Latest / Defense B2B podcast news daily - for industry professionals subscribe to develop your skillsets / The New Defense Stack: Five Frontiers Rewiring the Industrial Base

Transcript

The New Defense Stack: Five Frontiers Rewiring the Industrial Base · Defense B2B podcast news daily - for industry professionals subscribe to develop your skillsets

• 0:00
  Welcome to Copilot Podcasts. A drone just landed
• 0:03
  on a tank trail in South Korea, not a runway,
• 0:05
  not a tarmac, a literal rough dirt path used
• 0:08
  by armored vehicles, and then it took off again
• 0:10
  and kept flying the mission. No base required.
• 0:13
  Today, we're pulling apart five connected developments
• 0:15
  that are quietly rewiring the defense industrial
• 0:18
  base. A runway independent ISR drone, a flying
• 0:21
  hypersonic test network, orbital security gaps
• 0:23
  that Space Command just admitted out loud. AI
• 0:26
  that needs to train like a soldier, and cyber
• 0:28
  decoys that make your network's floor creak when
• 0:30
  an intruder steps on it. By the end of this episode,
• 0:33
  you'll understand what these five things have
• 0:34
  in common, why they matter right now, and what
• 0:36
  the defense stack actually looks like when you
• 0:38
  zoom out. I've got someone here who has been
• 0:41
  deep in this material, so let's get into it.
• 0:43
  Walk me through what actually happened in South
• 0:45
  Korea. The Mojave Stoll, short takeoff and landing,
• 1:01
  flew from a conventional South Korean Army runway,
• 1:04
  transited to a live fire training range, landed
• 1:07
  on a rough maneuver trail that tanks and artillery
• 1:09
  vehicles normally use, and then relaunched and
• 1:12
  continued the mission. A tank trail? Like the
• 1:15
  ground was not prepared for an aircraft at all?
• 1:17
  Exactly. And that's the whole point. The aircraft
• 1:20
  had dual weapon rails with inert payloads, so
• 1:23
  this was a controlled demonstration, not a combat
• 1:25
  event, but the terrain conditions were representative
• 1:28
  of what a real battlefield forward site would
• 1:31
  look like. The durable headline here is runway
• 1:34
  independence, because reconnaissance becomes
• 1:36
  far more resilient when aviation can disperse
• 1:39
  beyond large, predictable airfields. So it's
• 1:42
  not just a cool trick, it's a strategic shift
• 1:45
  in where you can put eyes. Right, and Mojave's
• 1:47
  value isn't simply that it can land short, it's
• 1:50
  that it can reconnect sensing with dispersed
• 1:52
  ground operations. A forward site could potentially
• 1:55
  support intelligence collection, communications
• 1:58
  relay, resupply, a whole range of approved mission
• 2:00
  packages. The aircraft uses what's called an
• 2:03
  open mission architecture, which means sensors,
• 2:06
  data links, autonomy software, and payloads can
• 2:08
  evolve on different schedules without requiring
• 2:10
  a whole new airframe. Okay, so what does open
• 2:13
  mission architecture actually mean in plain terms?
• 2:16
  Because that phrase gets thrown around a lot.
• 2:18
  Think of it like a smartphone versus a flip phone.
• 2:21
  A flip phone does one thing and you can't add
• 2:23
  apps. An open architecture platform lets you
• 2:26
  swap in new sensors or software without redesigning
• 2:29
  the whole aircraft. That makes integration engineering
• 2:32
  as important as the airframe itself, especially
• 2:34
  when communications or navigation are stressed
• 2:37
  in a contested environment. That's a really useful
• 2:39
  way to think about it. And there's a partnership
• 2:41
  angle here too, right? This isn't just a US program.
