TRAIN AS YOU FIGHT · EXERCISE · NOTIONAL EVENT · UNCLASSIFIED · NOTHING IS TRANSMITTED
SHIELD | ATLAS · Program office evaluator view
New program-office reviewer? Understand the system, choose your mission, and follow the proof in six minutes →.
Joint DoD · evaluator's starting point

Find your mission function, owning office and proof requirements.

Army, Navy, Marine Corps, Air Force, Space Force and joint missions have distinct operational doctrine and acquisition portfolios. Review the branch-by-branch WFF / program-office crosswalk before choosing an exercise. Candidate offices are not sponsors; no government integration or authorization is assumed.

Open service crosswalk, evidence gates and linked sources →

Army ground-vehicle example · four-click rehearsal

This page's interactive example focuses on a notional Army vehicle and fires workflow; it does not define the whole ATLAS platform. Review an illustrative crew report, degrade a link and rehearse the human approval handoff. Nothing on this page is transmitted to a government system.

The ±12 m, ±20 m, and ±23 m figures are exercise presets: soldier ±60 m, sight ±25 m, drone ±15 m, radar ±40 m. They are not a microphone measurement and they are not a radio-dongle measurement.

Loading the synthesis study…

Compare support responses, inspect dependencies across six staff functions, and define a value test. Synthetic staff rehearsal; no operational dispatch or measured ROI.

Fires program-office hypothesis · Ground program-office hypothesis

Loading offices…

PAE LEAVE-BEHINDS · ONE PAGE PER OFFICE · SAME NUMBERS AS THIS PAGE

Every office in the room gets one page and one ask

Each one-pager leads with that office's own problem and programs, then the ask written in this registry. The meters on them are the exercise presets on this page: one soldier ±60 m, all sources near ±12 m, drone jammed near ±20 m, jammed and congested near ±23 m. Those meters are not microphone or dongle measurements. Open one and print it to PDF. Orientation: who owns which beat →

PAE Fires · Integrated Fires Mission CommandThe packet, your threshold, a person releases or holds.One-pager · Office door →
PAE Maneuver Ground · CPE Ground (formerly PEO Soldier + PEO GCS) · PM Mounted Armored VehiclesThe crew confirms one location and hands Fires one packet.One-pager · Office door →
\n
VBK / Maneuver — integration evaluatorAuthenticated read-only source evidence, UAS groups 1–5 screening, protective-advisory review and C2 handoff gates; NO government VBK feed claimed.Open technical evaluator (sign-in) →
PAE C2 / Counter-C2 · Next Generation C2What arrived, what waits, and what the error becomes on a degraded network.One-pager · Office door →
PAE Agile Sustainment and AmmunitionAmmunition, fuel and vehicle state beside the released request.One-pager · Office door →
PAE Layered Protection · Counter-UASOne drone, one track, held for a person; incoming fire to counter-fire with a human release.One-pager · Office door →
THE RAG ASSISTANT · OPEN ONLY IF THEY ASK HOW ANSWERS ARE SOURCED

Shorter distance between a question and a sourced answer

The AI on this platform exists to get your staff from question to sourced answer faster. It is retrieval-augmented generation: each answer is assembled from pages retrieved out of the platform's intel corpus — cached intelligence files, ingested briefings, published sources — and the underlying files are one click away, so any answer can be checked against its source. The corpus is the same one the Intelligence Synthesis Engine reads, and it is what you can audit.

Where the corpus stands today, honestly: it is a cached, curated set of intelligence files (the live auto-research scheduler is currently off; pay-per-use lanes are gated). The platform carries two ingestion paths that feed it when enabled — an automated researcher that pulls material in from the wild, and a loopback that digests the operator's own briefings and alerts back in as institutional memory, clearly labeled and never treated as a primary source.

Guardrails we run on it: a probabilistic RAG drift gate (baseline eval, rolling drift detection, retrieval-faithfulness probes), hash-chained audit of what the AI produced, an AI SBOM, and runtime integrity checks. These are features of the platform — not certifications, and we claim no authority to operate.

Open the RAG console → · How the platform behaves when disconnected (DDIL) →

CHEATSHEET
✚
ASK AI
✚
PRESENCE ✓
⚠ SANDBOX / TRAINING MODE — Live read-only data. Write commands are inhibited (train as you fight, missile button safed).
NAV HISTORY