Adam Warmoth took this interview from Fort Carson. A division commander had called him the Friday before and asked him to be in Colorado on Monday, so he was in Colorado, sitting in a vehicle with a Starlink terminal doing the work of an internet connection. Ross Fubini, who wrote Chariot’s first check, opened by pointing out that this is every call he has ever had with Adam.

One guest measuring things in the field, while the other measures the field. Both agreed to be unusually specific about how they make their decisions.

About the guests

Adam Warmoth is the founder and CEO of Chariot Defense. He started in the Stanford lab of Ilan Kroo, who had just come back to teaching after founding the first electric aircraft company on Larry Page’s money. That led to Kitty Hawk, then to Uber Elevate as one of the early employees, then to Anduril at roughly 200 people, where he built the counter intrusion business unit and then took over counter UAS air defense.

Ross Fubini is the founder and managing partner of XYZ Venture Capital. He was first money into Anduril and first money into Apex Space. A very long time ago he introduced Palantir’s first employee to its second. When he started XYZ almost ten years ago, the premise was simply that he should back people coming out of the Palantir ecosystem. He describes that premise as having been correct, “shockingly so.”

About the company

Chariot Defense builds the power layer for the battlefield. The Amphora systems are modular, high voltage, and low signature, and they charge and run the things that now go forward with a unit: drones, radars, comms, sensors, directed energy. Founded in late 2024 and headquartered in San Francisco, Chariot came out of stealth in July 2025 with an $8M seed led by General Catalyst, and announced a $34M Series A led by Andreessen Horowitz in February 2026. XYZ was in both. Systems have been through JIFX at the Naval Postgraduate School, up to Alaska with the 11th Airborne, out to a National Training Center rotation with the Black Jack Brigade, and onto DIU Project GI.

Adam’s framing of the market is that there are three critical infrastructure layers under any robotic modernization story: comms, compute, and power. He thinks Palantir has taken the network layer and Anduril has taken edge compute, and that power is the third slot, and that after those three there are not really any more slots.

Five Things

1. Useful has a weight…and soldiers do the math

Adam does not use the term “Minimum Viable Product,” preferring instead “Minimum Useful Product.” For Chariot, there is a physical test for that determination: You hand a soldier your system. You hand him a box of identical weight and volume filled with MREs, or ammunition, or water, or Zyn. You tell him to pick one for the mission. Until your box wins that trade, you have something you can test in the field, and you do not have a product.

Lots of founders describe product-market fit, but I’ve yet to hear one describe it as a volumetric tradeoff against ammunition. It’s clean and simple, and requires no advanced analytics.

2. The feedback that counts comes from the guy who did not want to be there

Adam’s rule number one of new equipment training is to remember that the people in your training are not the people who will use your system. They are the people who happened to be available when a commander received the request to send some bodies. They are not your champion. They are not the innovation officer. They are more likely than not junior enlisteds trying to get through the day.

Which means their feedback is invaluable. They are the ones Chariot listens to. Not the enthusiast, not the sponsor, but the guys with no reason to care. Because if you can make that person’s day easier, you may really have something.

Adam is careful to say that this barometer is not sufficient, that you still owe the program office and the general officer engagement and the congressional work. But it is a much better north star for early feedback.

3. Be so good they refuse to give back the product

Chariot took systems to the 11th Airborne up in Alaska for a demo supporting DIU. Everybody used them. They plugged the Anduril Pulsar into them. When it came time for the team to head out, the soldiers said “not so fast.”

Adam ultimately left the gear behind (and did not charge for them) because they were early prototypes. Turns out, those prototypes (serial numbers 1 + 2) are still up there today. And the knock-on effect was that the first production unit went to Alaska as an actual sale (not SBIR revenue or an R&D proof of concept).

Ross calls that an incredibly uncommon signal, and I have to agree with him.

4. What the product does not do determines the speed of sale

Chariot booked early product revenue through DLA Tailored Logistics Support while OTAs and the other vehicles were still being set up. Adam is clear that this is not the right path for most products. It was the right path for his because of a list of things it did not do: the system did not fly, so it required no airworthiness cert. It does not emit, so it required no spectrum approval. It did not have to connect to a network for base functionality.

Hence, leveraging this pathway did not post undue business risk.

Bottom Line

Chariot walked into a capability requirement that had been stuck in requirements land for about four years. Some people who cared about power had seen the problem early, and had started to work the requirement. It was not funded, not approved, and had no field of vendors already bidding.

That, Adam says, is “luck.” If the requirement had already been funded and contested, he would have been late. If there was no semblance of a requirement, he would have been too early. Chariot arrived in the narrow band between those, and then used field credibility with senior leaders to help push the requirement across the line. In so doing, Chariot became a champion of the program office.

Bonus: What does it mean to be a first-check funder?

Ross is something of a legend in first-check defense land.

“I believe that I have one job, and my job is to get the next round of funding done,” he says.

Because he has watched enough companies go from seed to A to B to C, he can connect what the founder is doing now to what the capital markets will need to see later. The line that stuck with me was his example of the advice, which is not the usual encouragement. It sounds like this: because we did that eight months ago, we can tell this story now, which means this round gets done at a billion and not three billion.

In my opinion, most investors claim to be value-added while selling a commodity, but that level of pattern matching stands out. It only works, if the investor is close enough to the operating detail to know what happened eight months ago. Ross says most people do not have the patience to go that deep.

Sourcing Founders

For the aspiring VCs out there, Ross was pretty blunt about what’s enabled his success: he goes to the founders he has backed and says the problem is that he cannot find anyone as good as them. Then he asks who else they know. That’s it.

He doesn’t care about “seeing every deal” and he doesn’t build a fancy scanner.

He focuses on “walking a network” and asking its best members to name the next one.

On What Will be Built Next

XYZ has two internal views of defense. Either you become the standard, or you are wildly different. Ross believes Chariot becomes the standard for battlefield energy and Apex becomes the standard for satellite buses. Everything else has to be genuinely odd, such as his other investments: the “only secure 5G network,” and/or the “only ground autonomy integrator that goes from wheels to autonomy.”

Too many defense tech companies will die in the middle. “Ten percent better,” or “it works in this one environment,” or “I have this one key relationship” just will not work.

Chariot Defense: Website | LinkedIn
XYZ Venture Capital: https://www.xyz.vc
Crossing the Valley: https://www.valleycrossers.com

Adam Warmoth: https://www.linkedin.com/in/adamwarmoth/
Ross Fubini: https://www.linkedin.com/in/fubini/
Noah Sheinbaum: https://www.linkedin.com/in/noahsheinbaum/

Zachary Long and Eric Schwartz are the co-founders of Conductor AI. Both spent seven years at Palantir, where they met on a Department of Justice project supporting large-scale fraud prosecutions. But this wasn’t “start a company at first sight.” After collaborating, they split: Eric moved toward law enforcement and other USG components, then took what he calls “a bit of a rumspringa” into health tech; Zach went deeper into the Department of Defense. When Zach decided to leave Palantir, the two reconnected. Ben, the third co-founder, knew Zach from college and originated the idea with him; Eric joined shortly after as the first employee and built out the company.

The founding insight came from Zach’s own frustration. He wanted to write up a meeting in a Slack channel and was told he could not. He asked why. The internal security team said no, and there was no mechanism to challenge it. “Show me the document,” he said. Nobody could. Multiply that across the defense supply chain and you have put your finger on the systemic over-classification that riddles the national security enterprise. With a bit of good luck (and good timing), the LLM era arrived, and Zach found his solution.

About Conductor AI

Conductor AI builds software for high-stakes government decision workflows: security classification and declassification, foreign disclosure, export licensure, FOIA, and evidence review on the justice side. The platform ingests complex policy documents and institutional tradecraft, then orchestrates the decision tree a human expert would walk. The goal is for the product to generate more than just an answer; it also shares an underlying policy that supports it, so the reviewer can show the math to a starred superior.

Their customers are some of the least glamorous and most under-resourced offices in national security. One person at Nellis Air Force Base handles the entire technical manual transfer review process for the Air Force: 500,000 pages. One GS-13 adjudicates roughly $40 billion in export license approvals a year. These offices ask for more people annually and receive rotational staff who take six months to train and serve twelve months in seat.

The company has solid early traction: a direct-to-Phase-II Air Force Open Topic SBIR, deployment on classified networks with multiple releases per week, and two multinational exercises this summer. Those exercises were big wins, as they were able to take reviews that typically take a day down to an hour.

Key Takeaways

1. Sell at the friction point.

Conductor’s first thesis was to help people determine the correct classification of a document. They quickly learned this was directionally right but commercially wrong. Almost nobody sits down with a security classification guide and asks that question in the abstract. The acute pain was at the transfer point: when someone needs to move this information to a foreign partner, the question is, “can I?” That reorientation was a powerful business unblocker.

2. Cold outbound is a validation instrument.

Zach’s first contract came from spamming federal employees on LinkedIn. Enough of them assumed it was a phishing test that it became a running joke with customers. The fact that he was able to get strangers to reply to a cold message about classification review was pretty compelling evidence that the pain was real.

3. A cracked door is better than a closed one.

The Conductor team quickly found that attempting to convince an incumbent process owner to reform was a tall task. Instead, Conductor began looking for teams that were charged with improving classification management. These offices often came with some leadership buy-in, which made them qualified candidates for early adoption. Attitudinally, they were more likely to house people looking for a fresh approach, rather than defending an existing one.

4. Earn trust by grading yourself against decisions the customer already made.

Conductor was able to create a compelling sales motion by using backtests. If a prospective customer office processed 85 license applications last month and decided 16, Conductor could get their policy documents and tradecraft, and see how closely their software matched the 16. It converts an unfalsifiable AI claim into a measurable one, uses only the customer’s data, and leaves their judgment as the answer key.

5. The durable advantage is domain context, not model access.

Every competitor has the same frontier models. What Conductor has is what Eric calls “old school context, not model context.” Both founders completed the foreign disclosure training courses, they go on-site, and they build prototypes next to the person doing the work. Change management, accreditation, collaboration, security, and test-and-evaluation are not solved by an LLM. In defense, the willingness to learn someone else’s tradecraft is a signal of commitment, and is the secret sauce behind the “forward deployed” model. That is the reason that the Conductor team sees their primary bottleneck as hiring cleared people, not compute.

Learn more about Conductor AI: conductorai.com
Follow Zach: https://www.linkedin.com/in/zachary-long-20a78a21/
Follow Eric: https://www.linkedin.com/in/epschwartz/
Follow Noah:

For more Crossing the Valley: www.ValleyCrossers.com

They were four people. They had ~$150,000 in the bank. And John Conafay wrote a check to an accreditation partner. That check was easily a double-digit percentage of the company’s value at that moment. And he risked it all on the conviction that a Space Force contract was going to land.

Well, this is Crossing the Valley, so you already know that it did.

Three years later that relationship became a $25 million, five year SBIR Phase III with Space Systems Command.

We’ve covered a lot of acquisition orthodoxy these last few years on CTV. This conversation was fun because of how John went a different path, and bucked convention.

He spent money he didn’t have.

He started the transition conversation before he’d delivered value.

He hired strangers to knock on doors for him.

And he pitched the hardest version of his product first.

Here’s a bit more from my awesome conversation with John Conafay.

About John

John is a US Air Force veteran and the founder and CEO of Integrate. He enlisted after a failed run at a metal and hardcore record label he started in high school, and spent his service in COMNAV fixing AWACS on the flight line in Alaska. He calls it “a swift swift kick” that he needed. Those early days on the flight lane have paid dividends to this day, by helping the “business guy” hold his own in engineering conversations.

After leaving the Air Force John studied economics at Arizona State, where he took over the national chapter of Students for the Exploration and Development of Space (SEDS). It was during this time that he ran a coordinated campaign to win a donation from one particular individual: Mr. Jeff Bezos, himself a former SEDS chapter president. John and team had no idea they’d succeeded until Bezos folded up his $250,000 Heinlein Prize check and announced that it should go to SEDS instead.

His career spanned a variety of startups, including joining BryceTech, becoming employee number eight (roughly) at Astranis, Director of Business Development at Spaceflight, and Head of Business Development at ABL Space Systems. And he did all this before founding Integrate.

About Integrate

Integrate builds multiplayer program management software for organizations that build complex hardware. Founded in early 2022 and based in Seattle, the company sells what John describes as “a schedule that many organizations can share, without seeing each other’s private data.”

John describes the status quo they’re up against as “Microsoft Project,” a piece of software that shipped in 1984.

