Building Replicator: What it Took, and What it Meant
Nobody in Washington gets promoted for grading their own homework in public.
Dr. Kathleen Hicks and Aaron Sherman did it anyway. Three years after Replicator launched, they sat down and wrote an insiders’ history of the initiative. It’s called Move Fast and Scale, published by the Belfer Center, and it landed just at a moment when it seems that the whole building is racing to run some version of the same play (albeit with a lot more money behind it).
I’ve been wanting to have this conversation for a while, in part because of my own exposure to successors of Replicator, in the form of the Drone Dominance Program and Project Swarm Forge, two challenge-based acquisition initiatives focused on small, one-way attack drones. So I opened by asking Dr. Hicks for the single most important takeaway she’d pass along to someone trying to replicate Replicator. She didn’t talk authorities, and she didn’t mention contract vehicles.
She said:
“I think it’s for senior leaders, whatever that means in the particular context, to be as fascinated by the execution as they are by the rollout.”
Or put another way, “If your boss is interested in something, you should be fascinated by it.”
About Kath Hicks and Aaron Sherman
Dr. Kathleen Hicks was the 35th Deputy Secretary of Defense (aka the person who actually ran the Pentagon day to day). Before that, seven years at CSIS, and a PhD from MIT. She has been in and around the Pentagon long enough to have watched this kind of fight go south before.
Aaron Sherman was her speechwriter. He was inside the small group that built and launched Replicator, which means he wrote a lot of the words that the building, the Hill, industry and Beijing all read at the same time. That is an unusual seat, and the holder of the pen is also the first author of the policy.
About Replicator
In late August 2023, the Department of Defense launched what it called a “pathfinder initiative” to rapidly find, acquire and field new technologies. The target was multiple thousands of attritable autonomous systems, across multiple domains. The goal was to get them in warfighters’ hands inside 18 to 24 months. At the moment they announced the initiative, they named no systems. They identified no immediate “on-ramp” for companies. And they didn’t give Congress much of a heads-up. But they got a TON of attention.
The first tranche included AeroVironment’s Switchblade 600. Upon leaving DIU, T.S. Allen, the former Director of Replicator-1 said in August 2025, “we were able to deliver…hundreds of drones to war fighters” and “put thousands more on contract.”
But one thing both make clear is that drones were never the fully story.
Hicks describes Replicator as a pathfinder: the idea was to pick one capability area, drag it through every blocker in the system inside a two-year appropriation cycle, and leave the road wider for everything behind it. Authority to operate. Clearances and SCIF space for new entrants. Cyber hardening for small companies who become PRC targets the day they get named in public. Every one of those blockers had been quietly taxing non-traditional contractors for a decade. The only times they got solved were when a leader couldn’t unlock a program they cared about.
Key Takeaways
1. Not every valley is supposed to be crossed
This show is named after the valley of death, so I asked about the mapping Heidi Shyu’s team did showing several of them instead of one.
Dr. Hicks’ answer was that these valleys are not inherently bad.
“We need a big funnel of ideas and innovation,” she said. Things are supposed to fall out. Failure occurs only when we lose the good ones.
That reorientation turns a vague complaint into a targeting problem. Aaron says the earlier efforts in the Biden administration had already gone after earlier handoffs, research to prototype, prototype to production style transitions. Replicator deliberately went after the last one, the jump from “tens and hundreds” to thousands.
2. The clock determined the capability
“The major framework conditions for us were how to make change within the two years of an appropriation cycle, because that’s where the country is really stuck.”
Two years was not a goal, but a box. Once the team accepted the box, exquisite capabilities were off the table. By process of elimination, the capability area more or less picked itself. Attritable autonomy was one of the few places you could put something real in an operator’s hands, improve the domestic supply chain, and show a result before the budget turned over.
Most of the time acquisition works the other direction: pick a technology, or a capability area, and lay out the time it needs. Hicks started from a timetable, and worked backward to pilot a capability that could fit.
3. You cannot always change the people, but you can change incentives
DepSecDef Hicks signed more acquisition reform memos than almost anyone alive, so I appreciated her transparency about their worth:
“We’ve had dozens literally of acquisition reform reports and statutory changes. I’m not saying they’re meaningless. I’ve signed many, many, many of those myself. But at the end of the day, you really do have to incent change by elevating champions.”
