Innovation machine
By Cat Hofacker|March 2024
Ken Plaks, director of DARPA’s Tactical Technology Office
At most jobs, employees are judged by their successes. But at DARPA, the Pentagon’s technology incubator, “I actually get yelled at if my programs don’t fail enough,” says Ken Plaks, director of the agency’s Tactical Technology Office. If the agency’s programs succeed too often, the thinking goes, that means the challenges aren’t hard enough. “It’s a high bar,” he says, considering that the agency created the networking concept that became the public internet. So Plaks has a sizable list of potentially groundbreaking innovations that are receiving funds: The current portfolio includes a joint undertaking with NASA to demonstrate a nuclear thermal rocket and an effort to build a drastically cheaper seaplane the size of a C-130 for cargo flights in the Pacific. I visited Plaks at DARPA Headquarters in Northern Virginia to discuss these programs and how the agency stays at the cutting edge of technology. Here are lightly edited excerpts of our discussion, organized by topics.
On the ‘done’ pile
From a technology perspective, certainly space has moved from mostly governments to mostly commercial. That’s happened really only in the last couple of years, and DARPA kind of saw some of that coming. We had a program to do megaconstellations before megaconstellations, were a thing. And then Starlink passed us, right, which is fine. So we did our DARPA thing and said, “OK, cool. Put that in the done pile,” and we’re gracefully shutting down the program that we were trying to tease out.
He’s referring to Blackjack, an “architecture demonstration” in which DARPA has been operating a constellation of small commercially built satellites to determine whether they could be a lower-cost alternative to geosynchronous spacecraft. Plans had called for 20 satellites, but DARPA decided to conduct the demo with just the first four satellites, launched in June by a SpaceX Falcon 9. — CH
And it influenced the NRO [National Reconnaissance Office] and Space Force’s future architectures. Before, the idea was “We have to have the Battlestar Galactica satellite that does everything.” Now, we’re disaggregating into a bunch of small satellites that work together. It has been such a change that it went from “That’s impossible,” even as recently as five or six years ago, to “Of course, that’s how you do it.” And a lot of that’s just with how cheap it’s gotten to get to LEO in particular. On the engine side, we’re getting back into hypersonics, but that’s kind of back to the future. Most cargo or transport airplanes are still tubes with wings; we’ve talked about flying wings and stuff, but that really hasn’t happened. Our X-65 CRANE program has the potential in future generations to change how we do aircraft.
The CRANE demonstrator, short for Control of Revolutionary Aircraft with Novel Effectors, is being built by Boeing subsidiary Aurora Flight Sciences to test active flow control, in which flight control would be maintained via jets of pressurized air instead of external control surfaces. Plans call for remotely piloted test flights to begin in 2025. — CH
On the drone side, small, unmanned things are now doing missions that we never even contemplated — except, there were a couple of DARPA programs for the last couple of years that were looking at “How would I use small drones to go after armor and to go after an integrated air defense system and take it down?” So it’s interesting that if you look back four or five years, you can see the threads of the things that are actually emerging are things from DARPA programs.
Personal interests
A couple of things I’m noodling on and trying to push: Probably the biggest is how we deal with complexity and cost. We can build these really exquisite systems, but if I only have 10 of them against thousands, you know, eventually that breaks down. If you look at pLEO — proliferated low-Earth orbit — that is largely a way of dealing with complexity by doing simpler things. Pick one thing and be good at it and let someone else handle the other part, rather than putting it all together. Because as I start to get into all the size, weight, power-constrained volumes, one plus one frequently doesn’t equal two; it’s four, because there are all these interactions between the parts in nonobvious ways. The way we deal with that is systems engineering, which is necessary but kind of not sufficient. There’s all these stories of unintended consequences and problems — “We put the brake computer in the avionics computer, and now when your avionics crashes, you don’t have brakes,” or whatever. If it’s simpler, then I can buy more of them, and it’s cheaper and I can do it faster. And you get in this virtuous cycle. We also need to do a better job of dealing with the complexity. At some point there’s a finite limit, and you can only make it so simple, right? So if I’ve got to have a system that slices, dices and makes julienne fries, for whatever reason, then how do I deal with all that complexity? How do I manage it? I don’t want to sprinkle AI pixie dust on this, but it is kind of a big data problem. The ability to bring thousands and thousands of cores online to work on a problem without having to actually own thousands, for instance, may give you a computational way out of the wilderness.