The company won a $1.25 million SBIR Phase II with the Space Systems Command Mission Manifest Office in 2023, then a $25 million five year Phase III in June 2025. It reached IL6 accreditation and deployed into a classified environment in under twelve months.

Key Takeaways

1. Conviction comes from seeing the same problem multiple times

John describes how he hit the same wall at Astranis, at Spaceflight and at ABL, time and time again. At Astranis he was managing 180 suppliers while also running federal and commercial business development, and he spent his days on a loop: receive, download, import, reconcile, export, send, so that somebody else could do the exact same thing. He was too busy reconciling risk data to analyze it.

At Spaceflight he tried to build the fix internally. They were flying 20 to 60 customers on a single launch, 25 launches a year, 3,000 tasks per customer, and none of those customers could be allowed to see each other. He could not make it work inside the company. He tried again at ABL.

It was hard to see this as anything other than a massive opportunity after he’d failed to solve it three times from inside three different orgs.

2. Build for the worst environment first

John’s rule was that anything the team shipped for a commercial customer should also be deployable for / by the government. He didn’t want 5 different stripped-down versions; he wanted it to be the same software.

That conviction came from having lived the alternative. We see it again and again: when you’re looking to buy a commercial product you can get a great piece of software. But when you move it into a CUI environment, you get ten percent of the capability. John and Integrate wanted to build “something our warfighters and our acquisition professionals can actually use.”

That decision cost money in the short term, because it required removing external dependencies. The team had to exercise extreme discipline on the software bill of materials and examine where it’s open source code came from. They had to deploy to AWS GovCloud from day one, at meaningfully higher cost than the commercial alternative.

But John emphasizes a critical point, which is that Commercial to CUI was not the moat. Plenty of companies have made that jump. The money gets made when you move from CUI to SIPR, as this is where the market thins out. For Integrate, their Phase III conversation did not become real until Integrate got to SIPR. The customer told them directly: happy to talk pricing, but if you cannot get us into classified environments, none of it matters.

3. “Spend money to make money”

The truism is silly, but in the case of Integrate, it was just straight up accurate.

John paid Second Front to get them to prod before the seed round closed, before the Phase II money, and before anyone asked for it. He bet it all on getting working software deployed.

Three weeks later he had a customer on NIPR and in CUI environments, using working software.

There is a huge difference between a demo and a customer laying hands on keyboard. John still recalls the reaction: “oh, you are actually delivering things.”

4. Rent the door knocking to earn the room

John knew he did not have the government relationships up front, and he was unwilling to wait the months / years it would take to develop them for scratch. So he is candid: he paid contractors with connections to book meetings for him.

Access can be a cost in government go to market; you still have to close.

He tells a great story of a meeting that felt like it was destined for disaster. He was scheduled to meet with a lieutenant colonel and a major. They were busy, so he got pushed down to a second lieutenant fresh out of the academy. He pitched her anyway. She took it back to her chain of command, and a captain called him 45 minutes later to ask whether the product actually existed because it was solving such an important pain point.

The honest answer was “sort of” but that call changed the game for John and Integrate as they’d found a maniacal believer in what they were selling; someone with acute pain who couldn’t take it anymore.

5. Find the fastest path to impact

When I asked John what he would tell a young founder, he said, “go straight into the smallest startup you can find and skip everything else.”

The network and pattern recognition he built in those years is a large part of why Integrate moved as fast as it did. There is no substitute for the frying pan of a small company pushed to move fast, with radical accountability and high stakes.

Call for Startups

Go solve disinformation. John thinks the proliferation of bots and synthetic video is one of the genuine existential problems of the moment.

For More:

• Listen to the full episode:

• Integrate: integrate.co

• John Conafay on LinkedIn: linkedin.com/in/conafay

So when he tells you not to build hardware, it should mean something.

Josh started Gambit in 2023, in the middle of the drone gold rush, and deliberately did not build a drone. The decision to forego hardware has shaped the company he’s built, and the impact they’re having just a few years into the journey.

About Josh

Josh grew up in Pittsburgh, the son of a nuclear reactor engineer and an electrical engineer. His sister is an engineer. His wife is an engineer. His brother-in-law is an engineer. Suffice to say, the dude has forever been surrounded by tech nerds.

His formative moment came at twelve, in the garage, watching his dad fix the car. Josh assumed his father was the smartest man alive who knew everything. But his dad corrected him that day: he actually did not know how to fix the car. But he knew how to solve the problem.

Josh went to Penn State, then Stanford for grad school, and interned at NASA Glenn at 21, which he describes as “underwhelming compared to the movies.” A colleague pointed him toward a small space company in Los Angeles, so he found himself interviewing with an up-and-coming entrepreneur named Elon Musk in the window between Falcon 1’s third and fourth flights, before SpaceX had ever reached orbit. Musk described building a reusable rocket engine rated for a hundred flights. Josh asked why anyone would want that. The answer was that they were going to land the rocket.

Josh went on to become one of about five engineers who built the Merlin 1D. It sits in the Smithsonian now. He left when SpaceX had grown from 400 people to 2,500 (still the largest company he ever worked for), because he could tell the founding-level experience was no longer available there.

Then came Virgin Hyperloop, where he served as CTO and later CEO across roughly seven years. Six years after starting, he sat inside the vehicle his team had built and put his own life on the line.

About Gambit

Gambit builds a platform-agnostic software layer that lets one operator orchestrate many autonomous systems across air, ground, and sea. The thesis is that robots are commoditizing, platforms are proliferating, and no single manufacturer is going to rule them all. The constraint becomes less about machine availability and more about the people to run them.

Gambit thinks about their product as an entire robot workforce rather than a “better robot.” Each entity is a specialist that works together toward a cohesive whole, much the same way a soccer team has forwards and a goalie, or a band has a guitarist and a lead singer.

The product architecture reflects the thesis: roughly ten software containers, from behavior building to composition to runtime interfaces, that can be split apart and sold in pieces. Behaviors are solved by platform class rather than by manufacturer, so a quadcopter class or a fixed wing class covers many vendors with a parameter file rather than a bespoke integration.

The company exited stealth in January 2026 with partnerships including AWS, L3Harris, RTX, and Sierra Nevada. Since then it has announced two multi-million dollar Air Force contracts, an Army GVSC OTA for counter-UAS simulation, a red force as a service partnership with the Pendleton UAS Range, and integrations with ModalAI and Psionic.

Key Takeaways

1. Disqualify your first market before it costs you two years

Josh’s first use case was missile defense. A salvo of incoming missiles, a set of interceptors, and the weapon target pairing problem between them. It seemed to be a clean technical fit.

He went to a Missile Defense Agency innovation conference and found them recycling slides from three years earlier. “Even the jokes were recycled,” Josh recalls. He walked out understanding that the organization was wedded to its primes, and he backed away.

Saying no in the early days is virtually impossible. There are quite a few founders who would have spent eighteen months trying to make that door open, because the technical fit was so obvious. Josh took one look, read the tea leaves, and pivoted. That willingness to act quickly on a disqualifying signal should not be overlooked.

2. If they will not buy the product, sell them a piece of it

Government cannot easily buy software. It buys boxes. That reality has pushed a long list of software companies into building hardware they never wanted to build, and Josh watched Gambit run into the same wall in the small reconnaissance market.

His answer was architectural. Every customer Gambit talked to already had something. Lattice on one piece of hardware here, an Auterion Skynode on a piece there, a Skydio fleet in teh distance. Josh again took a contrarian path, recognizing that if he gave the government an all-or-nothing ultimatum, most of the time they’d choose nothing.

So Gambit built roughly ten containers and can shred out chunks of them for whatever the customer needs. Gambit’s runtime can sit underneath Anduril’s Lattice, so when an operator tasks a platform in Lattice, Gambit is doing the work on the edge. Gambit gets data and abstraction it would not otherwise have. The customer keeps the interface it already trusts.

3. Your best demo is the one you do not run

Anyone who has never touched Gambit’s software can walk over and be flying multiple platforms inside five minutes. The company frames demos as observer-led. It’s not “watch Gambit do it” but “here, you do it.”

It’s a bold bet but if you can change the narrative from “I saw a guy fly a drone swarm” to “I - a novice who has never held the sticks - flew a drone swarm with Gambit” you are installing yourself in that person’s cold storage.

4. A champion is someone who already believes what you believe

Josh’s definition of a champion is precise, and I loved it because it differs from a traditional defense tech contracting definition. A champion is not someone who likes you or appreciates your product.

A champion is “someone who arrived at the same conclusion independently.”

The one he names came out of AFRL in Dayton, and the belief was that more robots are coming and the force is woefully unprepared. The trap was that the organization was trying to solve everything at once. Josh’s contribution was to break it down to one specific problem: red teaming. Counter-UAS units need multi-drone threats to train against and cannot fly people to generate them.

That narrowing turned into Phase 1 SBIRs, which turned into Phase 2s, which turned into a set of red team contracts. There is also scar tissue to work through in this market. Programs like CODE, OFFSET, and AISUM promised collaborative autonomy years ago and underdelivered, and Josh has had investors who worked on those programs tell him flatly that what he is doing is not possible. His counter is that the mathematics were elegant but the techniques available then were not the ones available now.

5. Speed of modification is the actual product

Six weeks before we spoke, a technical point of contact told Josh he could see A and B, and asked whether A plus B could equal C. Specifically, could Gambit turn an electronic warfare capability from jamming into spoofing, so that a group of cheap objects could be made to look like expensive ones doing something they are not.

Gambit turned it around in three weeks.

Josh notes that almost every contract Gambit has landed in recent months followed the same shape: a conversation, a realization, a response. When they see the speed of delivery, programs find the money to add.

A closing note

Crossing the Valley is rarely an emotional podcast, but I want to share the part of my conversation with Josh that will stay with me well beyond the GovCon lessons.

It was the personal part.

Unexpectedly, we got into what the first company cost him. Separate bedrooms with his wife, a marriage that nearly ended, and the choice between building his company and saving his family. When the Eclipse seed round landed, the bottom fell out at home, and he called his investors to say he did not know if he could be what they needed him to be. Eclipse told him they were investing in him, and to take the time he needed.

As someone who has thrown myself into work at the expense of many other things in my life, I just want to thank Josh for being willing to dive into a difficult topic. And to share with you all the rule he landed on: “feedback within 24 hours.” If a partner, cofounder, or employee is surprised by something, that is a communication failure.

Thanks for reading - and hope you get something useful out of it.

For more on Gambit: https://www.gambit.us

For more Crossing the Valley: https://www.youtube.com/@crossingthevalley

Follow Josh Giegel: https://www.linkedin.com/in/josh-giegel/

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

About Brian

Brian Taylor spent about fifteen years as a mechanical engineer before he ever sat in the founder’s chair. He came up through aviation and aerospace, then landed at SpaceX, where he did development, qualification, and testing on Starlink satellites, including the structural qualification of the first stack of 60. From there it was Amazon’s Project Kuiper for high rate satellite production, then Loft Orbital.

But despite this tour d’force through a who’s who of space companies, by his own admission, Brian was never a space nerd. It really bugged him that you would work that hard on something, and unless you happen to be an astronaut, you would never see the hardware again. You do not get to iterate on it. You do not get to enjoy the fruits of your labor.

An engineer who wants his hardware back. That frustration became a company.

About Lux Aeterna

Lux Aeterna builds fully reusable satellites. While there’s been lots of attention heaped on the rockets, Brian and Lux are focused on the satellite itself. The goal is to fly the mission, survive re-entry, land intact, refurbish it, put a new payload on it, and send it back up.

Brian founded the company in Denver in August 2024. It came out of stealth in June 2025 with a $4 million pre-seed led by Space Capital, and closed a $10 million oversubscribed seed led by Konvoy in March 2026, bringing the total to $14 million.

The company’s first vehicle is called Delphi. It is roughly 200 kilograms, built around a rigid conical heat shield that doubles as the satellite’s primary structure, with a parachute assisted return. It will fly on a SpaceX Transporter rideshare in Q1 2027, with recovery at the Koonibba Test Range in South Australia through a partnership with Southern Launch. The company holds a Space Act Agreement with NASA Ames and two CRADAs supporting the re-entry and thermal protection work, and has stood up a Defense Advisory Board.

The entire payload capacity for that first mission is already sold out, across civil, defense, and commercial customers, hardware and software, small companies and large.

Also worth noting for the culture watchers: he did this interview from inside a sauna. An actual, off the shelf sauna, sitting on the production floor, because it was cheaper than the fancy sound-proof pods that Google puts all over its offices. The idiot index on that thing has to be excellent.

