Intel at the Edge

In this episode of Intel at the Edge, host Chip Usher sits down with Shelby Pierson, former Director for Analysis at the National Geospatial-Intelligence Agency (NGA).
Shelby joined the intelligence community in 1999 as an imagery analyst and rose through the ranks — working Iraq ground forces analysis, denial and deception, election security as the DNI's lead, and ultimately serving as NGA's Director of Analysis.
Shelby recently authored "Sensing the Threat: Operationalizing Computer Vision for Intelligence and Warfighting," and in this conversation she unpacks how AI and computer vision are reshaping the discipline — from the days of physical light tables and film to today's data deluge of commercial, national, and airborne collection.
Topics covered include:
- Her path into the CIA and career highlights, including Operation Iraqi Freedom
- How the analyst's job has evolved from imagery exploitation to "data creation"
- The origins of computer vision in GEOINT, from change detection to Project Maven
- Why analyst staffing hasn't kept pace with the data explosion
- Where human judgment and tradecraft still matter most
- How today's crises move faster than the 13-day Cuban Missile Crisis timeline — and what that means for AI-assisted decision-making
- Pushing AI compute to the edge — including onboard satellites in orbit
- Her advice for the next generation of imagery analysts
🎧 Listen and subscribe for more conversations on how emerging technology is reshaping intelligence, defense, and national security.
00:00 - Introduction: Shelby Pierson & "Sensing the Threat"
01:05 - Shelby's Path into the Intelligence Community
02:29 - Early Career: Iraq, Denial & Deception, and Working with Engineers
05:03 - From NIM for Russia to Director of Analysis at NGA
05:30 - The Light Table Era vs. Today's Desktop Analysis
07:40 - The Data Deluge: Scale of Modern Imagery Collection
08:37 - Integration: How Analyst Workload Has Changed
09:26 - Staffing Hasn't Kept Pace with the Workload
10:35 - Technology as Offset vs. Replacement for Human Analysts
11:35 - What Is Computer Vision? Origins in NGA
14:20 - Change Detection: The Early Analytics Story
15:34 - Project Maven and Full Motion Video
17:19 - Beyond Visual: Hyperspectral, Radar & Data-Rich Sensors
20:21 - From Imagery Exploitation to Data Creation
20:50 - Why Now? The Convergence of Software, Storage & Compute
22:32 - Tradecraft, Authorship & Trusting the Models
24:51 - Oversight Standards: Are AI Outputs Held to the Same Bar?
25:47 - Error Tolerance: Change Detection vs. Targeting
26:38 - AI in the Kill Chain: The Israel Example
27:57 - Golden Dome and Split-Second Defensive Decisions
29:30 - Case Study: The Chinese Embassy & Iranian Girls' School Incidents
32:36 - Cuban Missile Crisis vs. Today's Compressed Decision Timelines
34:22 - Piecing Together Complex Intelligence Networks
35:17 - Offense vs. Defense: Rules of Engagement for AI
36:15 - Scaling for a Potential China Contingency
37:10 - AI at the Edge: Onboard Compute and Sensors in Orbit
40:07 - Advice for Aspiring Imagery Analysts
42:15 - Closing & Where to Find the Paper

Intel at the Edge Podcast is a product of the Special Competitive Studies Project.

In this episode of Intel at the Edge, host Chip Usher sits down with Don Leka, Founder and CEO of Jumptuit, to explore a different kind of artificial intelligence — one built not on text, but on real-time data from the physical world.

Don shares his journey from building the Glide OS cloud platform in 2004, to founding Jumptuit, a company using a large dynamic reasoning model to track physical-world signals — from shipping movements and port activity to refinery output and geopolitical indicators — in near real time.



The conversation covers:

- Why Jumptuit's approach differs fundamentally from large language models

- How cross-sector data intersections can reveal probable events before they happen

- Using comparative, causal, and analogical reasoning to forecast geopolitical and economic shifts

- Case studies: Russia's GDP resilience despite sanctions, shifting oil demand, and instability in Niger and Albania

- What the intelligence community could gain from live, empirical, physical-world data

- A must-listen for anyone interested in AI, geopolitical forecasting, and the future of intelligence analysis beyond large language models.



0:00 Intro: AI Beyond LLMs & Meet Don Leka

1:16 Don Leka's Background & Building Glide OS (2004)

3:17 How Jumptuit's Approach Differs from ChatGPT & LLMs

5:10 What Data Jumptuit Collects (Persian Gulf Example)

9:50 Open Source Data & Cross-Sector Event Thresholds

10:48 Case Study: Operation Epic Fury & Iran War Assumptions

13:42 Large Dynamic Reasoning Model as Predictive Analytics Tool

16:16 Lessons from the Arab Spring & Traditional Intelligence Gaps

17:39 Why Historical Data Alone Misleads Forecasting

18:42 Case Study: Russia's GDP Resilience Despite Sanctions

21:10 Persian Gulf Tensions, Oil Prices & China's Reduced Demand

22:23 Ripple Effects: Japan, Albania & Global Instability

23:55 Is It the Data or the Algorithm? The "Secret Sauce"

24:40 Why LLMs Struggle with Outdated, Narrative-Based Data

26:25 Can Jumptuit's Model Be Distilled or Copied?

