4SHADOW ANALYTIX · Decision Surface Demo
ROUND 2 DEMO
4Shadow Analytix · AI-Driven Decision Surface

When does this airport's load cross its supply?

A working demonstration: an energy decision surface for a major US hub, anchored to published magnitudes, with the screening logic, the lineage, and the human sign-off visible. Not a wall of charts. One date, the levers that move it, and the reasons behind every exclusion.

SYNTHETIC DATA · REAL METHOD
Situation brief

The surface reads the data and writes the brief.

GENERATED NARRATIVE · CANNED FOR THIS STATIC DEMO · PRODUCTION REGENERATES FROM THE GOLD LAYER WITH ROW-LEVEL CITATIONS

This hub buys nearly all of its electricity from the utility, at a spend above $2 million per month[1], roughly 240 to 300 GWh per year[2] with an estimated 35 to 50 MW peak[2]. On-site generation is a single 160 kW floating solar array[3]. A $9 billion capital program pushes load upward through roughly 2031[4], and the airport holds staged percentage reduction targets with no net-zero pledge[5]. Under the modeled assumptions[8], projected peak demand crosses contracted supply capacity on the date shown below. Three levers move that date; one candidate source is screened out, and one sits beyond the planning window. Every verdict shows its reason.

The crossing date

One number carries the page.

—
Projected month when peak demand first exceeds contracted supply capacity, under the levers selected below.
Five-year monthly projection. Demand curve recomputes live; the dashed line is contracted capacity.
LEVER · SOLAR + STORAGE
0 MW effective peak shave. Requires FAA glint and glare study; on-site precedent exists at 160 kW scale[3]. Phased in over 24 months from decision.
LEVER · UTILITY FEED UPGRADE
Interconnection queue and construction modeled at 36 months from decision[8]. Capacity steps up when it lands, not before.
LEVER · ELECTRIC CENTRAL UTILITY PLANT
The reference build is DFW's $234M zero-carbon electric central utility plant[6]. Modeled as an 8% peak reduction phased over 36 months[8].
SCREENED · SMR
Not a toggle. NRC licensing horizon places first power beyond this 5-year window. Shown so the exclusion is a decision, not an omission.
The screening matrix

How things get ruled out is the analysis.

Candidate sourceVerdictGates and reasons
Solar + battery storageINFAA glint and glare study required near runways, historically passable. Utility interconnection queue applies. On-site precedent: the 160 kW floating array[3]. Peak-shave value depends on storage sizing, modeled 0 to 10 MW.
Utility feed upgradeINNo airspace constraint. Bound by the utility's interconnection and construction timeline, modeled at 36 months[8]. Adds contracted headroom, not resilience.
Electric central utility plantINProven at reference scale: DFW's $234M build, sized for future load growth[6]. Converts thermal distribution and cuts electric peak, modeled at 8%[8].
WindOUTPart 77 obstruction surfaces and Doppler radar interference make on-airport turbines effectively unpermittable. Off-site wind is a power-purchase question, not a siting lever, so it exits this matrix.
Small modular reactorBEYOND WINDOWNRC licensing and first-power timelines fall outside the 5-year planning window. Carried as a long-horizon option on the 10-year sheet, never as a near-term toggle.
The knowledge panel

Entities and relationships behind the numbers.

The graph is why the brief can cite rows instead of vibes: every load, source, and feeder is an entity with lineage, and retrieval carries the relationships with it.

UTILITY FEED contracted capacity SUBSTATIONS + feeder topology TERMINAL LOADS HVAC, lighting, vertical transport PLANNED CONCOURSE capital program load adds [4] GSE + EV CHARGING electrification adds [7] FLOATING SOLAR 160 kW, on-site [3] CANDIDATE SOURCES from the screening matrix PEAK DEMAND the number that crosses
Knowledge graph beside Postgres running hybrid retrieval. A pull carries its meaning and lineage with it.
The sign-off queue

The surface recommends. A named human decides.

Commission the FAA glint and glare study for solar siting
Gate item for the highest-value lever. Evidence: screening matrix row 1; on-site precedent [3]. Confidence: high that the study is passable at candidate sites.
Open the utility interconnection conversation this quarter
The 36-month lead time [8] means the feed-upgrade lever only moves the crossing date if it starts before the date locks in. Evidence: the curve above with the lever off versus on.
Scope the electric central utility plant feasibility against the DFW reference
Reference build documented at $234M [6]. Evidence: screening matrix row 3; modeled 8% peak reduction [8].
AUDIT LOG · demo instrumentation, in-page only · production writes Part 11-style signed entries to the estate
The architecture

One gold layer. Power BI beside the surface, not beneath it.

METERS + SENSORS S3-BACKED LAKEHOUSE (ICEBERG) BRONZE raw SILVER conformed GOLD KG + semantics POWER BI enterprise reporting, where the org lives THIS SURFACE briefs, screening, sign-off, lineage
Beside, not instead. Both consumers read the same governed gold layer.
Anchors and assumptions

Every number, tagged.

[MEASURED] [1] Monthly electricity spend above $2M (airport sustainability page). [3] 160 kW floating solar array, installed January 2020. [4] $9B capital program with major elements completing about 2031. [5] Staged reduction targets, 20% by 2035, no net-zero pledge. [6] DFW $234M zero-carbon electric central utility plant, groundbreaking August 2023. [7] Comparable hub anchor: 576 electric GSE charge points installed at SEA.
[ESTIMATED] [2] 240 to 300 GWh per year derived from spend at large-commercial rates; 35 to 50 MW peak derived at a 0.6 to 0.7 load factor.
[SYNTHETIC] [8] Contracted capacity, growth rates, program load-add phasing, lever timelines and magnitudes. Stated in the config block of this page and defended as modeling choices, not citations.
Full sourced research: the APA 7 research notes on the 4Shadow repo main branch, 2026-08-05 run.