What follows are the five ideas from this conversation that stuck with me:

1. The best deep tech companies are built on constraints.

Building reusable satellites did not require discovery of an entirely new set of physics; rather, the project required somebody to notice that the reason why nobody was building them finally ceased to be true.

In the early days of satellites, launch was so expensive and so rare that you optimized mass above everything else. Under that constraint, a disposable satellite is the correct engineering answer. Then rideshare and Falcon 9 reusability collapsed the cost and the scarcity of launch, and the optimization function changed underneath the industry.

As Brian says, “With launch much more plentiful than it was 10, 20, 50 years ago and much lower cost, you can turn the dial on that mass optimization a little bit to get some other features like re-entry and reusability.”

This meant deliberately accepting a mass penalty (e.g., heat shield, parachute, return propulsion) to buy reentry, and therefore, reusability.

Thus there was an assumption (and a constraint) that suddenly ceased to hold.

2. If nobody wrote the requirement, nobody else is competing for it. You also own the entire burden of proof.

Brian makes a point of saying that there was no program office asking for reusable satellites.

“There’s not a requirement written, there’s not an RFP that says, ‘We want a short duration mission.’”

Which poses a pretty significant challenge, because the responsibility of making the case lays with the company (and the founder). For Lux Aeterna, the pitch connects a couple of elements the Department of War thinks about:

Picture a timeline with launch in the middle. Lead time on one side, mission duration on the other. When you need a short lead time, whether that is prototyping, development testing, or a mission specific operation, you usually have a correspondingly short mission need. You can leave the satellite parked up there afterward. But at that point, as Brian puts it, you have built a target for one of our adversaries to go inspect.

As we’ve often said on this program, if you are responding to a proposal you did not shape, somebody else shaped it, and you are already late. Category creation is just requirements shaping at the earliest and most valuable possible moment.

But Brian does not oversell it, and I appreciated that.

“The onus is on us to prove that this is a useful thing and not be a build it and they will come kind of situation.”

3. Their moat is cycle time (not re-entry).

Every satellite company on earth iterates on telemetry, which is to say, on data downlinked from hardware they will never physically touch again. Lux Aeterna gets telemetry and the vehicle back on a bench.

“You can take it apart and look at everything. You can do non-destructive testing. You can do destructive testing if you want.”

Brian sees this as a compounding internal advantage. Every single flight generates a class of ground truth that competitors structurally cannot obtain, and it feeds straight into the next generation vehicle. His thesis is that a company capable of learning from full teardowns of flown hardware is going to out-iterate a company learning from downlinked packets, and that gap will widen with every mission.

4. Your hiring rubric does not travel. Your process rubric does.

After fifteen years in engineering, Brian thought he had hiring solved. He would look for the same things in engineers: Did they study mechanical engineering; did they work as a mechanical engineer; what did they build; did it work. A clean, quantitative, and repeatable process.

Then he had to hire sales, marketing, PR, and business operations, and found out the rubric fell apart when extended to other domains.

So instead of trying to create a rubric for each function, he sought to abstract from all of them:

“The idea of the sales process, pipeline, funnel, deal stages, that’s actually applicable to a lot of respects.”

Put hires through a funnel, and “focus on filling the company with people that are much smarter than me in every respect.”

5. “We don’t want to become a prime” is a bad excuse for having no process.

Brian made the observation that lots of startups treat “process” like the thing that turns you into a prime. But he rejects the framing entirely, and gives you a test you can use tomorrow:

“If you can fit it into a couple of bullets or a half-pager and it simply gets written down, how would I teach myself to do this six months from now? That is the right amount of process.”

Or said differently, “Process is just make your life in six months a little bit easier when you have to do this and you haven’t done it in six months.”

Process and bureaucracy are bad words that do not have to be bad. We are taking them back. Process is liberating when it is not enslaving, and it gives you the confidence that if you follow the cookbook, you land in the right place.

For more on Lux Aeterna: https://www.luxaeterna.com/

Follow Brian: https://www.linkedin.com/in/brianjtaylor/

For more Crossing the Valley: www.valleycrossers.com

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

Cyber operations have always lived in the shadows of the national security conversation. In the wake of the Department of War suspending CMMC Level 2 guidance, this episode unpacks the essential, high-stakes, and rarely understood world of cyber security, as Sam Jones, CEO and co-founder of Method Security, joins Noah to talk about building autonomous offensive and defensive cyber capability for the U.S. government and the country’s most critical enterprises.

We go deep on why the “post-Mythos” moment has changed the calculus for every CISO in America.

About Sam

Sam Jones is the CEO and co-founder of Method Security. His path into cyber started early: an internship at GDIT the summer after high school introduced him to network penetration testing and defensive systems work, and a SMART scholarship later routed him into Air Force Cyber in San Antonio as a civilian operator, where he progressed from a GS-7 to a GS-11 …in just one year. He left convinced the government's real problem wasn't talent acquisition but talent retention. ("It's nice to have these short tour duties to get people in, but you want to create the criteria to keep people around.")

Sam had wanted to work at Palantir since he was 18, after stumbling on early Gotham demos on YouTube. Rejected from his first internship application in 2010, he came back and joined after leaving government in 2014, spending five years there building cyber products and meeting his two future co-founders, Sean and Dan. After Palantir, Sam joined Shield AI as roughly its 25th employee, working on autonomy for robots operating in real-world, no-fail missions. This experience planted the seed for Method’s core idea: applying that same flavor of self-directed autonomy to cybersecurity.

That idea sat dormant, passed back and forth between Sam and his future co-founders for years, until GPT-4 arrived in 2023 and gave them the technical ingredient they’d been missing. Within six hours of Sam raising the idea, both co-founders were in, and the three quickly quit their jobs to found Method Security.

About Method Security

Method Security builds autonomous cyber systems for the U.S. government and critical enterprises, applying autonomy to a blend of offensive and defensive missions, from red teaming to vulnerability research and penetration testing. The company’s core bet is that organizations need to move faster than adversaries without sacrificing trust, governance, or control, and that this capability needs to be operator-owned and enterprise-governed. The unifying question: how do you secure systems faster, in-house, to outpace adversaries, while keeping the human operator in command?

Founded in September 2023 by three Palantir alumni, Method quickly closed a seed round with Andreessen Horowitz within about a week of pitching. Rather than immediately going to market, they went heads-down for roughly six months focusing entirely on building the product they knew they needed. When they finally came up for air, they quickly landing their first Fortune 500 design partner (around nine months in), and their first federal contract soon followed, in the form of a direct-to-Phase-II SBIR with the Space Force (around eleven months in). Deliberately, Method has never split its team into separate commercial and federal go-to-market functions — Sam describes it as intentionally absorbing the “pain” from both markets into one product organization so that tension shapes the platform from the ground up.

Today, Method serves both government and Fortune 500 customers, with a particular focus on institutions that have the most to lose.

Top 5 Takeaways

1. People are more important than the idea. Method’s founding team spent nearly five years trading startup ideas back and forth in shared Notion docs before landing on the one that stuck. The team and the working relationship mattered more than any single idea. Early stage investors are quick to highly cofounder pairings as a key determinant of success or failure, and Method’s team had de-risked their relationship through years of collaboration at Palantir and beyond.

2. One team, two markets. It is rather unusual to see a single team tackle both sides of “dual use” simultaneously, without hiring differently. Cybersecurity is likely one of the few markets in which such an approach could succeed. Method intentionally runs a single product team that feels pressure from both markets simultaneously. Sam compares it to the strongman “Hercules Hold” — enduring the pain of both sides at once and translating it directly into product.

3. Heads-down beats over-iterating. After closing their seed round, the founders deliberately limited outside feedback and focused on building conviction around what the market needed, rather than chasing every piece of user input. Too much feedback, too early, can pull a founding team away from the thing they were actually built to build.

4. The “post-Mythos” moment reshaped the market. As frontier-class AI models become available to adversaries, boards are now directly asking their CISOs what their response plan is. That question has become one of Method’s most effective go-to-market unlocks. At the same time, Sam is wary of blanket regulation of large language models themselves, arguing the real risk lies in how models get wielded by software platforms, not in the models alone, and that overregulating relative to unregulated open-weight alternatives could put U.S. companies at a disadvantage.

5. A correct prediction is worth more than a fast pivot.
   It’s worth reiterating that Method didn't iterate its way to product-market fit. The co-founders deeply studied the market for a decade, formed strong opinions, and stayed heads-down through a dark first year while outside voices pushed different directions. When the moment came, Method was an obvious answer with a deep, already-published technical answer. That kind of conviction is a massive unlock.

For more on Method Security: www.method.security

For more on Crossing the Valley: www.valleycrossers.com

Follow Sam: LinkedIn | X

Follow Noah: LinkedIn | X

Crossing the Valley is a podcast on technology, defense, and national security. Watch the full conversation with Sam Jones on YouTube, and find Method Security’s careers page if you’re interested in joining their mission.

Before starting Turion, Ryan spent nine years as an engineer at SpaceX, joining not long after the company’s sixth flight and working under Kevin Wu — an engineering leader Ryan credits as one of the best of his generation (now leads engineering at Relativity Space). Ryan’s interest in space dates back to 2002, when he first saw an NPR segment about the discovery of exoplanets that set him on his life mission.

About Turion Space

Turion Space is a space technology company built around a long-term vision: extracting economic value from space beyond satellite communications — ultimately, asteroid mining. Rather than raising all of the (massive) capital that vision would require up front, Turion is executing what it calls the “Starforge” strategy: decomposing that future architecture into discrete, scalable business portfolios — tactical space, space logistics, geospatial/ISR, and an underlying mission software and ground-operations stack (branded Starfire OS) — each capable of generating revenue and proving technology on its own timeline.

Turion went through Y Combinator after months of rejection from accelerators and VCs unfamiliar with space’s capital requirements. Since then, the company has closed an oversubscribed $85M Series B, secured a STRATFI, and won a position on Space Systems Command’s RG-XX IDIQ — a program for next-generation proliferated surveillance and reconnaissance in geosynchronous orbit.

Key Takeaways

1. Decompose the moonshot into fundable pieces. Turion’s founders concluded early that an end-state vision like asteroid mining “is not a financeable business” as a single bet; the infrastructure, time, and cost required to go from zero to one is too large to finance directly. Instead of chasing a massive up-front raise, they built a roadmap of smaller, scalable business portfolios that each generate near-term revenue while collectively building toward the larger architecture.

2. Don’t underplay your strongest customer for narrative’s sake. Turion originally positioned itself as commercial-first and avoided classified programs, believing government work would slow the company down. Ryan now describes that as underplaying the reality that U.S. national security is the largest buyer of new space technology. There was no avoiding this reality.

3. Execution is table stakes; relationships are a differentiator. Winning a seat on RG-XX against far larger, better-resourced primes came down to a simple baseline: demonstrating a plan and then executing it to cost, schedule, and performance. But execution alone isn’t sufficient. He still had to build the relationships and understanding of government acquisitions to convert their momentum into contract awards.

4. Sequence diversification deliberately, and build adjacencies on existing capability. Ryan saw early on that staying in just a single mission segment risked capping the company’s valuation, to the point where his only viable exit strategy would be an acquisition. At the same time, he needed the internal discipline to resist “shiny object syndrome.” Today, when Turion does move into adjacent portfolios, like space logistics, it does so by extending capabilities it already has (e.g., a maneuverable, vertically integrated vehicle capable of rendezvous and proximity operations) rather than starting a new capability from scratch.

5. When a customer shows you what they care about, listen. A pivotal moment in Turion’s early growth came during a Direct-to-Phase-2 SBIR briefing, when a government customer told the team plainly that they didn’t care about the low-Earth-orbit capability Turion had spent significant effort analyzing. Rather than defend the existing work, the team reran the full analysis for other orbits within roughly two weeks — a fast turnaround that Ryan credits as directly contributing to the STRATFI, and eventually the RG-XX award, that followed.

For more Turion Space: https://www.turionspace.com/

For more Crossing the Valley: LinkedIn | YouTube

Follow Ryan: https://www.linkedin.com/in/rwesterdahl/

Follow Noah: X | LinkedIn

Jackie is the founder and CEO of Whitespace. She began her career supporting the counter-IED mission during the surge in Iraq, working as a red-team analyst. She was a threat emulator who studied the tactics, ideology, and bomb-making techniques of adversary networks, often in Arabic on the dark web, then fused that with classified collection. That experience, and witnessing what timely intelligence does for operators downrange, became the through-line of everything she’s built since.