Defining ‘DARPA hard’
The first big thing is it’s got to have an outsized impact. When they hire you, the director shakes your hand and says, “Go change the world.” Stealth is a great example. The Air Force actually didn’t want stealth at first; they just wanted bigger jammers. The joke at DARPA is if you don’t invent the internet, you get a B. It’s a high bar, and we’re all A-plus students. The other big characteristic is our saying that there’s hard, and then there’s DARPA hard. When I first came here — I had been a colonel in the Air Force, and I had been in the acquisition community — I was pitching my first project, and I was like, “Well, this is low risk,” to illustrate that we could do this. And my boss said, “Well, if you’re sure you can do it, why are we doing it? The Air Force can do that.” You know, it’s not hard enough. I actually get yelled at if my programs don’t fail enough — which I’m pretty sure is not common in the rest of the government — the idea being that if I’m achieving my goals routinely, then I’m not setting hard enough goals. We run this whole building like “Shark Tank.” You go upstairs, and you get an hour to tell the big boss how you want to change the world with $50 million to build an airplane that has active flow controls or whatever. If she likes it, she writes the check, and then you go do your thing. Frequently, the result of the “Shark Tank” is homework — “Go make it bigger impact” or “You said you can make it twice as good; I want it 10 times as good.” And we even say in our solicitations, I’m not interested in things that are evolutionary. It’s got to be revolutionary. The other interesting thing about DARPA is that everybody’s temporary. If you’re doing a great job, after four years, they say, “Have a nice life.” And unlike the rest of the federal government, there’s really not any empires. We don’t have any X amount of money set aside for AI. The NSF [National Science Foundation] does that, partly so you can have the continuity with the grad students. But DARPA can turn on a dime. It increases our agility, and it makes it a little idiosyncratic. A quarter of our workforce changes every year. After two years, DARPA is almost a completely different place, and the things we’re working on are almost completely different. The other big thing is all our programs are finite — typically three or four years, then it’s over. We only do three- or four-year bites at the apple.
Handing off the tech
It’s easier to get it into the commercial side than the military, because they’re on two-year budget cycles. So even if it’s the best thing, a service can’t even begin to buy it for another two years. Whereas the commercial world can actually move quicker. It is a continual challenge. The DARPA mission is to prevent strategic surprise. How do you prevent surprise? Our solution is to make our own surprises and figure out what’s coming next. That way, the services or the government may make a decision not to invest in hypersonics or whatever, but at least they know it’s a possibility or that it’s not a possibility and we don’t need to worry about this. So while it would be tempting to go, “We have this fantastic missile and for another $100 million, we could get it to the point where it’s ready for the Air Force to buy,” that’s $100 million I’m not spending on inventing an internet, or whatever else. So it’s a balance. We do sometimes work with the services when they’re really interested to go halfsies on the funding and we’ll do an extra phase. We had a program here called the LRASM, Long Range Anti-Ship Missile, that the Navy was really interested in, but they just couldn’t get started on their own. So DARPA actually started the acquisition, and now the Navy is buying them, but they had some skin in the game too. Another DARPA saying is that transition is a full contact sport, and it’s super hard. We recently stood up a unit in one of our offices whose whole job is to help people transition [out of DARPA to the military or commercial applications] better. And again, part of that is understanding the landscape in the services. Because a lot of DARPA PMs [program managers] are from the military, they’re university professors and industry, so some of that is just connecting. There’s also a separate transition problem: Frequently, some university professor starts a company, that company performs on our contract and does a great job, has made a great piece of technology. What we were seeing was the Chinese were then coming in with a ton of venture capital — and sometimes it’s not even obvious that it’s a Chinese company or Chinese sponsorship. So we actually stood up a bunch of American venture capitalists.
He’s referring to DARPA’s Embedded Entrepreneurship Initiative, established in 2019. — CH
Having the DARPA stamp of approval is a pretty good thing for a startup, so we were doing a little bit of matchmaking of getting the venture folks together with the startups that are coming off a DARPA contract to try to get them capital without foreign entanglements. That’s a bit far from defense, but it’s critically important to the health of the defense industrial base, right?
Lessons from Ukraine
What you’re seeing is that anti-access/area denial really works good on both sides. Anything bigger than a quadcopter pretty much dies; they shoot down 30, 40 missiles and drones a night. If those were manned aircraft, those would be like World War II numbers. Part of managing complexity is I can do things faster and do it quicker, and you’re seeing the Ukrainians adapting very rapidly, and the Russians to a certain extent too. We have programs that are looking at how we do that; we have a program called REMA [Rapid Experimental Missionized Autonomy] that is looking at how to rapidly adapt a drone for a different mission. How do I rapidly add autonomy and other things to a drone at the speed of relevance, like a couple months. We actually turned the engines of innovation on ourselves. It was 70 business days from the “Shark Tank” pitch to on-contract performing, which is a land-speed record in the federal government and certainly fast by DARPA standards. Also, the United States as a nation is pretty good at stealth, and while we have things that we’re looking at doing better — and I’m not really going to go into that here — stealth only buys you so much speed, maneuverability and autonomy. It all really comes back to the complexity. Nobody says “I’d really like to go slow and take my time and do something in three years that I could do in a month,” but we’re driven there, by and large, by this vicious circle: It’s complicated, therefore it takes longer, which means I got to think about it more before I get started, which means it takes longer, which slows me down. So you’re trapped and it’s the F-35; you’re doing first flight in the ’90s and we’re still working on it. And the F-35 is a great airplane; I’m not throwing rocks at it; it’s super, super complicated. What I want to do is to the extent you can disaggregate, knock the complexity down by making simpler things great. And then I want to overtly design things so that they can be produced. If you told me I could go from making 20 F-35s a year to making 200 F-35s a year by a 5% decrease in the signature, that might be a trade that’s worth making. Those are all made up numbers, but nobody is actually talking that way.
