Field Notes • HyperQ deployment in PJM’s Cycle 1

This operational velocity is critical because PJM's reformed cluster process introduces two systemic constraints: compressed timelines and a strict "first-ready, first-served" framework that elevates site control to a primary gatekeeper node. In Cycle 1, PJM is managing an unprecedented surge, processing 811 generation applications representing massive potential capacity while staff concurrently handles other phases in legacy queues. According to Tapestry’s customer research into traditional regional transmission organization (RTO) workflows, a traditional, manual site control Application Readiness Review has historically required multiple hours of evaluation per application, scaling higher for more complex applications due to the constraints of sequential human review [2].

HyperQ's parallel processing compresses this initial screening window down to minutes per application. By evaluating computational runtime against historical baselines, HyperQ demonstrates a 20x+ acceleration for the site control portion of the Application Readiness Review phase across similarly complex cohorts. This structural velocity predictably flattens the timeline and reduces the linear administrative bottleneck when hundreds of projects drop simultaneously [3].

As application volumes rise to meet escalating energy demand, grid operators face unsustainable surges that—without scalable technology—require linear expansions of manual effort and processing timelines. HyperQ addresses this challenge by providing an AI-assisted solution that can scale for modern grid administration. Drawing on Stanford economist Chad Jones’s research regarding systemic “weak links,” HyperQ targets one of the primary upstream bottlenecks of the interconnection application process with agentic AI acceleration. 

As a result, the system can isolate non-viable projects early, preventing them from cascading downstream and reducing administrative friction across the full lifecycle. This streamlined review process will allow shifting additional focus to improving the ability of grid operators to examine improvements to other aspects of application processes.

While Tapestry's approach to comprehensive interconnection reform remains a multi-layered effort, HyperQ’s deployment serves as the industry’s first baseline validation of agentic AI performing at the scale required for the largest and most complex electricity market in North America.

Footnotes

1. Measures HyperQ's agentic execution window, specifically tracking advanced core processing milestones including document summarization, agreement and parcel relationship mapping, structural requirement evaluation, and visual citation extraction. This algorithmic runtime reflects pure computational analysis and excludes initial document transmission latencies as well as subsequent human validation or final regulatory determinations.
2. This directional analysis illustrates broader operational trends and historical orders of magnitude rather than an identical, localized baseline comparison. Manual benchmarks reflect specialized external evaluation time frames during previous lifecycle phases, whereas HyperQ metrics isolate the baseline algorithmic processing and system compute runtimes for the initial data ingestion layer. Because this architectural comparison evaluates a historical human workflow against a parallel computational layer across distinct points in time, it represents a macro-level capability trend within the Application Readiness Review phase and does not represent total end-to-end grid interconnection timeline or final human determination reductions.
3. Calculated by comparing the conservative lower-bound manual anecdotal historical baseline (4 hours / 240 minutes) against the median algorithmic runtime of 6.25 minutes.

Regulatory & structural context: Interconnection queue evolution

The length of the interconnection process is one of the most binding structural constraints on the modern electrical grid. Legacy processes have struggled to match the reality of unprecedented load growth, as engineers spend months painstakingly combing through applications by hand while surging demand outpaces the interconnection of new generation resources.

Tapestry demonstrated how agentic AI can streamline initial data processing and perform complex, multi-layered compliance analysis at scale

PJM Interconnection, North America’s largest wholesale electricity market and grid operator, recently initiated Cycle 1 of its newly reformed interconnection queueA waiting list of energy projects seeking to connect to the power grid.. This transition shifts PJM’s process from “first-come, first-served” to “first-ready, first-served” and represents an operational evolution, scaling to meet unprecedented application volumes within accelerated timelines. To assist in validating site control—the foundational prerequisite of ensuring applicants have control over the land parcels where they plan to build generating facilities and interconnect with the grid—PJM deployed Tapestry’s HyperQ.

The historic workload compression challenge

The magnitude of HyperQ’s deployment highlights the operational stress confronting modern grid administrators. PJM’s Cycle 1 encompasses 811 generation applications, representing 220 GW of potential generation capacityThe maximum amount of electrical power a power plant can generate at a given time.. To put this into perspective, the entire existing, installed nameplate generation capacityThe maximum theoretical power output a generator is designed to output. of the PJM grid—built out over nearly a century of utility engineering—stands at approximately 180 GW. PJM operators must ingest, sort, and validate a larger volume of potential generation capacity than exists across their entire transmission footprint today, in just a few weeks.