Jackie returned to grad school to study what changed after 9/11, and got pulled into an R&D community around Activity-Based Intelligence (ABI) — a methodology that made trained analysts roughly 300x more effective. Notably, Jackie has a liberal arts background, not an engineering one. And, fun fact: she founded Whitespace in 2014 when her first child was six months old (there’s something about young parents as founders…)

About Whitespace

Founded in 2014 and headquartered in Alexandria, VA, Whitespace builds sensemaking AI for defense and intelligence. The company bootstrapped for roughly a decade, funding technology development through consulting revenue, before raising a small seed round in 2025 (~$3.2M per public reporting) backed by family offices and a few VCs.

Its early IP, Worldline, is a simulation engine that creates a digital twin of a city-sized population observed by a synthetic sensor architecture, to generate realistic, multi-sensor training data. That capability won Whitespace its first prime contract with the National Geospatial-Intelligence Agency (NGA) in 2017 to build the ABI curriculum for the entire intelligence community, which the company then instructed for nearly eight years.

The flagship product today is Iris: an agentic AI “analyst” that lets operators self-serve pattern-of-life intelligence in natural language. Under the hood, Iris is built on ABI tradecraft translated into a toolkit of deterministic algorithms, wrapped in an agentic system carrying the persona and judgment of experienced analysts. Iris can run standalone, alongside human SMEs, or headless via API — and is designed for a future where her biggest user group may be other agents and autonomous platforms, not people.

Notably, by the time the company raised venture capital, it had flipped from ~80% services / 20% product to ~80% product / 20% services, 4X’d ARR, doubled revenue, and roughly doubled headcount.

Key Takeaways

1. Whether it’s services or product revenue, proximity is the most important early return. Whitespace’s consulting work kept Jackie embedded with the exact analysts and operators who would later define the product, and bought time to figure out the right way to deliver before committing. Revenue close to the mission is the next best form of R&D funding because it pays you to stay in the room.

2. Sequence hard problems rather than stacking them. Early on, Whitespace needed to navigate two meaningful transitions: services to product, and government-only to dual-use. Each is a company-defining transformation on its own. Founders rarely fail because a transition is impossible. Instead, we see time and time again they fail because they’re running several major transitions all at once and starving all of them from the attention they need.

3. Successful products are often the translation of expertise into software form. Whitespace’s hardest, highest-leverage bet was converting a qualitative analytic methodology into production-grade algorithms. This is where most expertise-driven companies stall, for two structural reasons: data science and engineering are genuinely different disciplines (“works in Python” is not “works at scale”), and the engineer has to deeply understand the analyst’s intent. The companies that bridge that gap (in Jackie’s case, by hiring the rare people who have both areas of expertise), are able to build something competitors can’t easily copy.

4. Discipline is knowing when to stop discovering. Growing in the software market can create a seductive trap: every new use case is a chance to prototype, deliver value, and move on. The risk is that you end up producing a pile of one-offs that never become a repeatable product. Crossing the valley requires deliberately cutting off the discovery process and forcing the work to coalesce, even when the next shiny problem is tempting and the customer paying you isn’t the one with the most interesting problem. In a word, focus is the difference between crossing or dying.

5. Brand is strategy when it tells the truth. Whitespace’s bright, almost psychedelic identity is a deliberate departure from defense tech’s default black-grey-gamer aesthetic. The logic is substantive, not cosmetic: the people who do this work are fiercely optimistic, and breaking the visual pattern creates room to rethink intelligence itself. In a crowded category of near-identical “thing-drops-from-the-sky, coming-soon” launch videos, refusing the template has garnered significant organic attention for Jackie and co. That’s because their brand aesthetic signals a genuinely different way of approaching a known problem.

What Stood Out to Me:

• The decade-long bootstrap. We are in an era of fast, rapid raises. Jackie’s path — 11 years, one small seed by choice — is almost unfathomably today.

• A non-technical, liberal-arts founder built a hard-tech AI company. And she did it before the ubiquitous LLMs made it such that every one of us is a coder. Jackie is candid about the false starts (an entire AR/MR company she founded alongside Whitespace in 2018 that was too early) and the scars from chasing one-offs.

• Young parents = great founders. Something about having a six-month-old requires radical focus. The realization that time, not money, is the truly scarce resource.

• The product philosophy is sharp and forward-leaning: Iris is “she,” headless, and built on the bet that the biggest future user of intelligence agents will be other agents.

For more on Whitespace: inthewhitespace.com

Follow Jackie: https://www.linkedin.com/in/jackie-barbieri/

For more Crossing the Valley: valleycrossers.com | youtube.com/@CrossingTheValley

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

Phil Duong is the CEO and co-founder of NODA AI. The son of Vietnamese immigrants who arrived in the U.S. with little money and speaking little English, he grew up in a family of service members who had fought for their freedom. His heritage left him with a deep sense of patriotism and a drive, as he puts it, to repay the country “multiple lives’ worth.” He was appointed to West Point but walked away from it when a childhood medical issue blocked him from branching aviation, calling the Marine Corps instead because they needed pilots and were willing to give him a shot. After leaving active duty, he built a career at enterprise-AI company C3, rising to general manager of its federal practice before leaving to start NODA. He frames himself first and foremost as “a Marine and patriot,” a self-description that shapes how he runs the company and what he’s willing to say in public.

About Paige Craig

Paige Craig is the founder and general partner of Outlander VC. Growing up homeless, and raised in what he describes as a violent world, he was named for a grandfather who rose from enlisted soldier to an officer under General Patton. Paige was recruited to West Point — where he claims the distinction of being the most-punished cadet in his class — and left in his third year, recognizing he was “the wrong guy” for the institution’s barracks mentality. He moved on to the Marine Corps and the intelligence community, and in 2003 built a private military company operating at and beyond the forward edge in Syria, Afghanistan, and Africa, which he sold. As an investor, his entire lens is the founder and the mission, evaluated against a proprietary 38-point framework built around vision, intelligence, character, and execution.

About NODA AI

NODA AI is a startup building the orchestration and decision layer for autonomous warfare. Rather than building drones or vehicles, NODA aims to overlay government-designed tactics and strategy onto the autonomous platforms other companies are building. Their software seeks to re-create what a strike cell or operations center does today, fast enough to run in real time as robots take over more of the battlefield. The idea is that battle plays for defeating advanced threats need to be baked into deployable algorithms in advance, ready to push down to autonomous boats, planes, and submersibles the moment competition turns to conflict. The company started as a five-person team that prototyped for roughly a year unpaid, and is now working across what Phil will only describe as “very topical” combatant commands.

About Outlander VC

Outlander VC is an early-stage venture firm focused on writing first checks into exceptional, often low-signal founders others haven’t yet recognized. The firm runs a high-volume, high-selectivity funnel. Paige says that over 10,000 founders come in through the top of the funnel each year, with roughly 5,000 passing AI filters to a first call, a few hundred meeting the team in person, and 10 to 12 ultimately receiving funding. Paige keeps more than half his calendar deliberately unscheduled so he can take long walks with founders and go deep on the person rather than the pitch deck. The firm now has seven partners Craig has trained.

Top 5 Takeaways

1. A relationship forged over quitting (West Point) and joining (the Marines)

Both Paige and Phil got into West Point. Both left before graduating. Both then went to the Marine Corps. That’s where the similarities end; Craig left because he chafed against the rules and the “barracks mentality” — he says there “wasn’t a rule I couldn’t break” and he simply didn’t fit. Duong actually loved the regimented life but left anyway, giving up the prestige because a medical issue blocked him from flying and he wanted to be an aviator badly enough to walk. When you think about how rare that series of events must be to begin with, the backbone of this relationship becomes a bit clearer.

2. The first check got done over 72 hours in Florida without a pitch deck (opened)

Craig flew Duong out to his place in Rosemary Beach, on the Florida Panhandle, and they spent three days together without opening a single PowerPoint slide. Phil recounts falling for the investors as much as Paige fell for him. Because they’d lived somewhat parallel lives with shared values derived from their youth, Phil was convinced that Paige would “turn over every stone” to help him succeed.

The degree of connection between these two made me question what I thought I knew about a “first check” relationship.

3. The NODA AI team worked (mostly) for free for the first year

When Phil resigned from C3 AI, his team followed him without so much as a product thesis. A group of senior engineers, data scientists, product designers, and his co-founder got up and simply left because he was leaving (what he calls his “Jerry Maguire moment”). The inspirational showing of belief gave way to concern as the VC they’d been counting on fell through after 6 months, but the team stuck together. This group of folks with kids and real opportunity costs simply kept building, unpaid, for nearly a year. Paige shared that this level of belief and sacrifice is in fact simply normal for the rarest 0.001% of founders. But that in itself is a rarity.

4. Leaders eat last

It’s perhaps a truism that in the Marine Corps, leaders eat last. Not as a metaphor but as practice: people line up in reverse rank order so that if food runs short, it’s the senior people who go without. Phil runs NODA the same way. When the early checks were small, the people building the product got paid first; he ate last. He hands every manager and executive he hires a copy of the book on the idea. His larger point is that leadership is the most abundant, untapped resource a founder has — it costs nothing but character — and that most founders, and most VCs, have no idea how to recognize or use it.

5. The “triad,” and the beach demo that beat the people with real budgets

Duong’s framework for surviving government acquisition is what he calls the triad: the warfighter who owns the problem, the program office that represents them, and Capitol Hill that funds it — and you have to work all three vertices at once, not in sequence. But the most vivid moment is how NODA broke through. With no money to market, in a field where billions were being poured into autonomy and a lot of “vapor” was being sold, Duong told the government to simply show up to a beach. There, a five-person startup demonstrated with real vehicles and real integration the thing that better-funded players had been writing articles about for a decade. Craig’s point is that the government has created a mile-wide, inch-deep surface area where lots of people win small programs. But it’s seriously challenging to land the big one fast.

For more about…

• NODA AI: https://www.NODAintelligence.ai/

• Outlander VC: https://outlander.vc/

• Crossing the Valley: www.valleycrossers.com

Follow:

• Phil Duong: LinkedIn

• Paige Craig: X | LinkedIn

• Noah Sheinbaum: X | LinkedIn

His appearance on Crossing the Valley has been a long time coming!

This week, John joined us to talk about Cape — a privacy-first cellular carrier he started in 2022 after a decade at Palantir and a tour as a Green Beret communications leader.

The simple version of the story is “ex-Palantir guy builds secure phone company.”

The truth, of course, is a whole lot more interesting.

Here’s what stood out:

1. John chose to build the version of the company that’s almost impossible to copy.

There were at least two easier companies hiding inside Cape’s problem. One was a privacy app. The other was a cybersecurity consultancy that helped existing mobile carriers (e.g., AT&T and Verizon) patch their own networks. “That would’ve been a great business,” John admitted. He built neither.

See, the vulnerabilities John and co care about are network problems — they come from “transitive trust” in cellular: your phone trusts the network, the network trusts every other network, and the whole patchwork leaks your location and activity to anyone who knows where to look. You cannot fix a network problem from a device or an app. You need network footprint. Follow that one premise to its end and you don’t get to pick the easy company. You’re building a global cellular network whether you like it or not. Cape doesn’t build towers or own spectrum; it writes all the software that runs the network and rents towers wherever it needs them.

So in the case of Cape, correct diagnosis led to the founding team to a logical conclusion that required them to do the most difficult thing they could imagine.

2. Naïveté is a gift.

John raised a $5M seed and genuinely believed it would get Cape live. “I was like, we just need a couple more engineers, and a couple off-the-shelf vendors to round out the stack, and then we should be able to get live.”

It took $50M.

He calls this “the luxury of naivety.” If he’d fully priced the build up front, the spreadsheet may have talked him out of starting. So that early ignorance wasn’t a mistake to be overcome. In fact, it was the grease in the skids, and the only reason the Cape team stands to build a moat at all.

The painful founder lesson here was John’s advice on conviction: first, decide whether it’s an important problem. Second, decide whether you believe in the solution. If both are yes, “don’t let how hard or how expensive it’s gonna be even enter the calculus.”

3. John would gladly help a competitor.

Asked about fast-followers, John said his instinct would be to help anyone trying to start the same business. “I’d help them go faster and avoid all the pain and the crazy things you get wrong the first time.”

It’s not hubris, nor is it stupidity, it’s just a man who has spent four years and $50M learning exactly how long the road is. Advice alone isn’t going to shortcut the work.

