By Dawn Zoldi
Stephen Winchell doesn’t run a typical government office. As the 24th director of the Defense Advanced Research Projects Agency (DARPA), he leads an organization that doesn’t just tolerate, but expects and even celebrates it. Speaking as a keynote at the 2026 GEOINT Symposium, Winchell openly invited anyone with an idea too wild for anyone else to touch to bring it to DARPA. That spirit of bold yet structured risk-taking has defined DARPA since its founding. Winchell’s keynote offered a rare inside look at how one of the most consequential research agencies in the world actually operates, and how industry can tap into it.
Discoverer to GEOINT: How DARPA Built the Foundations
Winchell traced DARPA’s role in the birth of geospatial intelligence (GEOINT). The Corona program, publicly known as Discoverer, was a secret CIA and U.S. Air Force initiative that DARPA ran starting in the late 1950s. The first thirteen flights either carried no payloads or failed outright. Discoverer 14 launched on August 18, 1960, and photographed the Soviet Union with a clarity that five years of dangerous U-2 overflight operations had never achieved. The program ultimately produced more than 860,000 images covering 750 million square miles.
“These images are unbelievable,” Winchell said. “They’ve helped humanity as well as the U.S. government.” Beyond intelligence gathering, the images revealed ancient civilizations buried beneath jungle canopy.
He drew a direct line from Corona to his own career. After leaving active duty in the U.S. Navy where he had been a submarine officer, he joined Johns Hopkins Applied Physics Laboratory, where his supervisor had worked on the Corona program. Years later, Winchell served as chief engineer for Project Maven, the Pentagon’s Algorithmic Warfare Cross-Functional Team, during what he described as the early days of deep neural networks. His team built data pipelines, conditioning workflows and assurance capabilities for object detection and classification from full-motion video (FMV), synthetic aperture radar (SAR) and electro-optical (EO) infrared imagery.
The Space Window America Cannot Afford to Miss
Winchell walked the audience through a stark economic shift relating to the space economy. In 1980, placing one kilogram of mass into low Earth orbit (LEO) cost $85,000. The Falcon Heavy brought that down to approximately $1,400. With Starship and New Glenn coming online, Winchell projects the cost will drop another “order of magnitude,” to somewhere between $100 and $400 per kilogram.
He believes the strategic implications will prove enormous. Last year, the United States launched 2,296 tons of mass into orbit. China launched 226. That near-10-to-1 advantage will widen, and Winchell believes the United States has a narrow window to convert that lead into lasting warfighting and economic advantages.
“We have a very special moment in (the space) industry,” he said. “Private industry is going to be in the driver’s seat. The government is no longer going to dominate the design, development and testing of these capabilities on orbit.”
In response, DARPA has launched a three-pillar space strategy. The first maps out technology trees and key dependencies, from thermal management and propulsion to on-orbit construction and positioning, navigation, and timing (PNT). The second builds new analytical tools to support force design, maneuver planning and unexpected contingencies. The third puts funding directly into the hands of researchers through open solicitations aimed at companies, universities and individual innovators.
DARPA’s space-specific programs are genuinely unlike anything in the mainstream defense conversation. Atmos Sense uses disturbances in the ionosphere’s electron fields to detect hypersonic reentries and explosions around the globe. A robotic servicing satellite called RSGS, equipped with manipulator arms capable of refueling and maintenance, is set to launch later this year. DARPA is also conducting a study to evaluate whether the agency should capture and de-tumble an asteroid, then hand it off to private industry to kickstart space resource extraction. Here is where weird meets reality.
The Scientific Method on Steroids
Winchell also explained how DARPA decides what to fund. He cited a framework established by DARPA’s sixth director that continues to animate the agency’s program evaluation process. Every proposal must answer four questions:
- What exactly are you trying to do (with no jargon)?
- How is it done today?
- What is your core technical insight that makes your approach better?
