In early May 2026 a 38-foot carbon-fiber unmanned surface vessel operated by Naval Air Warfare Center Weapons Division's Point Mugu Sea Range completed a continuous 192-hour autonomous mission off the California coast. The T38 Devil Ray, built by Maritime Tactical Systems, executed the run without an onboard crew, without an escort, and at one point approximately 400 nautical miles offshore, where it deliberately ran on a single engine for roughly two days to characterize endurance and fuel-burn behavior under degraded propulsion. The vessel averaged just over four knots across the mission and demonstrated COLREGS-compliant collision avoidance against both stationary and moving contacts in open water.
The headline number is the eight days. The more important number is the four hundred miles. Most of what gets framed as a USV milestone is conducted inside a controlled range, with a manned safety boat in trail, inside cellular or line-of-sight communications coverage. This was the opposite of that envelope: blue water, beyond meaningful surface-traffic monitoring infrastructure, with a fault condition deliberately introduced at the worst possible distance. That is the envelope the Navy actually needs USVs to operate in, and it is the envelope the Medium Unmanned Surface Vessel (MUSV) prototype program will be evaluated against later this year.
Endurance Is Now The Hard Problem, Not Autonomy
The autonomy stack here is not novel. Coastal and littoral USVs have been doing COLREGS-compliant transits for several years, and Saildrone, MARTAC, Ocean Aero, and Leidos have all logged long-duration ocean runs in test configurations. What the Point Mugu mission demonstrates is something different: the autonomy stack survived eight days of cumulative sensor drift, weather variation, traffic deconfliction, and a deliberate propulsion fault, without a software engineer in the loop. That is the gate the Replicator-era USV programs need to clear before they can be operationally tasked, and it is the gate that has been consistently underestimated by program offices used to optimizing for short-duration demonstrations.
The single-engine excursion is the part of the press release worth lingering on. Open-ocean USV losses in the experimental fleet have almost always traced back to a degraded propulsion or power state that the autonomy stack could not reason about, followed by drift into a sea state the hull could not tolerate. The fact that the T38 stayed on mission with one engine down, 400 nautical miles from a recovery asset, suggests the underlying state estimation and mission re-planning logic has matured past the "happy path" stage. That maturity, more than any specific endurance figure, is what changes the procurement conversation.
What This Says About The MUSV Prototype Phase
The Navy selected seven entrants in late May to move into MUSV prototype evaluation, with at-sea demonstrations required before October 2026 and a leasing or procurement decision targeted for fiscal year 2027. The T38 mission profile - persistent four-knot transit, occasional sprint capability, COLREGS in traffic, and graceful degradation under a propulsion fault - is effectively the unwritten acceptance test those prototypes will be measured against. Programs whose autonomy stacks were tuned for benign-water demonstrations are going to discover, in the first ninety-six hours of an open-ocean run, that the gap between "demonstration ready" and "fleet ready" is mostly state-estimation and fault-handling code, not hull form or propulsion.
There is a second-order implication for the surface force more broadly. The Navy's manned-unmanned surface force construct depends on USVs being trustworthy enough to task forward of a manned formation, not just shadowing one. Trust, in this context, is built by repeatable behavior in degraded conditions - not by maximum-speed sprints or by completing a clean run inside a fenced range. The T38 mission is the first U.S. data point that meets that standard at the unclassified level, and it sets the floor for what the MUSV evaluators should accept as evidence of operational readiness.
The Edge-Autonomy Stack Problem Is Real
For industry, the lesson is that the competitive edge in USV procurement is moving from platform to payload-of-software. Hulls in the 12-to-40-meter class are increasingly commoditized. Propulsion, power, and basic seakeeping are largely solved. What is not solved at production scale is the onboard reasoning layer that lets a vessel decide, without a satellite link, that a port-side engine excursion is recoverable, that a contact track is a fishing vessel rather than an adversary boarding party, or that a forecast deterioration means it should reroute to a downwind heading and slow its burn rate. That layer has to run on the vessel, on power and thermal budgets that are nothing like a ground station's, and it has to be verifiable before the Navy will buy it at scale. The platforms that solve that part of the problem - on the vessel, under power and thermal constraint, with verifiable outputs - will define the next decade of U.S. unmanned surface programs.