4. The deal that made Cape was won in a red team.

Cape’s inflection point was a single exercise in Kansas City: a realistic simulated hostage rescue where rescuers used ordinary Pixel phones on Cape’s network, layered over 100 of another carrier’s towers. The opposing red team featured US Marshals doing close-in geolocation, plus a credible group with insider access simulating a compromised US telco provider. They couldn’t locate, let alone track, the operators.

That result was the Series A, and it was the leverage that got the carriers to take meetings. But the deeper lesson is about sequencing.

Cape was fairly confident it could beat the close-surveillance team because a lot of its early identifier-rotation work was built for exactly that. It was much less sure about the insider/compromised-carrier threat, because that’s nearly impossible to test outside a real scenario.

Stepping forward into the breach, despite the risk and uncertainty, ended up being the takeoff moment that turned the tide for the company.

5. Wins compound and unlock new doors.

After Cape beat the insider-threat red team, the company caught the attention of Navy CTO / PEO Digital Justin Fanelli. Justin owned a very specific problem: cellular connectivity in the Pacific. Even more specifically, connectivity on the island of Guam, which traditionally relied on using carriers compromised by China.

Justin recognized the implication of the KC result: if you can provide trusted connectivity over a known-compromised carrier, that changes how you think about communications in theater. So he sent Cape to Guam to prove it. They did. And that single validated hypothesis — trust over compromised infrastructure — opened a whole new set of use cases that drove much of the last year’s growth.

This is the most repeatable strategic pattern from the conversation: each win exposed another, more valuable problem.

John was clear onwhat he’s after: becoming “the United States’ answer to Huawei” — the place where American capital and talent accumulate around telecommunications, the strategic asset the US conspicuously lacked during the 5G race. It’s a much bolder vision than “a better privacy plan,” and one of the more audacious infrastructure bets in the game today.

For more on…

Cape: https://www.cape.co/

John Doyle: https://www.linkedin.com/in/john-doyle-48633227/

Crossing the Valley: LinkedIn | YouTube

Noah Sheinbaum: LinkedIn | X

Share

Featuring Cameron McCord, CEO and co-founder of Nominal

Share

About Cameron

Cameron McCord has seen the defense ecosystem from more seats than most. He started his career as a young submarine officer, operating, in his words, with “the best of 1980s technology” and learning what it feels like to be beholden to antiquated tech when the mission is on the line. He carried that frustration into operating roles at Anduril and Saildrone, where he watched fast-growing hardware companies scale and saw the same testing problem appear at every one of them. Then he went to Lux Capital, where he saw the same pain across a thousand more companies and turned a hunch into a thesis.

Today, that thesis is a company called Nominal, one of the fastest growing, well-capitalized, and highly touted defense tech companies in the industry.

About Nominal

Nominal is a software and data platform for hardware testing operations. The hardware test market is a bit anathema for us non-technical folks. But the reality is, anyone who builds anything in the real world, whether for a commercial market or the warfighter, has to test and validate it. Until recently, that was mostly done via spreadsheets, MATLAB scripts, and hard-to-reach data. Nominal took many of the processes that befuddle founders and seasoned executives alike, and automated them. The company was incubated at Lux Capital. They intended to serve both commercial and defense customers from day one. And they’ve been on a tear, recently announcing a sole-source $53 million IDIQ for the Air Force Test Center. Today, Nominal is active in the Navy at Pax River, developed partnerships with MIT, and has some exciting news with DARPA.

What follows are the five ideas from this conversation worth stealing.

1. It’s time to stop arguing about “speed versus quality” and just make risk a number instead.

Today, when a program wants to put a new weapon on an old aircraft, the default answer is to rerun the same three thousand test points, because in the absence of better information, nobody can defend doing less. That takes three years, sometimes five, at a couple million dollars per sortie.

Cam’s vision is to replace that with a single, defensible number. Imagine a program manager opening Nominal and seeing that a weapon system is at TRL 8.45 today, backed by every individual data point, every packet of telemetry, every log behind it. Run next week’s tests, burn down four or five test points, and the platform projects you to 8.72. At which point a leader can look at the cost of chasing the last fraction and decide it is not worth it.

The ability to quantify how much risk you are actually retiring with each test unlocks a totally different decision calculus in the boardroom, in the Pentagon, and on Wall Street. The job becomes eliminating, with maximum rigor, the test points you can prove you no longer need.

2. The customer who criticizes you is sometimes worth more than the one who loves you.

We all want to be loved. But in the early days of Nominal, Cam learned to chase the opposite. The most valuable champion in Nominal’s early life was a Brigadier General who would send long, direct emails laying out exactly where the product was falling short.

That feedback became the company’s product roadmap.

Though they found themselves having to go through a crash-course in aerodynamics and modeling to speak their customer’s language, the early Nomineers benefitted immensely from finding someone who cared enough to tell them when they were wrong.

As the old saying goes, the opposite of love is not hate, but ambivalence.

This has been a recurring theme among valley crossers: “The worst thing that can happen is you get death by yes. You just keep getting told it’s great, yes, and then you kind of look and you’re like, no, like no one’s using this. They’re just saying yes to the PowerPoint.”

The definition of a champion is someone willing to try, and who cares enough to tell you the truth.

3. Slowing down is death in venture-backed defense.

Cameron has a specific warning for the wave of veteran founders entering defense tech. The exact discipline that makes operators excellent, their ability to quantify and manage risk, becomes a liability when it turns into taking endless baby steps.

“Pretty much every time I have convinced myself into slowing down at Nominal, I’ve kind of regretted it.”

Cam occasionally found himself talking himself into pushing a raise or a decision off two or three months, then watching how fast a performing business actually moves and wishing he hadn’t waited.

NOTE // caveat emptor: not every defense business is a venture-scale business, and the reverse is just as dangerous. The skill is not simply “going fast” but knowing the market, understanding the opportunity cost, and building the kind of business that attracts the kind of movement you need to be successful.

4. If you stumble into dual-use, you have already lost.

“I generally think if you stumble into a dual-use business, it’s probably a bad idea.”

Nominal was architected to serve commercial and government customers from day zero, because the decision shapes everything from how you build the product, to how you sell, to how you handle feedback.

Although “dual use” continues to be a hot investable category, it’s pretty precarious on the inside. What happens when the systems get out of whack? When commercial customers flood you with feedback, feature requests, and bug reports as usage starts to grow? In an accidental dual use company, that may mean pulling resources from the federal side, and delaying growth.

A company architected for day one dual use must build to survive that asymmetry.

For Nominal, the strategy was to start small, narrow, and focused, to build to the bigger vision. Nominal’s first government work was a small Phase I SBIR with the Air Force Test Pilot School. They knew that this institution that was never going to be a massive revenue source. But they solved one very narrow, very real problem, turning students who spent ninety percent of their time as data engineers into students spending ninety percent of their time as test pilots. And that single, specific win became the foundation for a $53 million enterprise deal.

5. If you make the requirement testable, the competition runs itself (much more effectively).

As we’ve seen in recent months, defense acquisition is (finally) moving away from the white paper. Programs like Drone Dominance and the hypersonics effort from the Test Resource Management Center (TRMC) and T-REX, to name a few, are putting vendors on the range and asking blunt questions that are answered by performance.

Cam (like many of us) is bullish on this shift, and Nominal has a particular vision for how this will play out.

Challenge-based acquisition is great, but there’s a way to go even faster than getting every company out to a range. Let’s say you have five vendors competing for the same contract. They all say they can do the thing. They show up with data to “prove it” - but it’s in different formats, using different metrics, and they each have a different way of describing their performance. Short of architecting an exercise (which remains a bit more time intensive), the government still has to figure out who is actually telling the truth. If you can’t wrangle all the engineering data into something comparable, that final decision still gets made with a surprising amount of subjectivity.

Cam’s answer is to make the requirement itself measurable before anyone competes. The government writes down what it wants a system to do, and that requirement gets translated into a concrete set of outputs: workbooks, checklists, defined test points. He calls it digitizing the Design Reference Mission.

It’s a vision worth keeping an eye on.

After speaking time with Cam, I left convinced that Nominal really is going after the single biggest (remaining) roadblock to meaningful acquisition reform. If we don’t solve test, we cannot meaningfully shrink the divide between cool ideas in R&D and capabilities that are validated and safe to deploy. If we’re entering a tech boom cycle (and if you’re even a slightly AI-pilled reader, it’s not hard to imagine that we are), then the need to quantify risk, readiness, and the value of a test (especially one you can skip) becomes a whole lot clearer. Cameron’s whole approach is a bet that the way across the valley of death is not to test less or test faster in a vacuum, but to finally be able to prove, with data, exactly how much testing is enough.

For more on Nominal: LinkedIn | Website | X

For more Crossing the Valley: valleycrossers.com | YouTube

Follow Cam: LinkedIn | X

Follow Noah: LinkedIn | X

About Ed Mehr

Ed Mehr is the CEO and co-founder of Machina Labs. A builder since childhood (he attended a school that emphasized welding, carpentry, and craft) Ed started his career in software before crossing over into hardware as an early engineer at SpaceX in the early 2010s. The cultural shock of moving from “ship a prototype in days” software to “wait months for a single iteration” hardware became the founding wound that drove him toward agile manufacturing. After SpaceX, he joined Relativity Space to work on 3D-printed launch vehicles, then founded Machina Labs to build the missing infrastructure he believed was the real bottleneck to America’s next industrial age: a general-purpose manufacturing platform that didn’t require part-specific tooling or CapEx. He’s a self-described capitalist and car guy, and prints small batches of his own shirts when an idea strikes him. Kinda badass.

About Machina Labs

Machina Labs is building “RoboCraftsman.” These are autonomous robotic cells that form sheet metal parts the way a potter shapes clay. Two robots come at a flat sheet of metal from either side and progressively deform it into complex 3D geometries, eliminating the need for the dies, molds, and multi-story stamping presses that have defined sheet metal manufacturing for a century. The company serves a dual-use customer base spanning commercial (Toyota is both a major customer and an investor) and defense (Air Force depots, missile primes, NASA). Machina has already deployed production cells at Air Force depots, where they’re making C-130 panels and other legacy aircraft parts that previously required hunting down donor airframes in junkyards or waiting four years for new tooling. The company is roughly 70–80 people, hiring aggressively, and approaching the end of qualification testing with parts on track to fly soon. Machina is now entering its scale phase, and Ed teases that a missile-program prime announcement is imminent…

Key Takeaways

1. The door to bits is closing. The door to atoms is opening.

Ed thinks we’re at a generational inflection. For the last twenty years, the smartest people went into software because that’s where iteration was free and the upside was infinite. AI is now eating that world by saturating it. The next cohort of ambitious people is going to look at code and think, what’s left for me to do here that the model won’t do better?

So they’ll look up. And conveniently, the sky just opened back up too. Starship is real. Hypersonics are real. Space is suddenly tractable in a way it hasn’t been since Apollo.

In this new reality, Starship is the new Harley-Davidson. After WWII, GIs came home, bought motorcycles, and went looking for national parks. We’re about to do that again, just with a different machine and a much bigger map.

Whether you buy the prediction or not, it’s a clean explanation for why so many talented people are suddenly making things you can drop on your foot.

2. We have a bit compiler. We don’t have a matter compiler. That’s the opportunity.

As Ed explains, a CPU is a bit compiler: you give it intent, it gives you computation. We’ve spent eighty years getting absurdly good at this. The entire software economy runs on the fact that turning an idea into bits is essentially free.

But there is no current equivalent for matter. If you want a physical part, you still need tooling, dies, factories, supply chains, lead times measured in quarters. Every physical company in America is bottlenecked by this, even SpaceX. That’s why their Falcon and Starship lines live in two completely different buildings with two completely different sets of tooling.

So Ed explains that Machina isn’t really about sheet metal at all… it’s his attempt to build the first general-purpose matter compiler. That’s a much weirder and more ambitious bet than “we make aircraft panels faster,” and it explains why Toyota is on the cap table next to the Air Force.

3. We are relying on the junkyard to keep military cargo planes in the sky.

When a C-130 — the workhorse cargo plane the U.S. military has been flying for 60+ years — loses a panel, the Air Force has two options. Option one: pay six figures for new tooling and wait four years for the part. Option two: go to a junkyard, find another damaged C-130, and cannibalize the panel off it.

That’s it. Those are the choices. The most-used aircraft in the U.S. military is being kept alive by scavenging.

