Kortex
snapshot · 2026-07-18
Kortex · published research · one graph, eight analyses

What connected infrastructure data can reveal

These are published research, not product pages: original observations computed live from the same graph that powers the applications; assets, grid topology, markets, ownership, water and digital infrastructure. They exist to prove the substrate generalises: a new validated relationship improves several of these at once. Each states the question it answers, its evidence class, geography, vintage and maturity; they are not equally authoritative, and they say so. Cite them, link them, challenge them: every one carries a method & receipt footer and a reproducible API route.

Market timing & value

what the market pays, hour by hour and year by year; observed price and generation series, deterministically joined

The grid's day

What does one day on a grid actually pay each technology; and who earns above or below the average?

PRODUCTIONobserved → transformed
38 grids · calendar 2024 · ENTSO-E hourly · GET /v2/spatial/grid-day

A clean year can have borrowed hours

A grid can be 95% clean by annual energy and still lean on imports or fossil backup for hundreds of hours. Which hours?

PRODUCTIONtransformed
hourly 2024, 8,784 calendar slots · nulls never interpolated · GET /v2/spatial/clean-hours

The renewable that eats its own price

As solar's share grows, what happens to the price solar earns? A decade of capture rates, per grid.

PRODUCTIONtransformed
43 grids · 2015–2025 · unit-free capture rate · GET /v2/spatial/capture-decline

Infrastructure constraints

what limits a project besides price; queues, cooling water, digital ecosystems

The queue is not a pipeline

Of the capacity that enters an interconnection queue, how much is ever built; by cohort, technology and region?

PRODUCTIONtransformed
US only (LBNL, end-2024 vintage) · active projects right-censored · GET /v2/spatial/queue-cohorts

Diversified assets. One river.

How much of an operator's thermal fleet depends on a single water system that fails in the same drought?

BETAinferred
US cooling links validated vs EIA-860 (35% quarantined); non-US inferred-unvalidated · GET /v2/spatial/water-portfolio

Power and digital infrastructure rarely agree

The cheapest clean-power zones and the existing digital ecosystems are different places. How far apart?

BETAtransformed (proxy)
294 zones · facility-count ecosystem proxy, not fibre routes · cleared sources only · GET /v2/spatial/power-fibre

The grid's load-bearing nodes

Which substations are structural single points; remove one and part of the network is cut off, generation attached?

RESEARCH PREVIEWderived
28,823 single points of 78,786 networked substations · topology, not power flow; screening for N-1 review · GET /v2/spatial/grid-criticality

Half the power takes the long way round

When Germany sells France 100 MW, which borders physically carry it? Loop flows through third countries, quantified.

RESEARCH PREVIEWmodelled
DC power flow, 2,741 buses · validated out-of-sample vs observed 2025 flows (median r +0.88) · GET /v2/insights/corridor-sensitivity

Demand context

what pulls on the system; counted signals, never forecasts

Where the demand is coming from

People, buildings, vehicles electrifying, datacentres; the countable ingredients of load growth, per country, under the financing backdrop.

BETAsource-reported → transformed
60 countries · EV (IEA/OWID) · GHSL population · 1.41B buildings · FRED rates · GET /v2/spatial/demand-signals

Portfolio concentration

whether diversification on the map survives the physics underneath it

Your wind farms are farther apart than their weather

How many effectively independent weather positions does a renewable portfolio really hold?

BETAmeasured (transformed)
measured ERA5 daily-anomaly correlations, 2015–2025; portfolios rebuilt on measured R (basis measured_corr_v1) · GET /v2/spatial/weather-portfolio
How to read the badges. PRODUCTION = deterministic transformations of observed or source-reported series, validated where a reference exists. BETA = the relationships involved are inferred or proxied, disclosed on-page, partially validated. RESEARCH PREVIEW = a disclosed heuristic stands in for a measurement; the upgrade is declared and the numbers will be restated. Evidence classes (observed · source-reported · transformed · inferred · heuristic · modelled) ride on every material metric in the API. Method & receipt footers on every page; misses published at /validation.