• 2:44
  That's actually one of the most consequential
• 2:46
  parts of the story. General Atomics and Hanwha
• 2:49
  Aerospace plan to co -develop and co -produce
• 2:51
  a production version for South Korean and global
• 2:53
  customers. They're expecting a critical design
• 2:56
  review in November and a first production aircraft
• 2:59
  flight in 2027. Those milestones create demand
• 3:02
  signals across the entire supply chain, structures,
• 3:05
  engines, avionics, sensors, secure communications,
• 3:08
  tooling, life -cycle support. So the demo itself
• 3:11
  is almost like a proof of concept for a much
• 3:14
  larger industrial ecosystem. Exactly. And the
• 3:17
  framing that really stuck with me from the source
• 3:19
  material is this. Expeditionary autonomy is an
• 3:22
  ecosystem sale, not merely a platform sale. Programs
• 3:26
  still need maintainable ground stations, training
• 3:28
  environments, field diagnostics, spares planning,
• 3:31
  and trustworthy software updates. The aircraft
• 3:34
  is almost the easy part. Okay, so we've got a
• 3:38
  drone that doesn't need a runway. Now let's talk
• 3:41
  about something that moves a lot faster than
• 3:43
  a drone, hypersonic vehicles. And there's a really
• 3:46
  interesting infrastructure story here that I
• 3:48
  don't think most people have heard. Yeah, and
• 3:50
  this one genuinely surprised me when I dug into
• 3:52
  it. Evivex Corp was selected for a contract worth
• 3:56
  up to $92 .2 million over three years to support
• 3:59
  the Pentagon's Sky Range program. Skyrange is
• 4:03
  managed by the Test Resource Management Center
• 4:06
  and its whole premise is that the strategic bottleneck
• 4:09
  in hypersonic development isn't the missile,
• 4:11
  it's the test network surrounding the vehicle.
• 4:13
  Wait, the bottleneck is the test infrastructure,
• 4:16
  not the actual hypersonic technology? That's
• 4:19
  the insight. Advanced propulsion cannot mature
• 4:21
  through equations and wind tunnels alone. Programs
• 4:24
  need repeatable flight opportunities that return
• 4:26
  usable evidence from extreme speed, heat, vibration
• 4:29
  and communication conditions. And the old way
• 4:32
  of doing this, fixed land sites and specialized
• 4:34
  ships, was slow, expensive, and inflexible. You
• 4:37
  couldn't reposition quickly, and tests were spaced
• 4:40
  out by months. So what does Sky Range actually
• 4:43
  do differently? It moves major instrumentation
• 4:45
  onto high -altitude airborne platforms. Former
• 4:48
  RQ -4 Global Hawk aircraft are being repurpished
• 4:51
  as what they call Range Hawks, carrying sensors
• 4:54
  and communications equipment for flight test
• 4:56
  observation. These airborne nodes can reposition
• 5:00
  across long corridors and relay telemetry while
• 5:02
  a test article crosses distance rapidly. According
• 5:05
  to reporting from Warwings Daily, SkyRange supported
• 5:09
  two hypersonic flight tests in just 10 days,
• 5:11
  one over the Atlantic, one over the Pacific.
• 5:14
  Ten days. That's a completely different tempo
• 5:17
  than months between tests. And that cadence matters
• 5:20
  enormously. Higher test frequency means you can
• 5:23
  reveal weak designs earlier, reward manufacturable
• 5:26
  solutions, and make counter -hypersonic sensing
• 5:28
  more realistic. The source material puts it really
• 5:32
  well. The race worth winning is the ability to
• 5:34
  learn safely, quickly, and honestly at the edge
• 5:36
  of physics. I love that framing. It's not about
• 5:40
  having the fastest missile, it's about having
• 5:42
  the fastest learning cycle. Exactly. And AVEX's
• 5:46
  role under the contract covers airborne range
• 5:48
  integration, flight operations, telemetry relay,
• 5:52
  resilient communications, payload integration,
• 5:55
  and mission data. Mike Jackson, their senior
• 5:58
  VP of Global Solutions, said it directly, rapid
• 6:01
  hypersonic development depends on scalable airborne
• 6:04
  infrastructure, resilient telemetry architecture,
• 6:07
  and the ability to collect operational test data
• 6:10
  at speed. So SkyRange is essentially a mobile
• 6:13
  laboratory distributed across aircraft, networks,
• 6:16
  ranges, and analysis teams. That's a wild reframe
• 6:20
  of what a test range even is. And every layer
• 6:23
  of that system has to preserve data provenance,