Her method was to find the people already dying to make a difference, show them a visible path to success, and then actually follow through on who advances, who gets more mission, and who gets more budget. That pattern of recognition and promotion actually does make it a difference.
Aaron called it a heist movie. A team assembling, somebody in each corner spinning a lock. The Vice Chairman, Admiral Chris Grady, took requirements off the table every single time. Solved, next. Solved, next.
Hicks’ trademark line was: “have pen, will sign.” You have a barrier and you’re telling me about it? Let’s fix it. If it’s not us, who’s going to fix it?
The image presented is a compelling example of top cover. It shows the GS-14 that raising a hand is not a career-ending move.
4. “Announce it before the building can water it down”
The Replicator launch attracted a lot of attention, and courted some controversy. I asked the team whether they’d run it back, given what they know now.
Hick said she would.
“The risks are much higher of having what you seek watered down. I have not only seen that, but I have lived it and been a part of it since the end of the Cold War. So do you create some expectations that have to be met? Do you have some questions you can’t answer right away? Sure. But in this context of major culture change, I’d rather moonshot.”
Hicks watched that happen with Ash Carter as Deputy Secretary and Bob Work in the building. Two of the strongest advocates for disruptive innovation, and autonomous systems that the Department ever produced, and culture still ate much of those efforts.
So Hicks resolved, “We’re not gonna let the culture kill it. We’re gonna change the culture.”
5. You’re always deferring a bill. Know which one it is, and prepare to pay for it
Aaron and Kath knew they short-changed congressional engagement before the announcement. They were up against the clock of a fixed event. They knew it would come with tradeoffs.
“The trust between the department and the legislative branch is crucial to getting things done and to moving fast,” Aaron said, “We knew going in that we [would have to] do the congressional engagement after the fact.”
Dozens of briefings followed. In SCIFs, on the Hill, at various level. There was some belly-aching about not being notified, but ultimately, the Hill did get on board.
This was still a good news story: speed came at a cost, and trust had to be re-earned. There’s a difference between deferring a payment and skipping it entirely, though. Even moonshots have to pay the bills!
What does it all mean?
A few closing reflections…
The “I’d rather moonshot” idea can work when you’re the deputy secretary, have the secretary and the NSC lined up behind you, and the resources to boot. A program manager three levels down who announces before coordinating is far more likely to face discipline than culture change.
Second, I think it’s fair to call the motion a success in terms of catalyzing speedier action, backed up by performance rather than performative marketing. However, I’m also a big believer the question ought always to be, “what could have been better” or “how could we have made a bigger dent, faster.”
My personal view (and part of why I am as involved with Drone Dominance as I am) is that we can always go faster. The commercially-derived nature of many of these autonomous systems means that usual OpSec requirements, while important, are increasingly dwarfed by the need to build exceptionally fleet-footed industry partners capable of continuous iteration on hardware and software, alike. The best way to do that is to balance two priorities: 1) keeping the top of the funnel open, so newcomers can come in and push the incumbents to try new approaches, faster. 2) Rewarding the winners with sizable contracts that justify their up front investments. I LOVE the fact that the current incarnation is showing what’s going on in the test range, publishing a leaderboard, and making it clear to companies how to play and how to win. And still, there is more to be done.
If we’re saying the capabilities are going to evolve so fast our system has to evolve to keep pace, I’m less worried about telegraphing our secret sauce to Beijing (in this domain). The structure and system is the thing, more so than any individual tech platform.
Thanks for reading - and hope you get something useful out of it.
Read the Report: https://www.belfercenter.org/research-analysis/move-fast-and-scale-brief-insiders-history-replicator-initiative
Follow Crossing the Valley: www.valleycrossers.com
Follow Dr. Kath Hicks: https://www.linkedin.com/in/kathleen-h-hicks/
Follow Aaron Sherman: https://www.linkedin.com/in/aaron-sherman-04017817/
Follow Noah Sheinbaum: https://www.linkedin.com/in/noahsheinbaum/