We are in a moment of serious strategic competition, the Pentagon is talking constantly about production capacity, and the actual sustainment model for legacy fleets is we hope there’s a wrecked one in the desert with the part we need. Ed’s company turned that four-year wait into a one-week wait, which is great, but the more interesting takeaway is what it reveals about the baseline. The industrial base atrophied so quietly that the workaround became “go shopping in a graveyard.”

4. The stockpile math is genuinely alarming.

In one recent conflict, the U.S. used 30–40% of its stockpile of certain missile systems. To replenish those stockpiles at current production rates would take three to four years. And that’s replenishment. That doesn’t account for the kind of surge production we’d want to see if things escalated.

This is not about manufacturing efficiency. It’s about our ability to simply produce what we need as a country. The story has been told a bit in the media, but what Machina is doing is fundamentally different than other approaches. It’s attempting to give us a new method to make material here at home.

5. Ed’s “Han Solo” test beats out the company “values” page.

Ed has two unrelated answers that I think are actually the same answer.

First: how does he know his technology is genuinely cool? Machina formed a 12-foot sheet metal Han Solo in carbonite and brought it to a trade show. Kids stopped and took pictures. “Kids are very honest critics,” he said. “They don’t have the social norms. They’re just doing their thing.” If kids think it’s cool, it’s cool. If they walk past it, you’re lying to yourself.

Second: how does he know who to hire? “If I gave this person a credit card and an objective, would they get it done?” If yes, hire. If you have to think about it, don’t.

Both are tests designed to bypass the part of your brain that wants to be polite, professional, or socially graceful about a judgment call. No need to construct a fancy rubric.

I don’t know if these are universally applicable, but whatever the heuristic for your company, knowing what you’re looking for, and communicating it simply, goes a long way.

For more on Machina:

• Careers: machinalabs.ai/careers

• Follow Ed: X | LinkedIn

For more Crossing the Valley:

• Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

• Substack: valleycrossers.com

• YouTube:

About John Hong and Tallis Chang

John Hong is CEO and Tallis Chang is COO of Obsidian Sensors. The two have worked together for over 40 years, with backgrounds in aerospace before joining Qualcomm. At Qualcomm, they were part of the team that spent nearly a decade developing low-power displays using MEMS (Micro Electro Mechanical Systems) built on glass substrates. When Qualcomm decided the program no longer fit its core business, John and Tallis were given the runway to find a new application for the manufacturing know-how they had built. That pivot became Obsidian Sensors.

About Obsidian Sensors

Obsidian Sensors, headquartered in San Diego, manufactures microbolometers — the core sensor inside thermal imaging cameras — using a fundamentally different platform than the rest of the industry. Where incumbents builds them on silicon wafers, Obsidian builds these sensors on display glass, leveraging LCD fabrication infrastructure in Japan and Taiwan that already produces millions of phone and watch displays. This shift dramatically reduced unit cost while preserving sensor performance.

The company spun out of Qualcomm in 2017 with seed capital, transferred equipment, seven team members, and five patent applications. It initially targeted the automotive market before pivoting to defense after receiving a DIU OTA at the start of COVID. Today, Obsidian is a critical supplier of thermal cameras to the US drone ecosystem, with production scaling from 5,000 units in 2025 to a planned 1 million units per year by the end of 2026 at its Top Gun Street facility in San Diego. A Phase 2 facility targeting 10–20 million units annually is planned for 2028.

Key Takeaways

1. The structure of your spinout determines the trajectory of your company. Qualcomm transferred five patent applications outright to Obsidian in exchange for equity, not royalties. That clean structure made it possible to raise from outside investors later without a strategic acquirer’s claims hanging over the cap table. These favorable IP terms gave the company a viable path, where it otherwise may have been stuck on the launchpad.

2. The market that funds you may not be the market that scales you. Obsidian was built for automotive, and automotive remains part of the long-term plan because of its volume profile. But automotive qualification cycles are too slow to validate a startup. The DIU OTA that arrived during COVID gave Obsidian working capital, a reason to stay operational during shutdowns, and a proof point they could carry into the rest of defense.

3. Cost transformation often comes from changing the manufacturing platform, not the product. Thermal imaging has been a defense-critical capability for 50 years, but the cost curve plateaued because the industry kept building on silicon. Obsidian’s insight was that the sensor problem could be moved entirely onto LCD glass, which meant the fabs already existed at scale and could run microbolometers alongside their normal display production.

4. Build the IP wall before the imitators arrive. Obsidian has 40+ patents granted and roughly 80 more pending, and their tech transfer to fab partners happens under export license. John mentioned the inbound interest from “weird emails” with “strange English grammar” almost in passing, because the legal infrastructure to handle that interest was already in place. Especially for hardware companies with defense applications, the IP architecture is a core part of the product. It cannot be retrofitted under pressure.

5. This is a crucible moment for American supply chains. Obsidian is building Phase 1 for 1 million units a year while current orders are at 5,000, and Phase 2 for 10–20 million while Phase 1 isn’t yet complete. The bet is that demand signals from customers walking into their facility are stronger than any forecast a procurement office will publish. Even without signed deals, the trends are clear: make it here in the US, at scale, as fast as possible. That’s a function of newfound clarity from the Department of War and the US government focusing much more attention and capital on domestic supply chains. And that’s pretty darn exciting.

Follow John: https://www.linkedin.com/in/john-h-973b372/

Follow Tallis: https://www.linkedin.com/in/tallischang/

Learn more about Obsidian: https://www.obsidiansensors.com/

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

Follow Crossing the Valley: YouTube | LinkedIn

Brian Spurlock is VP of Growth and Strategy at Persistent Systems, and a self-described “boomerang” employee, meaning he’s had two separate tenures at the company. He came up as a Special Operations acquisitions officer at Fort Bragg, which means he’s seen the Persistent Systems story from both sides: as the government buyer evaluating the technology and as the company executive selling it. He left for a stint at AWS, returned, and is now the one figuring out how to bring tactical mesh networking from Special Operations into Army divisions. He is based in the Fort Bragg / Pinehurst area of North Carolina.

About Persistent Systems

Persistent Systems builds tactical mesh networking hardware and software: specifically, the Wave Relay and Cloud Relay systems, delivered on hardware called the MPU (currently MPU5, with MPU6 in development). The idea is simple and hard: create a mobile RF network that moves with operators, passes real data at real speeds, and requires zero thought from the person holding the radio. Founded in 2007 by two Johns Hopkins PhDs — Herb Rubens and Dave Holmer — the company has never taken outside capital. It manufactures in Manhattan. It has customers across U.S. Special Operations, conventional Army (currently scaling into division-level exercises with 4th Infantry Division), the Air Force, NATO allies, and public safety. It is on the verge of Army program-of-record status after nearly 20 years.

Key Takeaways

1. Persistent Systems bootstrapping a hardware company in defense for 18 years, and lived to tell the tale.

Persistent Systems has never taken a dollar of outside investment. That’s a constraint that shaped the company’s entire character. When you can’t hire ten people, you decide carefully about one. When you can’t burn cash proving out a market, you make sure the product works before it leaves the building. Brian describes the early years as a “patchwork of stories” — LAPD, John Deere, Army National Guard soft units — not a clean TAM story, just finding pockets where the technology solved a real problem and using that revenue to fund the next iteration. The absence of a VC clock means there’s no turn in the road where the car has to go 105 mph. It also means every dollar of growth is going directly back into the business.

2. SOCOM was the best place to stress-test a product for Persistent Systems. It is not necessarily a pathway to conventional Army adoption.

This distinction matters more than most people realize. Persistent Systems spent years earning trust in the Special Operations community, replete with honest feedback, serious users, and operationally intense environments. That feedback made the product better. But when the Army’s 4th Infantry Division started running NGC2 exercises at Fort Carson, something completely new appeared: 700 nodes on a single network. No SOCOM unit would ever need that. The RF physics, the network management, the latency behavior — all of it changes at that scale in ways that smaller networks simply don’t expose. So you may go to SOCOM because they’ll tell you the truth and they’ll use the product hard. But you cannot go there assuming that love from tier-one units automatically converts into Army programs of record. Those are different customers with different requirements, and you have to earn both separately.

3. The most underrated growth metric in defense tech is a question: are customers still bringing you their hardest problems?

When pushed on north star metrics (e.g., what does a VP of Growth at a bootstrapped company use to know if the business is heading in the right direction?) the answer wasn’t ARR, contract value, or unit count. It was: are the people we work with continuing to bring us problems they can’t solve anywhere else? Brian frames the inverse compellingly: when a user stops bringing you hard problems, that tells you something. You’ve either stopped being useful, or they’ve found someone else. The fact that the Army is now doing that — presenting Persistent Systems with problems at a scale they’ve never encountered — is itself the signal that the relationship is healthy and the trajectory is real. It’s a qualitative north star, but it’s a creative way of evaluating where a company stands.

4. Staying quiet while you’re figuring it out is a legitimate competitive strategy.

Brian makes an observation that cuts against the current defense tech marketing moment: if you stay quiet and fail early, you can recover. If you get loud — press, LinkedIn, podcast circuit, conference panels — and then deliver something that doesn’t work, that failure becomes your brand. In the Special Operations community especially, a product that doesn’t perform gets talked about. You don’t get a second shot at a first impression with that user group. Persistent Systems spent years being almost invisible publicly (indeed, this is their first podcast!) building reputation through product performance rather than narrative. The tradeoff is slower awareness growth. The return is that when you do show up, you show up with 18 years of product that actually works.

5. The next unsolved problem in tactical networking has less to do with connectivity and more to do with how to use data once it flows.

Brian’s “call for startups” moment is the clearest articulation of the gap he’s watching from his seat. The pipe problem is largely solved: there are many players ensuring that data can move. However, the knowledge problem remain. While platforms can display dots on a map, how to signal to a commander which dots require human judgment and which can be handled by machine, in real time, under cognitive load, in a kinetic environment, is still wide open. The companies that solve the data-to-decision layer — not another dashboard, not another COP, but genuine decision support that knows the difference between machine-executable and human-required — are the ones he’s watching for.

BONUS: Engineer where the conditions are hardest, not where they're cheapest. As a fellow New Yorker, I have to mention that Persistent Systems didn't pick Manhattan for talent or proximity to capital. Instead, the company located here because NYC is the toughest RF environment on the planet. Dense urban canyons, signal reflections, interference, and mobility all conspire against tactical networks. By choosing to engineer in the worst-case environment every day, the company built products that just work when they get to Mosul, the Hindu Kush, or a Pacific archipelago. See, told you NYC was the new defense mecca ;)

Follow Brian: https://www.linkedin.com/in/spurs/

For more Persistent Systems: https://persistentsystems.com/

For more Crossing the Valley:

• LinkedIn: https://www.linkedin.com/company/crossing-the-valley/

• YouTube: https://www.youtube.com/@crossingthevalley

ABOUT ARIYAN KABIR

Ariyan Kabir is the co-founder and CEO of GrayMatter Robotics. He started the company in early 2020 alongside two co-founders, building on PhD research conducted at USC’s Center for Advanced Manufacturing, where he and one co-founder studied the intersection of robotics and AI. The third co-founder, chief scientist Satyandra K. Gupta, has spent more than 35 years working at the intersection of AI, robotics, and manufacturing and sits on multiple national boards.

Ariyan’s framing of the problem is shaped by a realization that came from the academic environment: that manufacturing (long assumed to be “solved”) is in fact one of the largest unaddressed application areas for modern AI. The US is currently short half a million skilled manufacturing workers, a gap projected to reach four million in seven years and put $2 trillion of annual GDP at risk.

ABOUT GRAYMATTER ROBOTICS

GrayMatter Robotics builds the autonomy layer for high-mix, high-variability manufacturing, the roughly 90% of factory work that cannot be addressed by traditional pre-programmed robotics. The company’s robots perform tool-manipulation tasks: sanding, grinding, polishing, buffing, coating, blasting, and inspection. Underneath the robots sits the company’s real differentiator: a foundation model for materials and processes, trained on what Ariyan describes as the largest manufacturing process dataset in the world: 7 petabytes of multimodal sensor data spanning 14–18 modalities per cell.

The business is now six years old. This interview was recorded at the company’s 100,000 sq ft Physical AI Experience Center in Carson, California, which is the company’s fifth facility. Today, GrayMatter’s customers include Boeing, Raytheon, Oshkosh, Caterpillar, Riddell, the US Navy, the US Air Force, and Huntington Ingalls Industries, where GrayMatter is a partner on the HYPR (High Yield Production Robotics) program designing the next-generation American shipyard.