Inspired by boat building
For our Liberty Lifter program, I have a naval architect building an airplane. A C-130’s cost per pound is something like $550. You buy airplanes by the pound. We’re looking at being $200-something per pound. It turns out there’s a couple ways you could cut down $300. One is we’re making a wing-in ground effect airplane — so very heavy lift, but it can still get up and away to 10,000 feet. That gets you over maritime obstacles, small islands and stuff like that. That also means I don’t have to pressurize the cabin, which implies a certain level of fit and finish on a cargo airplane. So a C-130-sized airplane, that gets expensive. For Liberty Lifter, we’re probably using marine grade composites — they’re still in the trades phase — which are significantly cheaper than aviation composites. We’re looking at using marine aluminum, which you can weld, versus aircraft aluminum that you have to rivet, which is a lot of touch labor. It’s like 50 times more per pound just to use. The other thing is I want to be able to build large portions of this seaplane at boat yards — not like Huntington Ingalls, but Billy Bob’s Boat Emporium. I want him to be able to make the wingbox because it’s just aluminum and steel and maritime composites that he’s used to using. It’s a different shape, but he’s got all the tooling and everything he needs. As we look at going to the Pacific, the Army and the Air Force and the Marines are talking about hiding on islands and playing a shell game. And that’s great, except if everything’s flying into a runway or a port, if I blow that up, you can’t get supplies and you’re going to be combat ineffective. But if I could bring in a C-130-sized seaplane or maybe even a C-17 to land anywhere on the coast or in the water, that changes the terms of the engagement. It’s the transportation factor: how efficiently I can transport stuff, comparable to a ship but speeds comparable to an airplane. When you airmail or FedEx something, it’s really fast, but it’s also really inefficient in terms of price per pound. When I send something on a cargo ship, it’s super efficient, but it would take a month to send a letter to Asia. Liberty Lifter is kind of in the middle, so it’s almost as fast as a jetliner, but not quite, and it’s almost as efficient as a cargo ship, but not quite. I don’t know that there would be necessarily a civilian market for this, but from a military perspective, it solves a lot of problems in an innovative way — if we can build it.
The right time for a nuclear-thermal rocket demonstration
Space from a military perspective is more contested; it used to be kind of rainbows and unicorns out there. We are not expecting that to be quite the same in the future, particularly in cislunar beyond geosynchronous orbit. There’s commercial interest in a way that there wasn’t previously, and NASA, our partner on DRACO [Demonstration Rocket for Agile Cislunar Operations], wants to go to Mars. This is similar to the seaplane: Right now, your choices are chemical rockets — which give tens of thousands of pounds of thrust but are inefficient — and you’ve got ion drives that are super efficient and give much less thrust. DRACO is in that middle region where I get thrust like a chemical rocket but two to three times more efficiently. And there’s a couple different flavors of nuclear propulsion. In the long term, if I’ve got a nuclear reactor up there, I think you want nuclear thermal, which gives you the magnitude of the thrust. For station-keeping, I’d rather use ions and be super efficient. And so it’s not an either; it’s a yes, both. And once I’ve got the nuclear reactor, I can decide how to use that power —an open loop as a thruster, or closed loop as an electricity source for an ion thrust. I think you would want both for a real mission. We know how to build closed-loop nuclear reactors; we’ve been doing that for years. Moving into space, there’s all sorts of challenges. We’re focusing on the nuclear thermal side of it, which is, to a certain extent, the thing with more uncertainty. There’s nowhere on Earth where you can test those things, and so we’re never turning the reactor on till we get into space. That solves the problem of “What if the rocket blows up?” Well, OK, fine; we’re using high-assay, low-enriched uranium. It’s not an innate radiation hazard in the way that the highly enriched, weapons-grade stuff is. So even if the rocket blew up on the pad, we’re not going to have like a Chernobyl-type disaster.
