Joshua's 2026-06-08 direction, GDH Holdings / Turschl are recorded **as documented, without inference**. - **Effect on status: none.** The dispositive count of Tier-1 Arkansas channels naming the Conway/Port principal stays at **zero**. Status remains `bracketed-because-no-Tier-1-channel-names-the-principal`; the un-bracketing trigger (a Tier-1 Arkansas channel naming the principal) is unmet. ## New input (2026-06-11): Google's own project site claims the Port land purchase in the first person — surfaced for maintainer review, status unchanged Google operates an official first-party project website for the Port of Little Rock site, **littlerockdatacenter.com**, whose archived text states: **"In April 2026, Google confirmed that we have purchased land at the Port of Little Rock for a potential data center"** — with the project described as in an "Evaluation" stage and "[a] final decision to build is only made after all diligence and public reviews are complete" (web research 2026-06-11, [littlerockdatacenter.com capture](../../web%20archive/2026-06-11/www.littlerockdatacenter.com/google-littlerock-purchase-confirmation-2026-06-11.md)). On 2026-06-10 Google also held an invitation-only public meeting on the project, with public affairs representative Laurel Brown stating "[t]here are a number of permits that still need to be filed" ([[2026-06 Google Holds an Invitation-Only Meeting on the Port Data Center]]). - **What's categorically new:** every prior Willowbend↔Google channel was third-party (DCD/ADG reporting, registry inference, vendor overlap). This is the **putative principal identifying itself, in its own published first-person voice, with the specific land whose recorded deed runs to [[Willowbend Capital, LLC]]** (Pulaski County Inst. # 2025033561, $11,505,600). It is no longer accurate to describe the Willowbend–Google link as resting only on "uncorroborated third-party reporting." - **What it is not:** a Tier-1 Arkansas record channel. A company website is Tier-3 under the source hierarchy, and the statement does not name Willowbend, explain the nominee structure, or constitute an Arkansas public record. The D006 un-bracketing trigger — *a Tier-1 Arkansas channel naming the principal* — remains formally unmet. - **Disposition (per the D006 discipline — surface, don't auto-un-bracket):** recorded as a new input; **status unchanged** at `bracketed-because-no-Tier-1-channel-names-the-principal`. **Flagged for Joshua:** this input plausibly warrants a follow-on dialectic (prospective **D007**) on whether a first-party public self-identification changes the Willowbend framing even without a Tier-1 record channel — and, separately, whether the trigger definition itself should distinguish "no channel names the principal" from "the principal has publicly named itself outside the Arkansas record." Note the Forgelight/Conway and Spark questions are **unaffected** — the site statement concerns the Port land only. ## New input (2026-06-12): the item-6 supplemental — three additions, no principal named, bracket holds The [[deq/pre-permitting-correspondence-2026-06-11/_overview|DEE-DEQ item-6 pre-permitting supplemental]] (78 files) was swept for hyperscaler-principal naming; **no record names Google or any hyperscaler as the beneficial principal of Forgelight, Willowbend, or Spark** — the un-bracketing trigger is not met. Three additions bear on the tension: - **Willowbend formally certified operational control.** Its Individual 401 application answers **Yes** to "Does the Applicant… have the primary management and ultimate decision-making responsibility over the operation of the activity?", with [[Michael Montfort]] signing as "Manager" (after signing the February pre-file as "Counsel"), and describes the project only as a "**light industrial complex**" ([[Project Boar CWA 401 Pre-File and Application]]). The shell's own filings thus assert the very operational primacy that Statement A doubts — strengthening the *formal* record of shell-as-principal while the practical-control question stays open. - **State officials pressed the opacity and accepted the nominee answer.** DEQ's air-permit engineer refused to draft Groot's permit while the disclosure "does not list anyone as officers or owners," then accepted "Manager" for Montfort ([[Groot Air Permit Pre-Draft Correspondence]]) — documenting that even the regulator could not get past the nominee layer for the *confirmed-Google* entity. - **Google personnel appear inside the named entity's file, not the bracketed ones.** [[David Essex]] (`[email protected]`) rides Groot's permit correspondence; [[Stuart C Hindmarsh]] appears a third time, as Groot's "Attn:" on USACE notice MVM-2024-216 with a (479) number. Both concern **Groot** (already independently Google-confirmed) — corroborating structure for the apparatus, not channels naming the Forgelight/Willowbend/Spark principal (per the [[D006 Synthesis]] channel/structure distinction). **Disposition:** status unchanged at `bracketed-because-no-Tier-1-channel-names-the-principal`. The prospective-D007 question flagged in the 2026-06-11 input remains with Joshua. ## Notes - Statement A and Statement B both engage the same evidentiary base — [[The Conway Data Center Project]] Evidence section, [[The Port of Little Rock Data Center]] Evidence section, the Arkansas Secretary of State entity records for both LLCs, and the *Data Center Dynamics* report relaying the *Arkansas Democrat-Gazette*'s reporting. The contestation is over **what standard of evidence justifies what strength of attribution**, not over what the underlying record contains. - The Willowbend attribution and the Forgelight attribution are at **structurally different evidence levels**. Willowbend has direct Tier-3 reporting; Forgelight has only inferential bridges. The dialectic should consider whether a single verdict should cover both, or whether the syntheses' framings should be calibrated separately to their distinct evidence levels. - No HSPI-sealed material is at issue. Unlike T001, the dialectic can engage the full evidentiary record without an HSPI bracket. - The closest parallel in the corpus is [[Identifying the Unnamed Cooling-Water Data Centers]] — where the synthesis explicitly declines to make an identification ("the circumstantial lead toward AVAIO / Project Leo for the Arup project is **not** an identification"). T003 asks whether that same restraint should apply to Forgelight and (in a weaker form) Willowbend. - Unlike T001 and T002, T003's resolution path runs *inside the wiki* (a reporting-standard decision) rather than primarily through an external forum (a Commission ruling). This is itself a heterogeneity vector relevant to the 3-domain test.