Summary
Autonomous systems are no longer emerging. They are already shaping how modern defense operates. The challenge now is not innovation, but how these systems scale under real world constraints, including navigation, communication, and production. As the margin for error narrows, the ability to accurately perceive the full environment becomes foundational to performance and decision making.
This is Part One of a three part series by Jennifer Sertl
On March 26, leaders across defense, robotics, and venture gathered in New York City at the offices of Pillsbury for the annual Drones and Robotics event hosted by GENIUS NY and ff Venture Capital.
This marked Circle Optics’ fourth year participating in the event. That continuity matters. It reflects the role this convening plays in bringing together founders, operators, investors, and policy leaders who are actively shaping how autonomous systems are built and deployed.
Special recognition is due to Kara Jones for her continued leadership in building the unmanned systems ecosystem and to Oliver Mitchell for his role in connecting early-stage innovation with capital and strategic support.
The panel, moderated by David Skinner of the Griffiss Institute, brought together Ben Levinson, Founder and CEO of Heaven Aerotech, Ryan, Co founder of Root Access, and David Franco, Co founder and CEO of Wild West Systems.
What emerged from the discussion was not speculation about the future.
It was a clear picture of what is already happening and what is not yet working.
The Shift Has Already Happened
The panel made one point unmistakably clear. The transition to machine driven warfare is not ahead of us. It is already underway.
David Franco framed it directly when he said there will be far more machines than humans on the battlefield and that the next arms race is robotics . That statement is not theoretical. It reflects what is already visible in current conflicts where low cost drones are responsible for a significant portion of battlefield outcomes.
At the same time, Franco challenged one of the dominant patterns in today’s deployment models. Many systems are designed to be used once and discarded. He pointed out that you would not go to war and lose all of your soldiers in the first engagement , which reframes the conversation toward reusability and sustained operations.
Ben Levinson extended that thinking by asking what happens when scale increases dramatically. He described a near term reality where millions of systems could be operating simultaneously and asked how those systems function at that level . That question changes the nature of the problem. It is no longer about individual system performance. It is about whether entire networks of systems can operate reliably under pressure.
This is where the technical challenges become structural. Endurance becomes critical because battery powered systems cannot support long distance missions. Navigation becomes unreliable when GPS is denied or unavailable. Communication creates risk when signals can be detected or disrupted. Each of these factors introduces fragility, and at scale that fragility compounds.
Ryan brought a different dimension into the conversation by focusing on time. He explained that mission critical systems cannot wait years for upgrades and that the goal is to move from development to field testing in months, not years . This is a significant shift. The bottleneck is no longer the ability to design systems. It is the ability to integrate, test, and iterate fast enough to keep up with real world conditions.
The Real Constraint Is Not Innovation
What made this discussion valuable is that it did not present innovation as the primary challenge. The constraint is everything around it.
Supply chain limitations continue to restrict access to key components. Many critical parts are not produced within the United States or allied nations, which creates dependency and slows production. Even when components are available, replacing or reconfiguring systems introduces additional engineering complexity that is often underestimated.
Regulation adds another layer. Export controls, compliance requirements, and approval processes introduce delays that do not align with operational urgency. These frameworks serve an important purpose, but they also create friction that is difficult to ignore.
Manufacturing is where the challenge becomes most visible. Franco explained that building a prototype is not the hard part. The real difficulty is building the system that produces those prototypes at scale . That distinction matters because it separates demonstration from deployment.
Cost adds further pressure. Systems must be both capable and affordable, but those objectives often conflict. The concept of cost per outcome is increasingly shaping how systems are evaluated. Low cost systems can be produced in volume but may lack durability. High performance systems deliver capability but are harder to scale.
Cybersecurity introduces a different kind of risk. Ryan highlighted that many vulnerabilities exist at the firmware level where visibility is limited and verification is difficult. As systems become more interconnected, the ability to secure every component becomes increasingly complex. The implication is clear. As deployment accelerates, exposure increases.
What This Means for How Systems Are Built
Across the discussion, one pattern became clear. These challenges are not independent. They are interconnected.
Scale increases dependency on reliable navigation and communication. Speed increases exposure to integration and security risks. Manufacturing constraints influence design decisions from the beginning. Cost pressures shape how systems are deployed and reused.
At the center of all of this is a simple but critical requirement. Systems must be able to operate with confidence in what they perceive.
If the input is incomplete or distorted, every layer built on top of it becomes less reliable. At small scale, that may be manageable. At large scale, it becomes unacceptable.
This is where the idea of perception as infrastructure becomes relevant, and it is a position Circle Optics has been building toward. Not as an added feature, but as a foundational requirement. When systems can capture the full environment accurately and in real time, without relying on reconstruction or correction, the rest of the system becomes more stable.
Decision making becomes faster because less interpretation is required. System performance becomes more consistent because variability is reduced. Operational risk decreases because there are fewer unknowns.
The panel did not explicitly frame the conversation this way, but the implication was clear throughout.
As autonomous systems scale, the margin for error narrows. And when that margin disappears, the quality of perception determines the outcome.