• 6:26
  because an impressive flight with unreliable
• 6:28
  evidence teaches very little. That's the discipline
• 6:31
  that makes this more than just a cool drone story,
• 6:34
  it's a data infrastructure story. Alright, let's
• 6:37
  go up even further, literally. Because while
• 6:39
  all of this is happening in the atmosphere, there's
• 6:42
  a whole other domain that's becoming increasingly
• 6:44
  contested, and Space Command just said some pretty
• 6:47
  candid things about where the gaps are. Yeah,
• 6:49
  General Steven Whiting of US Space Command was
• 6:52
  pretty direct. He said tracking has improved,
• 6:55
  but targeting quality understanding of orbital
• 6:57
  objects still has a long way to go. He specifically
• 7:00
  identified gaps in commercial and allied data
• 7:02
  integration, sensor -to -command links, and surveillance
• 7:05
  capacity. The problem isn't just seeing an object
• 7:08
  in orbit, it's understanding behavior well enough
• 7:10
  to make accountable decisions about it. So we
• 7:13
  can see things, but we don't necessarily understand
• 7:16
  what they're doing or why. Right. And Whiting
• 7:19
  also called for expanded awareness beyond geosynchronous
• 7:23
  orbit and into cislunar pathways, the Earth -Moon
• 7:26
  neighborhood. That matters because future commerce,
• 7:29
  science and security will increasingly share
• 7:31
  that space. A peaceful space economy needs traffic
• 7:34
  awareness and resilient communications just as
• 7:37
  much as military readiness does. And there's
• 7:39
  a workforce dimension to this, that's pretty
• 7:41
  striking too. Very striking. Space force leaders
• 7:44
  described goals of 25 ,000 uniformed Guardians
• 7:48
  and 12 ,500 civilian personnel by fiscal 2031.
• 7:53
  They linked that growth to new missions, more
• 7:55
  satellites, more combat crews and greater training
• 7:57
  demand. Officials also emphasized live, virtual
• 8:01
  and constructive environments for testing skills
• 8:03
  before operators face real consequences. That's
• 8:06
  a massive expansion. And it connects directly
• 8:09
  to the next piece of this stack, because you
• 8:11
  can't just hire 25 ,000 people and hand them
• 8:14
  a manual. You need to train them. And increasingly,
• 8:16
  that training involves AI. And this is where
• 8:19
  it gets philosophically interesting. Shield AI
• 8:22
  published an essay arguing that defense autonomy
• 8:24
  is fundamentally an experience problem before
• 8:27
  it's an algorithm problem. Their central claim
• 8:30
  is that public data rarely contains the exact
• 8:32
  threats, sensors, geography, and degraded communications
• 8:36
  links that real missions require. So you can't
• 8:38
  just train a defense AI on the internet the way
• 8:40
  you'd train a chatbot. Exactly. Generative systems
• 8:44
  can imitate abundant patterns, but mission autonomy
• 8:47
  has to behave predictably where examples are
• 8:49
  scarce, in degraded conditions, in contested
• 8:52
  environments, in scenarios that have never happened
• 8:55
  before. The argument is that high fidelity simulation
• 8:58
  can supply controlled repetitions, rare conditions,
• 9:01
  and inspectable outcomes that real -world data
• 9:04
  simply can't provide at scale. But here's where
• 9:06
  I want to push back a little. Can a simulation
• 9:09
  really substitute for the chaos of an actual
• 9:11
  battlefield? Like, isn't there always going to
• 9:14
  be a gap between the virtual and the real? That's
• 9:16
  the right question to ask, and the honest answer
• 9:18
  is yes, there's always a gap. The source material
• 9:21
  is clear that virtual success doesn't equal field
• 9:24
  readiness. But the argument isn't that simulation
• 9:27
  replaces reality, it's that simulation can shorten
• 9:29
  feedback loops, expose failure modes early, and
• 9:32
  make joint training safer and more frequent when
• 9:35
  live events are scarce or too dangerous to run
• 9:37
  repeatedly. So it's a compliment, not a replacement.
• 9:41
  And the slingshot aerospace work with the Space
• 9:43
  Force is a really concrete example of this. They
• 9:46
  were awarded a $27 million contract to inject
• 9:49
  AI into space warfare training through something
• 9:52
  called TALOS, an autonomous AI agent that serves
• 9:56
  as a realistic red force in orbital war games.