Tackling this historic surge in application volume under a highly compressed calendar timeline, HyperQ successfully executed 9,312 parallel tariff readiness assessments at a median algorithmic processing speed of 6 minutes and 15 seconds. This means HyperQ simultaneously compared the site control provisions in new applications against the region’s tariff rules and user manuals, analyzing and flagging deficiencies so that the operator’s expert reviewers can focus on final determinations.

Establishing site control is the critical baseline of the entire interconnection pipeline; if developers cannot secure the proposed land, subsequent infrastructure modeling and technical studies cannot proceed effectively. PJM’s live deployment of HyperQ signals a scalable approach for queue administration at large, demonstrating how an intelligent software layer can synthesize and pre-validate massive amounts of data across sprawling portfolios of disparate real estate agreements.

This shift toward concurrent computational review provides an operational model for navigating the current energy boom, supporting finite human resources as they handle expert legal and engineering decisions.

The shift to "first-ready, first-served"

The U.S. Federal Energy Regulatory Commission (FERC), which oversees regional electricity markets like PJM, originally designed its open-access transmission framework in 1996 to manage what was then a lower-frequency stream of interconnection applications for centralized power plants. During that era, PJM received fewer than 15 requests annually, and the entire U.S. saw only a few hundred requests a year.

The generation mix has evolved since then; today, operators must navigate retirements of those large, centralized resources as well as an influx of distributed resources and inverter based resources. Increased complexity means every application requires checking fragmented property deeds and independent mandates against strict Open Access Transmission Tariff (OATT) criteria.

To keep pace with this evolution, the industry has fundamentally shifted from the legacy “first-come, first-served” serial process to a “first-ready, first-served” cluster framework, outlined in FERC Order 2023. Many regional operators' transition cycles entail their own queue reform, making site control data readiness a prerequisite before an asset can advance to interconnection to the physical grid.

Strategic sequencing: Why HyperQ prioritizes site control

To maximize systemic efficiency, HyperQ targets site control validation as its primary "Gatekeeper Node." This strategic sequencing focuses agentic capabilities to what Stanford economist Chad Jones identifies as the system’s most crucial ‘weak link’. By eliminating early structural vulnerability, HyperQ prevents manual backlogs from compounding downstream. 

When parallel AI agents tackle the primary compliance gate first, the system isolates non-viable projects before proceeding onto more technical reviews.

Prioritizing this specific phase relies on three core operational principles:

• Eliminating speculative "phantom capacity": There is no functional advantage to running complex simulations on a generation project that lacks legal rights to the underlying land footprint. Enforcing strict validation at Step Zero filters out speculative placeholder applications immediately, ensuring that only viable projects proceed.
• Capturing AI’s comparative advantage: While electrical models follow rigid technical schemas, land rights evidence is heavily unstructured and legally heterogeneous (encompassing deeds, lease agreements, site plans with Geographic Information System coordinates, and local amendments). Document variability often leads to cascading human review backlogs, whereas agentic AI excels at this exact task: extracting intent, entities, and geographic boundaries from complex PDFs.
• Accelerating speed-to-power: With commercial and data center electricity demand escalating exponentially across the territory, capacity injection is a macroeconomic priority. Developers with secured land rights represent the assets closest to breaking ground; prioritizing this gate fast-tracks projects most likely to achieve commercial operation.

Distributed agentic architectureSoftware using specialized models to autonomously execute multi-step workflows. & system infrastructure

Engineering intelligence for grid-scale complexity

HyperQ’s site control pre-validation functions as an intelligent software layer purpose-built for complex legal and structural reasoning, designed to absorb the grid's data-intake constraints. Tapestry's high-throughput architecture deploys a distributed, multi-tiered workflow of specialized AI agents capable of navigating information-dense regulatory and technical submissions.