28:42 Beyond Geopolitics: Climate, Health & Individualized Intelligence

31:12 What the National Security Community Should Be Doing

32:40 How Jumptuit Complements (Not Competes With) LLMs

33:53 Key Customers: Finance, Energy, Defense & More

Intel at the Edge Podcast is a product of the Special Competitive Studies Project.

Narrative intelligence is emerging as one of the intelligence community's most important new disciplines — and few people understand it better than Joe Stradinger. 

In this episode of Intel at the Edge, Joe Stradinger — founder of Edge Theory and creator of the National Center for Narrative Intelligence in Oxford, Mississippi — joins Chip to break down what narrative intelligence actually is, how it differs from traditional social media monitoring, and why he believes it's become a central front in modern cognitive warfare.

They cover:

• Joe's path from CPA to early internet entrepreneur to national security advisor

• What "narrative intelligence" means and how it's detected, classified, and measured

• The concept of "intercontinental ballistic narratives" and how adversary messaging spreads over time

• How China, Russia, and other actors exploit real-world events to sow division

• Washington's growing (but still unsettled) response to cognitive warfare

• Why storytelling — from Joe's Mississippi Delta roots to America's own national narrative — may be the most powerful (and underused) tool in this fight

Whether you work in national security, business, or just want to better understand how information shapes belief and behavior, this conversation offers a rare look inside a discipline that's quietly reshaping how the U.S. thinks about influence, deterrence, and truth.

00:00 Intro

00:44 Joe's background & how he got into this space

03:12 Mississippi, storytelling, and why fiction is a weapon

05:04 What is narrative intelligence?

09:06 Tracking narratives over time (the CCP example)

13:21 Case study: the data center narrative war

17:29 How AI has collapsed the influence chain

18:40 Case study: Iran's Hollywood-style war narratives

21:14 Attribution: tracing a narrative back to its source

24:22 Federal vs. commercial — why U.S. citizens are never targeted

26:53 Real-world exploits: Baltimore Bridge, Silicon Valley Bank

27:41 The new NSC role & "putting the I back in DIME"

30:20 Why America needs to tell its own story

34:00 Educating the next generation

36:34 The National Center for Narrative Intelligence

Intel at the Edge Podcast is a product of the Special Competitive Studies Project.

Four and a half months into the Iran war, what's actually left of Iran's nuclear program — and what would it take to keep it that way?

In this episode of Intel at the Edge, host Chip Usher sits down with Amy McAuliffe, who recently retired from the CIA as the Assistant Director for Weapons and Counter Proliferation. Amy also served as Chairman of the National Intelligence Council and Assistant Director of CIA for the Near East.

Amy breaks down:

·       Why she describes Iran's program as "down but not out" after the June 2025 and 2026 strikes on Natanz and Fordow

·       The uncertainty around Iran's 60%-enriched uranium stockpile and where it may be located

·       The impact of losing key nuclear scientists — and how much technical knowledge Iran has actually retained

·       How the death of Supreme Leader Khamenei and the rise of new IRGC leadership could shift Iran's calculus toward weaponization

·       What a credible future nuclear deal would need to include, from enrichment caps to IAEA verification

·       Whether a purely military solution can ever fully resolve the nuclear question

·       Her advice for young people considering a career in the U.S. intelligence community

Intel at the Edge Podcast is a product of the Special Competitive Studies Project.

Kimi K3, China's largest open-weight AI model yet, just hit the public — and it's rewriting the terms of the US-China AI race. In this emergency episode, host Chip Usher sits down with David Lin, Senior Director for Tech Platforms at SCSP and CIA veteran, to unpack what K3's release really signals. 

They dig into how the competition has shifted from a simple US-vs-China contest into a two-axis battle of open weights vs. closed weights, why K3 dropped right alongside Xi Jinping's World AI Conference speech, the eye-popping cost gap between K3 and frontier US models, and the question of large-scale model distillation and IP theft. 

They close out with a look at Nvidia's Jensen Huang and a growing coalition of US companies pushing back against a closed-weight future. It's a fast-moving landscape — and this conversation captures exactly where things stand today.

00:00 — Intro: Kimi K3 drops

01:20 — What K3 is and how it compares

03:18 — Open weights vs. closed weights

07:14 — Beijing's role and export controls

12:42 — Security risks of open weights

15:32 — Distillation and IP theft

21:22 — Nvidia and the US open-weight push

24:25 — Closing thoughts

Intel at the Edge Podcast is a product of the Special Competitive Studies Project.