The company sells outcomes, not hardware. Robots are deployed on a subscription basis with multi-year commitments, and performance compounds quarterly through software updates, recipe upgrades, and hardware refreshes.

KEY TAKEAWAYS

1. Sequence the market. Commercial first, defense second.

GrayMatter spent its first three-plus years building almost entirely in the commercial world, despite the obvious dual-use relevance of factory automation. Their reasoning was that defense customers tolerate zero failure, and an unproven autonomy stack cannot debug itself inside an aircraft program. By the time the company turned toward national security work, they had already validated the technology, the business model, and the deployment muscle.

2. The business model can be the product.

Performance Composites — GrayMatter’s first customer — had been searching for a robotic sanding solution for five years before the conversation with GrayMatter. IT wasn’t so much a capability they were missing as it was a different risk appetite: manufacturers cannot tie up millions of dollars in Capex on unproven technology. GrayMatter’s response was to invert the deal entirely: zero upfront cost, deployment first, subscription payments only after the system was producing value in production. In exchange, the customer committed to a multi-year contract. The structure became the wedge that opened the entire Fortune 100.

3. Adoption became the geostrategic moat.

Asked how 7 petabytes of US manufacturing data compares to South Korea or Shenzhen, Ariyan reframed the question. The race isn’t about dataset size. The US ranks near the bottom of the top 10 in robot-to-worker adoption ratios; South Korea is #1 and China is climbing fast. Tonnage-wise, China has built more ships in the last 12 months than the US has built since WWII. This reframing has direct strategic implications: a defense industrial base strategy that focuses on technology sovereignty without obsessing over diffusion velocity will lose to one that does the opposite.

4. GrayMatter builds AI into every layer.

GrayMatter identifies eight distinct layers of engineering work that sit between a customer problem and a delivered outcome: industrial engineering, solution engineering, process recipe engineering, cell design, tool/fixture design, implementation, sustainment, and digital-twin operations. Most robotics companies put AI in one layer: the robot. This means the other seven can become bottlenecks. But GrayMatter is now deploying domain agents at every layer, coordinated by an orchestration layer the company calls Factory Super Intelligence (FSI). The focus on under-appreciated elements of the solution set has helped them differentiate.

5. Hire from the intersection. The mixing room doesn’t work.

GrayMatter’s most expensive hiring lesson was that you cannot assemble a great physical AI team by recruiting the best AI specialists, the best manufacturing engineers, and the best roboticists into a room. The translation cost is too high; each discipline optimizes for its own definition of the problem. What does work is hiring people who already live at the intersection of at least two of the three domains, then leveling them up on the third. The company’s head of customers came from Tesla. Their head of aerospace and defense was two steps below Boeing’s CTO. Their head of hardware also came from Tesla. This was a counterintuitive lesson, but one that stuck with me most from the conversation.

Follow Ariyan: https://www.linkedin.com/in/ariyankabir/

For more GrayMatter: https://factory.graymatter-robotics.com/

For more Crossing the Valley: www.valleycrossers.com

Ben Cipperley spent 26 years in the Navy. The first 18 he served as an Explosive Ordnance Disposal office, with three tours across the Indo-Pacific, three years between Iraq and Afghanistan, and command in Guam. Which is to say, he has more experience than most.

What makes his background unusual isn’t the operational experience; plenty of veterans make the transition. It’s the Pentagon chapter that came after. A fellowship at the Stimson Center led to a resource sponsor role writing the Navy’s budget, which led to being pulled onto the team drafting the 2022 National Defense Strategy (mid-process, while Russian forces were marshaling in Belarus). Which led to working under the CNO on what became the Navy’s first force design document in 50 years. It’s a classified vision for the fleet in 2045 that, he notes, had quite a few robots in it.

He put an “Open to Work” banner on LinkedIn when he retired. The CEO of Havoc called him 20 minutes later.

About HavocAI

HavocAI is a collaborative autonomy software company. The pitch is deceptively simple: one operator, thousands of autonomous systems. The technology is the connective tissue, the software layer that lets a surface drone, an aerial drone, and a ground vehicle share a common operational picture and respond to commands from a single interface, rather than requiring separate operators, separate control systems, and separate data pipes for each platform.

The company was born from a conference convened by then-Undersecretary of Defense Heidi Shyu around the concept of “Hellscape,” IndoPaCom’s deterrence concept for a Taiwan scenario involving massive swarms of autonomous systems. The question in the room wasn’t whether the platforms existed. It was: how do you make thousands of them work together when a human brain can only track so many things at once? Founders Paul Lwin and Joe Turner quit their jobs the next day and started building the answer.

Early validation came through Silent Swarm, an exercise run by Naval Surface Warfare Center Crane. Havoc showed up with 12 prototypes of their Rampage platform. Representatives from NIWC Atlantic watched, then bought the prototypes on the spot. Those units shipped to Portugal to support Task Force 66. That’s the loop Havoc is now trying to close, and why they recently acquired Mavrik (air) and Teleo (ground) to make good on the multi-domain part of the pitch.

The business model is worth understanding clearly. The government is bad at buying software. It tends to need software wrapped in a hardware container before it can conceptually acquire it. Havoc has internalized this and partnered with eight different maritime platform builders, plus air and ground hardware partners, so that their autonomy stack arrives pre-integrated with best-of-breed hardware. They’re trying to be the OS layer that makes everyone else’s platforms work together. But they’ll sell you the hardware too, because that’s how you get the check.

The Lessons

1. The problem the government actually has isn’t the problem it says it has.

Ben draws a sharp distinction between what defense procurement documents say and what operators actually need. There are no formal capability development documents specifying what a medium unmanned surface vessel needs to do. The requirements that exist are top-level (e.g., speed, payload, endurance) without a mission profile to lay them against. This creates a strange situation: companies are trying to satisfy requirements that aren’t fully written yet, for operators who know what they want but can’t always articulate it in the language acquisition needs.

Havoc’s origin story reflects a clearer version of this: the problem wasn’t the robot. It was the software to make robots work together. Nobody had written that as a requirement. A senior leader named it in a room, two founders heard it, and they went and built it.

2. “Works” is the word doing a lot of heavy lifting.

I am always curious about what people mean when they say “the product works.” I pushed Ben on this at length. His response was that “works” doesn’t mean end-to-end, zero failures. It means you’ve demonstrated a high enough Technology Readiness Level that the government believes you’ve solved a piece of their problem and are worth investing in further.

But there’s a harder version of this. Noah describes watching companies show up to exercises, fail three times, succeed on the fourth attempt, and walk away saying “see, it works” — while the warfighter in the room is mentally writing them off. Ben’s answer: if it doesn’t work, it goes in a Conex box and you never see it again. The military’s decision calculus is binary in a way that startups often underestimate. One bad demo doesn’t just cost you that contract. It costs you credibility with that community for years.

His rule of thumb for whether to attend a government test event: if you can’t define what your success state looks like before you arrive, don’t go. If the government hasn’t clearly outlined the operational problem, and you can’t articulate how your technology solves at least part of it, you need more turns before you’re ready to demonstrate.

3. Scale is the argument, not capability.

The case for autonomy isn’t that autonomous systems are better than destroyers at every task. It’s that you can lose 10% of a drone fleet in a day and replenish it in weeks. You can deploy hundreds of systems fast, replace them fast, and iterate on payloads fast. The learning cycle compresses from years to days. Against an adversary who has spent 30 years developing anti-access/area-denial weapons specifically designed to threaten large surface combatants, the risk calculus of putting a $2.3 billion target in a contested strait is, as Ben puts it, off the charts.

The argument is economic and strategic. And it’s an argument defense tech founders often undersell, because they’re focused on what their product can do rather than what the alternative costs.

4. The insider path is real, but it’s not a shortcut.

Ben’s transition looks clean from the outside: senior Pentagon official with deep relationships joins autonomy company, brings credibility and access. The reality is more complicated. He spent years building understanding at every level of the system, from budget to requirements, acquisition, strategy, and operations.

When he was in the CNOs office, he saw Havoc at an investor conference, brought it back as a recommendation, and helped the CNO’s office start paying attention to what defense tech startups were actually building.

That kind of insider value takes decades to build. Which means the shortcut version (e.g., hiring a recently retired flag officer to open doors) usually doesn’t work the way founders hope. The doors might open, but what matters is what you do once you’re inside.

5. The multi-domain bet is a systems integration bet, not a platform bet.

Havoc’s goal isn’t to own the best aerial drone or the best ground vehicle, but to be the software layer that makes a surface drone, an aerial drone, and a ground vehicle operate as a single coherent system under unified command.

Ben described a hypothetical operational scenario: a maritime ISR platform sees a fast-moving target but can’t track it over the horizon. If it can launch an aerial drone and pass targeting data between them through a common autonomy stack, you’ve created networked warfare from a single operator interface. The value derives from the connections between platforms, and in making those connections simple enough that the person running the operation doesn’t need three separate control consoles and three separate teams.

For more about HavocAI: https://havocai.com/

Follow Ben: https://www.linkedin.com/in/bencipperley/

For more Crossing the Valley: valleycrossers.com

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

Crossing the Valley explores the journey from proof of concept to production in defense technology. New episodes weekly.

Luke Fischer is the CEO and co-founder of SkyFi. He is a former U.S. Army helicopter pilot and Special Operations veteran who transitioned through DIU (Defense Innovation Unit), Uber Elevate (as Head of Flight Operations), and Joby Aviation (leading government operations) before founding SkyFi in late 2021. He also spent time as an Entrepreneur in Residence in VC, which gave him direct exposure to what gets funded — and what doesn’t. That multi-domain journey across military, Big Tech, and venture is core to how he built SkyFi.

About SkyFi

SkyFi is a geospatial intelligence marketplace and platform that aggregates satellite imagery and analytics from hundreds of commercial satellite operators — making it accessible to any customer, anywhere, via web or mobile app. Rather than owning or launching satellites, SkyFi sits at the aggregation layer: abstracting away which provider owns the satellite and letting users task imagery or pull analytics with a button press. The business is currently approximately 80% government / 20% commercial, serving customers ranging from top global hedge funds and agriculture companies to Special Operations Forces, NATO, and allied governments across Asia. SkyFi closed its Series A in December 2024.

1. The most strategically valuable position in a fragmented market isn’t owning supply; it’s owning demand data.

Every satellite company in the geospatial ecosystem has a narrow, siloed view of what their customers want: just the buyers who came to them, for their specific imagery type. SkyFi sits above all of them. Because they aggregate demand across hundreds of providers, they accumulate something no individual satellite company can build — a behavioral map of who is buying what, where, and when, across the entire market. Luke is explicit about this: it’s how Uber used location data to place bikes, scooters, and Uber Eats in specific cities. The analytical products SkyFi builds aren’t guesswork; they’re derived from four years of observed customer intent at scale. The business that processes demand will always know more than the business that only fulfills it.

2. When a government customer tells you their own internal tools are useless, write it down. That’s your pitch.

Luke didn’t make up the claim that commercial imagery can solve 97% of what classified assets are currently used for. A government customer told him that. Classified imagery systems are slow, bureaucratic, and rationed; commercial imagery is instant, cheap, and available to anyone with a credit card. The mismatch between what exists and what gets delivered to the warfighter is a distribution problem. SkyFi’s entire government business is built on that one honest sentence from an operator who was frustrated with their own system.

3. Protecting customers becomes a competitive strategy.

Satellite providers — who are simultaneously SkyFi’s supply chain and potential competitors — routinely asked for SkyFi’s customer names, framing it as diligence. Luke refused. He reverse-engineered why they were actually asking (Salesforce, i.e., direct outreach) and built a better diligence process to neutralize the argument. The suppliers eventually came back around because they trusted SkyFi more than the brokers who gave up their customer lists. In this aggregator model, the moment Luke let his suppliers touch the demand side, he would’ve handed them the roadmap to cut him out. The customer list is the moat.

4. Dynamic pricing for geopolitical risk is coming to satellite imagery.

This one didn’t get much airtime, but it’s worth sitting with. SkyFi tracks heat maps of customer tasking behavior. When tasking activity spikes around a particular geographic area, that’s a signal that imagery of that area is becoming more valuable. Luke has Uber’s former dynamic pricing engineers on his team specifically to build this. In a world of escalating conflicts and rapid geopolitical shifts, the pricing of intelligence will eventually reflect real-time demand, just like surge pricing reflects a concert letting out.

