GridSignals maps what a large electricity load owes the grid, who can make it respond, what evidence proves performance, and whether the data pipeline preserved that evidence. Fixed-scope transaction diligence, fixed-scope pipeline verification, and a live ERCOT system that turns grid stress into timestamped records checkable against public market data.
The common thread is evidence. A party's own logs document its position; they are not independent proof of it. No one grades their own homework. GridSignals builds the account a counterparty can check rather than take on faith, and states plainly what has not been established.
In a large-load transaction, three things can end up with three different parties: the obligation to respond to the grid, the records that prove what happened, and the equipment that actually executes the response. Under Texas SB6, the definition of a large load customer expressly includes a successor in interest, so a buyer can inherit the obligation while another party keeps the switch, or the records.
What the asset must or must not do, under which source, with what trigger, penalty, and successor treatment. Interconnection, forecasting, ride-through, registration, and curtailment regimes now key on different thresholds, so one asset can sit inside several regimes and outside others at once.
What proves performance: meter data, curtailment notices, QSE records, telemetry. Who holds each, who may disclose it, and whether the rights to obtain it transfer with the asset. The obligation may follow the asset while the rights to the data needed to verify performance may not.
GridSignals focusWho can actually cause the response: the operator, a landlord, a retail provider's desk, the serving utility, or a developer-owned substation. An obligation without control of the deciding equipment is exposure, not capability.
GridSignals focusThe Grid Obligation & Evidence Schedule: every grid obligation on one asset, with its trigger, economic exposure, successor treatment, evidence rights, and who controls the deciding equipment, built from the actual transaction documents.
A fixed-scope, source-to-storage review of whether your pipeline preserved what the source actually published, graded in a fixed findings vocabulary.
Most days, you do nothing. When monitored conditions cross a defined threshold, GridSignals distinguishes system-wide scarcity, localized congestion, and sudden disturbances so the site can apply its own operating policy.
A fixed-scope diligence sprint on one asset, built from the actual transaction documents: interconnection and service agreements, elections, classification letters, and data-rights provisions. The deliverable is a schedule of every grid obligation on the asset, with its source, trigger, required or prohibited behavior, who controls the deciding equipment, economic exposure, successor treatment, what evidence proves performance, who holds that evidence, and whether the buyer can obtain it.
Capital providers can already demand contracts, meter records, and engineering memoranda through a data room. What we have not found is a standardized transaction artifact that integrates obligation source, successor treatment, economic downside, technical capability, and evidence rights across the relevant counterparties; no single counterparty holds the whole picture. Whether transaction parties will pay separately for that integration is an open question, and it is the question GridSignals is testing.
This is a technical and economic evidence memorandum. It is not legal advice, and it is not an engineering certification. Form: one asset, fixed scope and fee, approximately ten business days, from named source documents. Deliverable: the Grid Obligation & Evidence Schedule plus an open-questions register for counsel and the independent engineer. Endpoint: a report date fixed in the scope.
In July 2026, ERCOT changed a market data feed's schema, and GridSignals' own deployed pipeline silently missed a newly published field: the parser discarded columns it did not recognize, and the drift alarm only watched for known fields going absent, so an addition raised nothing. A pre-release baseline had recorded the exposure in advance. We traced the loss end to end and wrote the full audit record.
That failure class, capability built but capture not deployed, published but not stored, changed but silently coerced, is what a source-to-storage engagement looks for: a bounded, fixed-scope review of whether a specific pipeline preserved what its source actually published, before and across a source's schema change. Findings are graded in a fixed vocabulary, and where deployed-artifact evidence is unavailable the verdict is reviewed, not verified. Nothing certifies itself, including us: the method was run on our own stack first.
Form: one pipeline, one deployed stack, fixed scope and fee agreed in writing before work starts, one engagement at a time. Deliverable: a findings report graded in the fixed vocabulary, from verified-preserved to not-gradeable, with the evidence basis for each finding. Exclusions: no continuous or automated assurance. Endpoint: delivery keyed to the first relevant changed rows after the source change.
GridSignals runs a live ERCOT classification system, now in its second summer of production: settlement-point monitoring on a five-minute cadence, mechanism-labeled classification of scarcity, localized congestion, and sudden disturbances, and tamper-evident, timestamped records checkable against public market data.
It is the reason the diligence and verification work is done by people who operate a production grid-data pipeline themselves, with the documented failure classes to show for it. The live system in detail →
A record you simply have to trust is not proof. The grid-side inputs and the integrity of each record are built to be independently checked. A claim about site response also depends on the provenance of the meter data behind it.
Built on ERCOT's public market data: the physical record of what the grid actually did, not a proprietary feed you have to take on trust.
The account comes from outside your operation. That is something a tenant, a regulator, or a noteholder can check, precisely because you did not write it.
A fixed, timestamped account of what the grid did, what GridSignals recommended, and, when source-traceable site data are available, what response was measured: not a verdict you are asked to take on faith.
The standard we hold ourselves to is the one we would want from anyone documenting our grid behavior: claims you can check, not claims you have to believe.
Illustrative values, shown to demonstrate the integrity method. On a real record, the hash chain shows whether the published record has changed. ERCOT references allow the grid-side inputs to be checked against public data; any site-response claim also requires source-traceable meter data.
ERCOT's RTC+B co-optimization went live in December 2025. Under the new market rules, batteries co-optimize across energy and five ancillary service products every five minutes. This fundamentally changed how prices form, when stress events develop, and which signals lead.
GridSignals' current research and evaluation are built around the post-RTC+B market.
Meanwhile, as its large-load queue expanded, ERCOT replaced sequential project-by-project review with a batched study process. Large flexible loads are increasingly settling at points where localized congestion can spike prices $400+ while the rest of the grid is calm. These events cannot be resolved from system-wide averages alone.
And the rules are catching up to the loads. The PUCT approved ERCOT's Batch Zero large-load interconnection process on June 18, 2026. Separate ride-through requirements for large computational loads (NOGRR282 and NPRR1308) took effect August 1, 2026. At the national level, NERC issued a rare Level 3 alert in May 2026 after repeated events in which 1,000+ MW of computational load dropped off the grid in seconds, and is developing registration criteria and reliability standards for large loads under a July 2026 FERC directive. Flexibility is increasingly entering the registration, study, and performance requirements applied to large loads. GridSignals creates a record a site can show if a counterparty asks what the grid did, what the system recommended, and, when source-traceable site data are available, how the site responded.
Our research process reviews observed events against the current models. Material model changes follow the documented evidence and change-control process. Patent pending.
Similar peak-cost and settlement-point questions arise in PJM. GridSignals has built and evaluated a PJM 5CP shadow-monitoring path. It is not a customer-facing deployment.
Designed and built the entire GridSignals platform: data infrastructure, ML models, real-time detection architecture, and autonomous operational systems.
Energy markets expertise across ERCOT, PJM, and other deregulated markets. Leads commercial strategy.