• 9:59
  Instead of following pre -programmed paths, TALOS
• 10:02
  responds dynamically to trainee actions, forcing
• 10:05
  Space Force Guardians to deal with unpredictable
• 10:07
  machine speed threats. So you're training humans
• 10:11
  against an AI adversary that actually adapts
• 10:13
  to what you do. That's a fundamentally different
• 10:16
  kind of exercise. And Talos is trained on Slingshot's
• 10:19
  library of real -world orbital data, drawn from
• 10:22
  tracking approximately 95 % of payload -sized
• 10:25
  objects across Earth's orbital regimes. So it's
• 10:28
  not a made -up adversary, it's modeling behavior
• 10:31
  that reflects how real satellites actually maneuver.
• 10:34
  That's genuinely unsettling in the best possible
• 10:36
  way. Okay, but there's a human authority question
• 10:39
  here that I think is really important. Who's
• 10:41
  actually in charge when you have AI systems operating
• 10:44
  at machine speed? This is where a separate commentary
• 10:47
  from September 17 comes in. It proposed explicit
• 10:51
  chains of command and delegation limits for defense
• 10:53
  AI agents. The idea is that autonomy should earn
• 10:57
  authority through evidence rather than novelty.
• 11:00
  Machines can search, correlate, navigate, recommend
• 11:03
  and execute bounded tasks faster than a person.
• 11:06
  But people still define objectives, interpret
• 11:09
  context, approve consequential actions, and remain
• 11:12
  accountable for mission outcomes. So the human
• 11:15
  isn't just a supervisor, they're the one who
• 11:17
  carries the moral weight of the decision. Exactly.
• 11:20
  And the procurement implication is significant.
• 11:23
  Tests should examine coordination, override behavior,
• 11:26
  logging, and degraded communications, not just
• 11:29
  accuracy averages. Trust grows when the interface
• 11:32
  makes uncertainty visible and the organization
• 11:34
  rehearses intervention without drama. That's
• 11:37
  a very different evaluation framework than just
• 11:39
  asking, does the AI get the right answer most
• 11:41
  of the time? Okay, so we've got drones that don't
• 11:44
  need runways, flying test ranges for hypersonic
• 11:47
  missiles, orbital awareness gaps, and AI that
• 11:50
  trains like a soldier. The fifth piece of this
• 11:52
  stack is cybersecurity, and CISA just dropped
• 11:55
  some guidance that I think reframes how a lot
• 11:57
  of people think about defense. Yeah, CISA published
• 11:59
  new guidance on September 16, 2026, specifically
• 12:03
  on cyber decoys for critical infrastructure.
• 12:07
  And the framing is really important. Decoys are
• 12:09
  positioned as a complement to zero trust, not
• 12:12
  a replacement for it. The problem they're solving
• 12:14
  is adversaries who use valid credentials, native
• 12:17
  tools, and what's called living off the land
• 12:19
  techniques after they've already gotten inside
• 12:21
  a network. So the attacker is already in and
• 12:24
  they're using tools that look legitimate. That's
• 12:27
  the nightmare scenario. Right. And that's exactly
• 12:29
  where decoys become powerful. A decoy might look
• 12:32
  like a system, an account, a credential, or a
• 12:35
  file, but it has no legitimate business purpose.
• 12:38
  So if anyone interacts with it, that interaction
• 12:40
  produces a high -confidence signal. Legitimate
• 12:43
  users don't touch well -designed decoys. Only
• 12:46
  someone who's exploring where they shouldn't
• 12:47
  be would stumble into one. It's like making the
• 12:50
  floor creak in a specific spot that only an intruder
• 12:53
  would walk across. That's a perfect metaphor.
• 12:56
  Ancissa breaks this down into a few specific
• 12:59
  tools. Tripwires alert on interaction. Breadcrumbs
• 13:03
  guide suspicious exploration toward monitored
• 13:05
  areas. Honey tokens are fake pieces of data that
• 13:09
  are being watched. And honeypots are more complete
• 13:12
  simulated systems, but those require stronger
• 13:15
  isolation, logging, and expertise to run safely.
