🛰️ Defense & Intelligence

See the connections no single session can

Pattern matching across 500 reports. Auditable to source. Air-gapped deployment.

Intelligence analysts develop intuitions after reading hundreds of reports. Those patterns live in their heads and walk out the door when they leave. Pensiv makes accumulated context a system property.

What you're ingesting

Ambient sources specific to intelligence work

Field reports

HUMINT summaries
Observer reports
Debriefs
After-action reviews

Signals intelligence

SIGINT summaries
Intercept analysis
Communications pattern reports

Imagery & geospatial

Satellite imagery annotations
IMINT analysis
Movement pattern reports

Open source intelligence

News aggregation
Social media monitoring
Public records
Academic publications

Internal collaboration

Analyst notes
Briefing materials
Coordination emails
Assessment updates

Structured intelligence

Diplomatic cables
Intelligence assessments
Warning reports
Threat briefs

What the vault builds

Domain-specific schemas discovered automatically

Entity networks across actors, locations, and events

The vault tracks who does what, where, and when. Entity linking surfaces when Actor A from Report #47 is the same as "regional commander" in Report #412.

Schema induction discovers recurring operational patterns

After ~50 reports, the vault knows: "this type of actor tends to follow this sequence before escalating." Patterns that were invisible become obvious.

Analogical retrieval surfaces structural similarities across theaters and time

"What patterns from the South China Sea resemble pre-escalation indicators from the Black Sea?" No keyword overlap. Identical deep structure.

Contradiction detection flags when new intelligence conflicts with established assessments

The system doesn't delete the old assessment — it marks it superseded, links it to the new one, and surfaces the contradiction if both are retrieved.

What you ask it

Example queries for intelligence analysts

"What indicators from the past 90 days structurally resemble pre-escalation patterns from [historical precedent]?"

What Pensiv returns Primary Scenario (0.68 confidence): Naval posturing + diplomatic rhetoric pattern matches Black Sea Q4 2021. Evidence from 7 reports (Mar-May 2026), cross-theater structural similarity 0.71. Stress Scenario: Supply line buildup resembles [redacted] pre-conflict indicators.

"Has Actor X changed behavior in the last 6 months compared to established baseline?"

What Pensiv returns Yes. Baseline (18mo, 42 reports): monthly cadence, civilian infrastructure focus. Recent (6mo, 14 reports): weekly cadence, military-adjacent targets. Schema shift detected Mar 2026. Contradiction flagged between Feb assessment ("stable pattern") and Apr field observations.

"What's the dissenting view on [assessment topic]?"

What Pensiv returns Consensus (5 analysts, Apr-May): Actor likely responds to economic pressure. Dissent (Analyst C, May 3 note): "Economic pressure historically drives brinkmanship escalation, not de-escalation." Links to 3 historical cases (2019, 2022, 2023) supporting dissent.

Proof & compliance

Built for classified environments

Active TS/SCI + CI Poly clearance. 20 years Special Forces (18F intelligence, 7th SFG). Zero external dependencies. SQLite-first. Deployable on SIPRNet / JWICS.

ITAR / EAR

Defense export control. Full provenance + user attribution. On-prem or air-gapped deployment.

ICD 203

Intelligence Community audit standards. Retrieval trace, temporal accuracy, write trail.

Air-gap ready

Zero phone-home. PostgreSQL + pgvector. Local LLMs via Ollama. No cloud dependencies.

FedRAMP pathway

Defense ATO established. FedRAMP Moderate target Q1 2027. Civilian IC + DoD use cases.

Why defense first

An unfair advantage competitors can't replicate

Clearance moat

Active TS/SCI + CI Poly. Takes competitors 12-18 months to replicate. Every other memory startup has zero cleared personnel.

All six capabilities map directly

Analogical reasoning (cross-theater patterns), surprise gating (filter noise in high-volume feeds), FSRS decay (age out stale assessments), schema induction (discover operational patterns), full audit (ICD 203 compliance), wiki synthesis (team knowledge doesn't walk out the door).

$13.4B DoD AI budget FY2026

First-ever standalone AI budget line. SBIR Phase I ($150-250K) → Phase II ($750K-$1.5M) → Phase III sole-source. Parallel OTA through DIU.

Ready to deploy a vault that remembers?

On-prem. Air-gapped. Federated analyst vaults. Full audit trails. Active clearance. We speak your language.