Overcoming linear human constraints, the system architecture relies on four primary engineering pillars:

• Two-dimensional parallelization: Executes high-throughput evaluation across multiple applications concurrently, while internally parallelizing individual document summarization and requirement-specific legal analysis.
• Agreement bundling: Algorithmically groups fragmented contracts, amendments, and cross-references into unified, legally coherent bundles.
• Deterministic control logicFixed code rules forcing an AI to follow a predictable execution path.: Distills complex institutional domain knowledge into a multi-stage agent pipeline, mapping specific tariff criteria to discrete evaluation blocks.
• Multimodal grounding, RAG, & precise citations: Utilizes a specialized Retrieval-Augmented Generation (RAG) framework to anchor agents in verified source materials, correlating text-based criteria with visual evidence such as wet signatures, corporate letterheads, and geospatial map data. This framework strictly requires agents to anchor every determination with direct document citations, elevating model accuracy. For grid operators, these embedded citations provide immediate, manual auditability, ensuring human-in-the-loop oversight remains seamless and efficient.

Early user feedback from engineers confirms that providing these direct, page-level citations significantly streamlines manual engineering reviews, removing the traditional administrative burden of hunting through dense documentation to verify application data.

Energy infrastructure & security perimeters

Data security forms the structural foundation of HyperQ’s intake pipeline. The system enforces an isolated, U.S.-based secure data perimeter governed by VPC Service Controls, ensuring all data ingestion, processing, and localized document indexing occur exclusively within a hardened perimeter. This strict isolation restricts access entirely to authorized personnel and ensures full compliance with Personally Identifiable Information (PII) and Critical Energy/Electric Infrastructure Information (CEII) data custody standards.

System optimization & operational alignment

Refining HyperQ for real-world operations

Perfecting HyperQ’s architecture—which deploys Google’s multimodal Gemini models within a guided agentic workflow—for a high-stakes, active regional transmission operator (RTO) environment demanded a rigorous period of foundational field testing. Tapestry engineers spent 10 months collaborating with PJM’s team to map out a precise compliance baseline, translating intricate tariff rules into a deterministic instruction set for the system’s control logic.

Their iterative development process required going beyond the “letter of the law” in official tariffs to capture the “case law” in practice, requiring nuanced historical knowledge, real-world edge cases, and expert judgment calls that standard rule-based software cannot handle. Crucially, this initial phase established a repeatable implementation playbook designed to streamline and accelerate deployment across other global energy markets. Data verification challenges and application volume surges are universal challenges, and HyperQ’s core architectural foundation is uniquely built to scale across these varied regulatory environments.

Context-window architecture & mitigating data drift

Dense interconnection materials strain the standard attention limits of AI models. The team resolved this by developing a specialized context-chunking strategy that mirrors a senior project manager's workflow, programmatically distilling raw, heterogeneous filings into focused sub-tasks to prevent context loss.

To directly mitigate real-world data driftObserved shifts in patterns / statistical distributions for a dataset. , HyperQ underwent an intensive, iterative “hill climbing” optimization process that systematically refined its performance. Utilizing a dataset of 234 historical queue submissions, this exercise exposed HyperQ’s underlying engine to a fragmented mix of legal agreements, unstandardized property deeds, and varied PDF formats. By evaluating the system against these historical cases, engineers successfully executed instruction optimization—refining the agents' specific guidance and context windowsThe maximum amount of data an AI model can read and process during a single invocation. to maximize reasoning accuracy across complex real-world variables without altering underlying model weights.

The team executed an intensive regimen of systematic prompt optimization and task alignment and ran thousands of automated evaluations against specific Term, Conveyance, and Exclusivity tariff requirements in order to finalize the architecture. This deep optimization successfully codified decades of domain expertise, aligning parallel macro-assessments with subject matter experts. As a result, HyperQ now delivers predictable, production-ready assessments across 99% of applications in under an hour.

Production execution metrics & validation

The precision of expertise at the velocity of computation

HyperQ successfully streamlined the initial Application Readiness Review for site control in PJM’s Cycle 1 by programmatically isolating and synthesizing 4,581 raw site control documents into 2,328 legally coherent agreement bundles at unprecedented speed.

Four compliance pillars

Processing these files required extracting and cross-referencing a dense matrix of legal and spatial logic. Each of the 2,328 generated bundles was evaluated against four non-negotiable Open Access Transmission Tariff (OATT-defined) site control elements, yielding a total of 9,312 distinct macro-assessments executed across the cohort after accounting for immediate structural exceptions:

1. Term assessment: Calculating active contract duration against cycle-specific regulatory phase minimums.
2. Conveyance evaluation: Verifying chain-of-title continuity.
3. Exclusivity validation: Confirming landowner covenants grant exclusive use and validating the absence of material interference with a developer’s ability to construct and operate the facility.
4. Minimum acreage verification: Calculating spatial viability against strict physical constraints.