• 13:18
  And they're tying all of this to MITRA frameworks,
• 13:21
  right? So it's not just ad hoc, there's a structured
• 13:23
  methodology. Exactly. CESA organizes the planning
• 13:27
  around MITRA Engage and MITRA AT &T CIC, so placement
• 13:31
  of decoys reflects real threat behavior rather
• 13:33
  than guesswork. The aim is to expose intruders,
• 13:36
  impose cost on them, or safely gather intelligence
• 13:39
  under controlled conditions. And because alerts
• 13:42
  from decoys are high fidelity, meaning they're
• 13:44
  almost never false positives, they can actually
• 13:47
  reduce alert fatigue for security teams. Alert
• 13:50
  fatigue is such a real problem. Security teams
• 13:53
  get flooded with noise and start ignoring things.
• 13:56
  And that's the design challenge with decoys,
• 13:58
  believability without creating new pathways into
• 14:00
  production systems. You want the decoy to look
• 14:03
  real enough to attract an adversary, but isolated
• 14:06
  enough that interacting with it doesn't give
• 14:07
  them anything useful. It's a careful balance.
• 14:10
  So how does this connect back to Zero Trust specifically?
• 14:14
  Because Zero Trust is already supposed to be
• 14:16
  this rigorous, verify -everything model. Zero
• 14:19
  Trust verifies every access decision, it assumes
• 14:22
  breach and checks continuously. But decoys add
• 14:25
  something Zero Trust doesn't inherently have,
• 14:28
  a sensor layer that catches post -compromise
• 14:30
  behavior. Zero Trust might let someone in with
• 14:33
  valid credentials and then keep checking them.
• 14:36
  A decoy catches them the moment they start exploring
• 14:38
  somewhere they shouldn't be. The two work together,
• 14:41
  Zero Trust narrows the attack surface, decoys
• 14:44
  illuminate what's moving inside it. That pairing
• 14:47
  is really elegant. One verifies, the other reveals.
• 14:52
  And CESA is clear that decoys don't replace identity
• 14:54
  management, patching, segmentation, or recovery
• 14:57
  controls. They're in addition to a mature defense
• 15:00
  posture, not a shortcut around building one.
• 15:03
  For defense suppliers and infrastructure operators,
• 15:06
  that distinction keeps the idea practical rather
• 15:09
  than theatrical. I want to zoom out for a second,
• 15:11
  because I think there's a thread connecting all
• 15:14
  five of these things that we haven't named explicitly
• 15:16
  yet. What do Mojave, Skyrange, Orbital Security,
• 15:20
  Synthetic AI Training and Cyber Decoys actually
• 15:22
  have in common? Evidence. All five of them are
• 15:26
  fundamentally about making claims testable and
• 15:29
  outcomes inspectable. Mojave proves a physical
• 15:32
  aircraft can do a difficult recovery under real
• 15:34
  terrain conditions, not a simulation, not a marketing
• 15:38
  slide. SkyRange exists so hypersonic programs
• 15:42
  can return usable data from extreme conditions
• 15:45
  rather than just claiming performance. Orbital
• 15:48
  security is about understanding behavior well
• 15:50
  enough for accountable decisions, not just cataloging
• 15:54
  objects. And synthetic training for AI is about
• 15:57
  inspectable outcomes. You can see what the machine
• 15:59
  inferred and how the team responded. And cyber
• 16:02
  decoys produce high fidelity alerts, evidence
• 16:04
  of intent, not just anomalies. The common thread
• 16:08
  across all five is a shift from assertion to
• 16:10
  verification. The defense stack is becoming a
• 16:13
  system that learns faster than its adversaries,
• 16:16
  and the infrastructure being built right now
• 16:18
  is what makes that learning possible. So when
• 16:21
  you step back from all of this, five frontiers,
• 16:24
  five different domains, here's what this really
• 16:26
  comes down to. The United States is investing
• 16:28
  in the infrastructure that makes claims testable.