5. The biggest adoption barrier for new technology is…

Awareness. As Luke and SkyFi went through NATI DIANA, they found that allied defense ministers and government officials consistently didn’t know that commercial satellite tasking was instant and cheap. They assumed the James Bond spy satellite model — months of budget process, classified clearances, exquisite assets. The awareness gap is real, and it’s not the gap between “interested” and “buying.” It’s the gap between “has no idea this exists” and “interested.” It’s a very different sales motion if you’re informing vs. persuading. And it means the addressable market is almost certainly larger than anyone’s current TAM model accounts for.

For more SkyFi: skyfi.com

For more Crossing the Valley: valleycrossers.com

Follow Luke: https://www.linkedin.com/in/lukemfischer/

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

Todd Stiefler is VP of Enterprise at WHOOP, overseeing the company’s B2B and B2G business lines. He brings an unusually cross-functional background to the role: a decade as a Senate defense staffer focused on missile defense and strategic nuclear deterrent, followed by product and commercial leadership roles at GE’s industrial IoT division, LMI, and Palantir. A self-described “tech nerd” and ultra-marathoner who finished the Leadville 100 in 27 hours, Todd found WHOOP the way most people find things that change their lives — by accident, scrolling LinkedIn late one night. He’s been building the company’s defense business ever since.

About WHOOP

WHOOP is a Boston-based wearable technology company focused exclusively on human performance and health optimization. Unlike general-purpose smartwatches, WHOOP does one thing: continuously measures high-fidelity physiological data — heart rate variability, resting heart rate, respiratory rate, skin temperature, and strain — and translates it into actionable metrics around sleep, recovery, and readiness. Its early customers included elite professional athletes, and it has since grown into a mass-market product worn by hundreds of thousands of consumers and, increasingly, military personnel. WHOOP is actively pursuing the DOD enterprise market, partnering with programs like SOCOM’s POTD and the Army’s Holistic Health and Fitness (H2F) initiative. The company’s enterprise division, led by Todd, handles everything from individual unit programs to large-scale institutional deployments.

Key Takeaways

1. Build for your most demanding user first.

In contrast to the “minimally viable product” framework, WHOOP’s early decision to build exclusively for elite athletes — with LeBron James and Michael Phelps among their first 100 customers — wasn’t a marketing stunt, but a product strategy. By targeting the most demanding performance environment, they built something rigorous enough to eventually serve everyone else. That same logic attracted special operators before conventional forces, and conventional forces before institutional DOD programs. They solved for the “edge cases” first rather than the average user, because they were solving their distribution challenge through their product design.

2. But the right product alone isn’t enough… you still have to solve the process.

Todd was direct about what WHOOP got wrong in its early defense push: they had a best-in-class product but hadn’t built the surrounding infrastructure. No contract vehicles. No data sharing agreements. No pricing architecture tuned to government procurement. “The DOD buys from people that solve their process, not their problem.” Getting that right — not just shipping a great device — is what actually enables scale. So while many of us laugh at high-ceiling IDIQs without real funding, the vehicle is a key ingredient.

3. The wearable itself is actually not the intervention.

This was one of the most important points from the episode: motivated, self-selected consumers who buy WHOOP on their own are signaling their desire to change their behavior. Warfighters handed a device in a conference room, with no context, no coaching, and no program design, often won’t. Even though the average warfighter may be more attuned to their body than the average civilian, in this case, if the warfighter doesn’t opt in, the product won’t deliver. This insight is why Todd never sees WHOOP going to a “push to all” rather than a “pull” based sales motion. If it’s foisted upon people, it will not produce the desired effect.

4. In federal sales, relationships and strategic partnerships beat conference booths.

Todd assesses that in the defense market, niche events where you can have real conversations, and partnerships with athlete management system providers who can pull WHOOP into their stack, have been the most valuable sources of leads. Trade show booths? Much less effective. As always, your mileage may vary, but it’s another reinforcing data point suggesting deep, niche relationships beat a larger breadth with high impressions.

5. Sell sustainable programs, not hardware deployments.

Todd’s north star for WHOOP’s next 12 months is helping customers build programs that will still be running in five years. That means bringing in WHOOP’s human performance experts and solution architects early, co-designing integration with existing data platforms, and being the vendor who asks, “What are you actually trying to achieve?” before taking the purchase order.

For more Crossing the Valley: valleycrossers.com | LinkedIn

For more on WHOOP: WHOOP.com

Follow Todd: https://www.linkedin.com/in/toddstiefler/

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

Byron Boots is the co-founder and CEO of Overland AI. Prior to starting the company, he was an associate professor in computer science at the University of Washington, where his research focused on machine learning and robot autonomy. Byron served as principal investigator on DARPA’s Racer program, one of the most ambitious off-road ground autonomy challenges the agency has run since its landmark Grand Challenge series in 2004–2005. It was in that role — out in the field, iterating under adversarial test conditions — that the insight and technology behind Overland AI took shape.

About Overland AI

Overland AI builds autonomous ground vehicle software and systems for the U.S. military. Their flagship platform, ULTRA, is a purpose-built unmanned ground vehicle with upgraded suspension, miniaturized compute, and a suite of sensors including stereo cameras, LiDAR, thermal cameras, and Starlink — capable of GPS-denied navigation in extreme off-road terrain.

The company operates across four primary concept of operations: ISR (intelligence, surveillance, and reconnaissance), logistics and resupply, layered protection (including counter-UAS), and breaching. Their go-to-market approach has been almost entirely driven by demonstrated field performance. Initially this came through DARPA experiments, then through DIU and direct work with Army and Marine Corps units. Overland has won contracts including the DIU Ground Vehicle Autonomous Pathways program and is currently focused on scaling into formations and establishing a permanent operational presence with the services.

Key Takeaways

Five principles from Byron’s journey that apply broadly to defense tech founders:

1. Field, first. One of the most counterintuitive decisions Byron made at University of Washington was to hire strong vehicle mechanics as some of his first team members. It’s a simple but counterintuitive logic. You don’t have to build the perfect software in the lab before you put it on the vehicle because if vehicles could be repaired rapidly in the field, the software team could test more aggressively, fail faster, and iterate without fear. Thus, the vehicle mechanics gave the rest of the team leverage.

2. The best BD is performance (with the lights on). Overland AI’s early business development was almost entirely organic. I pressed him on DARPA’s checkered history of transitioning programs, but in this case, DARPA invited the other services to watch their field experiments. Those experiments were credible, because the company often didn’t know what was coming. It wasn’t a PowerPoints brief or a table top conversation; it was an opportunity for customers to see the tech in action. That credibility dramatically accelerated their sales path.

3. Build on top of what works. Byron’s framework for capability development is deliberate: get one robot working reliably, then add a second, then build multi-vehicle coordination. Don’t architect for the end state. Prove each layer before adding the next. This mirrors successful product development in commercial software but is even more important in defense, where integration failures in the field are immediately visible and costly.

4. “If it’s not seamless, it won’t survive contact with the warfighter.” Despite the old saying “good enough for government work,” the operational bar for defense products is actually quite amorphous. Byron observed that “a slow or unreliable robot simply gets abandoned.” In other words, your project might die without you even knowing. The product actually has to perform well enough that operators don’t think about it. Only then can you expand the mission set, add payloads, or introduce multi-vehicle coordination.

5. It’s not about your tech. The sooner you embrace this, the faster you’ll win.

   Winning a contract gets you in the door, but a contract isn’t the same thing as winning warfighter trust. Byron draws a clear distinction between them. Overland's next 12 months aren't going to be defined by new contracts as much as they are going to be defined by creating "a permanent presence" with Army and Marine Corps units. I love this point because it suggests a totally different motion. It’s not about closing the deal or renewing it. It’s not about another slide or another meeting. It’s about doing things that don’t scale, having people side by side with the mission owners, and obsessing over the question of how to make your technology invisible inside existing workflows rather than celebrated as a novelty.

For more Crossing the Valley: valleycrossers.com

For more on Overland AI: www.overland.ai

Follow Byron: https://www.linkedin.com/in/byron-boots/

Follow Noah: https://www.linkedin.com/in/noahsheinbaum/

In this episode, Forterra CEO Josh Araujo is joined by Ari Schuler, a returning CTV guest, to take us inside Forterra’s acquisition of goTenna.

Company Overview

About Forterra: Ground autonomy company enabling war fighters to shoot, move, and communicate more effectively. Technical stack includes autonomous driving systems, mission payload integration, and now mesh communications through the goTenna acquisition. Working with military customers across operational theaters. Founded 20+ years ago (as Robotic Research), recently closed Series C while simultaneously acquiring goTenna.

About goTenna: Mesh networking communications company providing interoperable comms, born from Hurricane Sandy in Brooklyn. Pivoted from consumer to military/public safety after discovering product-market fit with operators in Afghanistan, Iraq, and the southern border. Technology enables high node-count networks (100+ devices) sending low-bandwidth position/location data, critical for blue force tracking and preventing friendly fire.

About our Guests

About Josh Araujo: CEO of Forterra, building autonomous ground vehicle technology for military operations. Retired Marine, and Jefferies investment banker covering aerospace and defense. Joined Forterra (then Robotic Research) in 2021 as the first non-technical hire, now leading a company with 20+ years of autonomy innovation transitioning to programs of record.

About Ari Schuler: Former CEO of goTenna, now leading the communications division at Forterra. Built goTenna’s pivot from consumer to B2G, deploying mesh networking technology in conflict zones globally. Former DHS official who created Customs and Border Protection (CBP)’s innovation team, bringing government operator perspective to commercial tech development.

Key Takeaways

1. Skip the vitamin, find the medicine.

Ari’s mentor gave him a deceptively simple framework: are you selling something people want (vitamins), or something they can’t live without (medicine)? goTenna started as a consumer product, but product-market fit became unmistakable once the technology hit Afghanistan, Iraq, and the southern border. Ari explains that when you found the true medicinal use case, you go all in. Straddling consumer and government markets wastes capital and credibility that companies don’t have to spare.

2. The ability to sustain operations in the field is the real moat.

Both Forterra and goTenna had extraordinary technical pedigrees: Forterra’s co-founder holds 100+ patents, and goTenna’s mesh protocols are proprietary and deeply differentiated. But Josh was explicit: deploying at scale into combat environments is “miles apart” from having a cool demo. The muscle required to do customer success, field maintenance, and operational support in places like Afghanistan or Ukraine is not taught in any graduate program.

3. In M&A, cultural alignment is what ultimately determines deal success or failure.

The core of the conversation centers on how Josh turned a term sheet around in 72 hours, from a Friday morning coffee between Ari and Scott Sanders (another CTV alum), to an afternoon board meeting, to a signed term sheet by Sunday night. This kind of speed is not only a function of high caliber financial modeling and expertise, but of a clear strategic rationale and cultural fit. Josh talked about his view that even the best business alignment in the world can’t overcome a cultural mismatch. Culture doesn’t mean squishy values in this case. It refers to the companies’ shared mission focus, values, and tech stack (NetSuite, Arena, JIRA).

4. The defense VC funding model is creating a structural problem for which consolidation is the only solution.

Josh flagged a concern that most defense tech founders are quietly aware of: many of the best companies in this space have great technology, great teams, and genuine product-market fit, but lack venture-scale TAMs. VC math requires one in a hundred to go 100x. At the risk of sounding trite, that calculus alone will not support the entire ecosystem of mission-critical hardware businesses. As a result, if venture is the primary tool, companies that shouldn’t fail will fail because of capital structure problems. This is why Josh is shaping Forterra into a leading consolidator to create a home for these companies.

5. In a post-merger integration, the first priority is “do not break what’s working for the customer.”

Too many post-merger integration efforts try to move fast, combine everything, and extract synergies. In this deal, Josh and Forterra explicitly rejected this. The approach has been phased and deliberate: first, validate the technology combination (goTenna’s X2m module was physically integrated into Forterra’s Vector platform within 30 days, achieving 25x the range of the previous radio). Second, resource the acquired team to execute on their existing pipeline. Third, after the teams have developed organic trust, start to combine go-to-market and engineering functions. This cadence matters because defense vendors cannot afford to push artificial disruptions on their customers. As Ari explains, “We can figure out the HR systems. What we can’t do is mess up operational deployments.” The customer relationship is paramount.

For more:

• Website: Forterra | goTenna

• LinkedIn: Forterra | goTenna

• Follow Josh: https://www.linkedin.com/in/josh-araujo/

• Follow Ari: https://www.linkedin.com/in/ari-schuler-1ab93239/

• Follow Crossing the Valley: https://www.linkedin.com/company/102082197/