Managing linear engineering constraints

Historically, evaluating these interlocking parameters has been strictly linear: analysts manually trace a decision tree one page, one clause, and one document at a time. Attempting to resolve the resulting 9,312 distinct macro-assessments creates a significant challenge for traditional manual workflows trying to scale.

Managing thousands of interlocking, dependent variables simultaneously is cognitively impossible for human reviewers; the sheer density of the cross-references guarantees critical path delays and information degradation. HyperQ bypasses this sequential barrier through two-dimensional parallelization, processing multiple applications horizontally while simultaneously evaluating dense, multi-layered decision trees vertically across pages.

Compounding capability vs. cycle depreciation

Beyond compressing immediate time-to-insight, an agentic system architecture alters the scaling economics of the queue intake process. In standard administrative workflows, managing massive surges in volume across multi-year cluster study cycles introduces severe institutional memory challenges. When queue cycles are staggered, specialized engineering knowledge and administrative edge-case precedents can easily become siloed or lost between active application windows, requiring repetitive team onboarding and alignment sessions to reset baseline operational standards.

HyperQ resolves this structural drift by permanently codifying domain expertise directly into the platform’s core task parameters. Because the software's underlying logic is optimized continuously, every unique edge case resolved or system refinement implemented yields permanent, compounding gains in operational capability. Rather than expending finite organizational energy just to maintain baseline functional parity from one queue cycle to the next, grid operators can permanently institutionalize complex regulatory logic, ensuring identical, uncompromised verification standards regardless of surge volumes or chronological gaps between cluster windows.

Anchoring reviews in a fixed instruction set establishes a standardized, neutral data-intake filter. HyperQ applies the exact same programmatic compliance rules across an entire batch of applications. Ultimately, the system eliminates the grueling manual search for signature blocks and execution dates to surface the exact data points requiring verification, freeing up expert engineers to make the final determinations.

System roadmap: Downstream architecture expansion

A blueprint for grid interconnection and beyond

While verifying site control data is structurally separate from downstream physical engineering studies, stabilizing the entry point ensures the integrity of the entire pipeline. In the coming months, HyperQ will expand to complex electrical data reviews, a phase that currently requires a time-consuming, labor-intensive, and costly manual review by engineers and consultants. By replicating the same proven automation framework used for site control, HyperQ will be able to programmatically audit single-line diagramsA simplified schematic blueprint mapping out a project's electrical layout. and equipment parameters, isolating configuration anomalies before they corrupt additional grid models.

As its processing infrastructure builds outward, HyperQ’s underlying tech stack will evolve to handle the sheer scale of global network complexity. Future iterations of the tool will migrate to Google’s code-first Agent Development Kit (ADK) framework, driven by a graph-based runtime optimized for the extreme concurrent throughput required by subsequent study cycles.

Establishing a clean ingestion protocol lays the critical groundwork for Tapestry’s Unified Model Manager: a novel system that weaves siloed, disparate grid models and data into the first unified, interactive model of the electric grid. This definitive, shared source of truth will allow planners, operators, and developers to execute decisions against a perfectly synchronized grid model, shifting the paradigm from reactive, snapshot-based workflows to active grid orchestration.

Though no single piece of software can act as a total silver bullet for the multi-phased interconnection queue, embedding Tapestry’s tooling within Cycle 1 demonstrates how the initial, data-dense phases of grid administration can be managed with algorithmic precision. To further optimize this entry point, Tapestry also aims to extend HyperQ access directly to the developer community. Providing a transparent interface to pre-validate applications will allow developers to identify and resolve compliance and readiness criteria prior to formal submission.

The operational complexity and regulatory density of North America’s largest electricity market provides the real-world telemetry needed to systematically refine our models. As we iterate alongside grid operators and engineers, HyperQ’s collaborative architecture serves as an evolving blueprint: one that will continuously expand its capabilities to optimize transmission capacity, adapt to changing market rules, and accelerate the transition to a modern grid infrastructure worldwide.

April 29, 2026

December 8, 2025