• 16:31
  That's the hopeful industrial policy underneath
• 16:33
  all of these headlines. And knowledge compounds
• 16:35
  when experiments are repeatable and data remains
• 16:38
  trustworthy. That's true whether you're testing
• 16:40
  a hypersonic glide vehicle over the Pacific,
• 16:43
  training a Space Force Guardian against an AI
• 16:45
  adversary, or watching a decoy honey token get
• 16:47
  touched by someone who shouldn't be in your network.
• 16:50
  The discipline is the same. See clearly, communicate
• 16:53
  honestly, act under accountable authority. I
• 16:56
  want to flag something that I think is easy to
• 16:58
  miss in all of this. None of these programs are
• 17:01
  guaranteed to succeed. The EVIX contract is a
• 17:04
  prototype award whose follow -on depends on validated
• 17:06
  accepted performance. Mojave's demonstration
• 17:08
  didn't prove every operational claim or survivability
• 17:11
  requirements. CISA's guidance explicitly says
• 17:14
  no source supports guaranteed detection or a
• 17:17
  universal deployment pattern. That honesty is
• 17:19
  actually part of what makes these programs credible.
• 17:22
  The source material is careful to separate demonstrated
• 17:24
  performance from marketing language and future
• 17:26
  aspiration. A drone landing on a tank trail is
• 17:29
  real. A production fleet operating in contested
• 17:32
  environments is still a future milestone. Holding
• 17:34
  that distinction is exactly what makes a defense
• 17:37
  workforce trustworthy to customers and crews.
• 17:40
  And that's a useful frame for anyone working
• 17:42
  in or adjacent to this space. The question isn't
• 17:45
  is this technology amazing, it's what has actually
• 17:47
  been demonstrated and what is still aspiration.
• 17:50
  Exactly. And the workforce implications of all
• 17:53
  five of these are worth sitting with. Across
• 17:55
  Mojave, Skyrange, Orbital Security, Defense AI,
• 17:59
  and Cyberdecoys, the roles that matter most aren't
• 18:02
  just technical. They're people who can translate
• 18:04
  between physics, software, mission authority,
• 18:07
  and coalition trust. That's a rare combination,
• 18:10
  and it's increasingly what the defense industrial
• 18:12
  base is competing to find. So here's the synthesis.
• 18:15
  Five things happened in roughly a 48 -hour window
• 18:18
  around September 16 and 17, 2026. A drone landed
• 18:22
  on a tank trail in South Korea. A $92 million
• 18:25
  contract moved hypersonic testing into the air.
• 18:28
  Space Command admitted orbital awareness gaps
• 18:30
  out loud. An AI company argued that defense autonomy
• 18:33
  is an experience problem. And CISA told critical
• 18:37
  infrastructure operators to make their networks
• 18:39
  creak in the right places. And individually,
• 18:41
  each of those is interesting. But together, they
• 18:44
  describe a defense stack that is becoming more
• 18:46
  distributed, more data -driven, more autonomous,
• 18:50
  more orbital, and more cyber resilient all at
• 18:52
  the same time. That's not a coincidence. It's
• 18:55
  a convergence. A convergence. I like that. And
• 18:59
  the actionable insight for anyone listening,
• 19:01
  whether you're in the industrial base, in government,
• 19:04
  in cybersecurity, or just trying to understand
• 19:07
  where defense technology is actually going, is
• 19:09
  this. The platforms are not the story. The infrastructure,
• 19:13
  the data, the training environments, and the
• 19:16
  people who can work across all of it, that's
• 19:18
  the story. And the organizations that understand
• 19:21
  that will be the ones building the next layer
• 19:23
  of this stack. Not the ones with the flashiest
• 19:25
  hardware, but the ones with the most trustworthy
• 19:28
  evidence. Remember that drone that landed on
• 19:30
  a tank trail? The one that opened this episode?
• 19:32
  That wasn't just a cool demo. It was a proof
• 19:35
  of concept for a whole philosophy. Resilient
• 19:37
  ISR begins where dependable engineering meets
• 19:40
  adaptable people. And that same philosophy runs
• 19:42
  through every single one of the five frontiers
• 19:44
  we talked about today. The floor is being built.
• 19:47
  The question is, who's going to be standing on
• 19:49
  it? Thanks for listening to Copilot Podcasts.