- And how will you hold yourself accountable (what are the metrics that prove or disprove the hypothesis?)
“It’s the scientific method on steroids,” Winchell explained. “Moving at speed and focused on the warfighter. That’s it.”
What additionally separates DARPA from a traditional research funder is how deeply leadership engages with each program. Leaders scrutinize whether the project is moving fast enough, whether it is risky enough and whether the risk is the right kind. Winchell described the agency’s program management culture as giving each effort a “loving and intimate and involved hands-on touch” from the entire leadership team.
You Are Going to Fail and That’s the Point
Failure at DARPA doesn’t signal a poorly designed program. It is, in Winchell’s assessment, the mechanism through which the agency generates knowledge that no other process can produce. He pointed to two examples:
- A multi-decade effort to develop synthetic whole-blood substitutes. The team worked through freeze-drying, freezing and preservation challenges. They repeatedly hit walls, but each failure exposed new science. The cumulative work eventually cracked open a path toward lipid-shelled synthetic hemoglobin and, potentially, shelf-stable blood for battlefield use.
- An optical atomic clock development that required four successive attempts before it matured enough to transition to the private sector and eventually be acquired by a publicly traded company.
“You don’t know how you’re going to fail,” Winchell noted. “You don’t know what it’s going to mean and what you’re going to learn. Having a boss who is okay with that is a really good start. Having a team that embraces it and celebrates it is huge.”
The artificial intelligence (AI) portfolio follows a similar trajectory. In 2018, DARPA made a $2 billion bet on AI. It launched 130 programs that Winchell credited with laying the groundwork for the capabilities that now reshape commercial and defense sectors. Today, DARPA is not trying to compete with frontier model companies. Instead, it is working on the edges of what commercial AI cannot or will not do: achieving thousand-fold speed improvements, cutting power consumption by orders of magnitude, applying formal methods to verify model behavior and convening industry and academia around core national security challenges. Three weeks before GEOINT, Winchell approved a new program focused on multi-agent consensus-making and security in advanced AI systems. He also oversees an internal experiment that pits a DARPA program manager’s human judgment against a team of experts using AI to accomplish the same tasks, faster.
It Depends: The Honest Answer on Transition
So what is the biggest barrier to moving DARPA programs into operational use? Winchell said, “It depends.” Every program, every program manager and every acquisition partner brings its own constraints. That said, he identified one structural chokepoint he believes offers the most leverage, the requirements generation and acquisition process inside the Department of Defense (DoD).
“If you can create a fair test of requirements that actually matter to the warfighter, there’s space for groundbreaking technology to truly displace what exists,” he said. “A lot of folks don’t want to set up fair tests because they don’t want things to be displaced.”
As an example of progress, he pointed to a program that transitioned at “negative three months” as his gold standard. In that case, another government agency started executing on a DARPA concept before the formal program even launched. The cost to DARPA was zero dollars.
DARPA’s Open Door for Industry
Winchell closed with a direct pitch to industry. The agency’s roughly $4 billion annual portfolio distributes resources across three roughly equal segments: small businesses and non-traditional contractors, academia and larger systems-engineering and classified program development organizations. The SBIR and STTR pipeline alone exceeds $190 million annually.
POP OUT QUOTE
“We are doing crazy stuff. If you’ve got a crazy idea, if you’ve got a vision of something that no one else wants to go after but you’re right, we are the right partner for you.” – Stephen Winchell, DARPA Director
The profile of a DARPA program manager, he said, is someone who is technically deep, driven, and obsessed, but above all, carrying a core insight they know is right even when no one else believes in it. That’s the DNA of every successful DARPA program in its history.
“Find the thing you’re really excellent at. Pour yourself into it. And once you have that nugget that no one else will believe in, bring it over to us,” he encouraged.
For an agency that helped invent the internet, GPS and stealth aircraft, the invitation is worth taking seriously. Learn more about DARPA at https://www.darpa.mil/.