SandboxAQ tests GPS-free navigation software on Northrop Grumman's expendable drone

Why it matters

SandboxAQ is turning magnetic navigation into software that can fit existing drones. That could make GPS-independent positioning practical on lower-cost aircraft, but the test disclosed no accuracy data or contract.

SandboxAQ announced Wednesday that it had flight-tested its AQNav magnetic navigation software on Northrop Grumman's Lumberjack, an attritable one-way attack drone built to operate in contested airspace. The more consequential detail in the September 9th announcement distributed by PR Newswire was the installation: SandboxAQ says its engineers added AQNav to Lumberjack's existing systems in less than an hour.

For founder and CEO Jack D. Hidary (@jackhidary), the test pushes a research-heavy navigation program toward a product that defense manufacturers could integrate without adopting a proprietary sensor and computing stack. Hidary studied neuroscience at Columbia University and held a clinical neuroscience fellowship at the National Institutes of Health, where he worked on brain imaging and neural networks. He later helped build the internet business EarthWeb and co-founded Vista Research, which Standard & Poor's agreed to acquire in 2005.

Hidary formed the group that became SandboxAQ during six years at Google, combining AI research with quantum technologies before Alphabet spun SandboxAQ out in 2022. His current thesis is that physics-, chemistry- and biology-grounded models can handle industrial problems that language models cannot answer reliably. AQNav puts that thesis in an aircraft: the software estimates location from measurements of the Earth's crustal magnetic field rather than depending on a signal broadcast by satellites.

Magnetometers aboard the aircraft detect local variations in that field. AQNav compares the readings with magnetic maps and applies physics-based models to calculate a position. Because the reference signal comes from the Earth, SandboxAQ describes AQNav as passive, all-weather and resistant to GPS jamming or spoofing. The software can operate alongside inertial, visual and satellite navigation rather than replacing every other source of positioning data.

The Lumberjack demonstration paired magnetic navigation with the drone's visual navigation system. SandboxAQ called it the first reported MagNav test on an attritable aircraft and the first reported combination of MagNav and visual navigation on an attritable attack platform. Those firsts remain SandboxAQ's claims.

The announcement did not provide the flight's date, location, duration, navigation error, route or environmental conditions. It also did not publish a comparison with GPS, inertial navigation or another magnetic navigation system. The test therefore demonstrates integration and flight operation, rather than establishing how AQNav performed against competing approaches. SandboxAQ and Northrop Grumman did not announce a production contract, operational deployment or purchase.

The software is the product

SandboxAQ describes AQNav as a software-first, hardware-agnostic system that can run on existing onboard computing and use open-architecture interfaces. That approach is central to Hidary's distribution bet. Magnetic navigation has traditionally depended on sensitive instruments and careful installation around the electromagnetic noise generated by an aircraft. SandboxAQ wants aircraft makers to treat MagNav as another input in an open navigation architecture, with SandboxAQ characterizing sensors and processing their output in software.

The approach is particularly relevant on attritable aircraft, which are designed around lower costs and acceptance that some airframes will be lost. Expensive, specialized navigation hardware becomes harder to justify as militaries buy larger numbers of uncrewed systems. Software that can use existing onboard compute and a range of sensors has a clearer path onto those platforms, provided its accuracy survives operational testing.

Northrop Grumman describes Lumberjack as a modular Group 3 uncrewed aircraft capable of loitering, precision strike, electronic warfare and intelligence missions. Its published mission systems include visual positioning and satellite communications. Lumberjack can be launched from the air, ground or sea.

The SandboxAQ announcement identified Max Schuster as Northrop Grumman's Lumberjack program manager. SandboxAQ navigation general manager Luca Ferrara positioned the sub-hour installation as evidence that AQNav can move across uncrewed platforms without lengthy hardware redesigns, according to the announcement.

A defense wedge for SandboxAQ's broader model business

Since 2023, SandboxAQ has reported AQNav tests involving the U.S. Air Force, Airbus and Boeing, including flights aboard C-17 and C-130J transports. AQNav was also selected for NATO DIANA's 2025 cohort and participates in the Defense Innovation Unit's Transition of Quantum Sensing program.

Those programs give SandboxAQ access to aircraft and defense-relevant test conditions, though SandboxAQ's disclosures do not indicate how many flights involved paying customers or operational systems. The next threshold is measured performance that aircraft manufacturers and government buyers can use in certification and procurement decisions.

Hidary has ample capital to pursue that process. SandboxAQ said it had raised more than $950 million after expanding its Series E to more than $450 million in April 2025. Investors in that financing included Ray Dalio, Horizon Kinetics, BNP Paribas, Google and Nvidia, joining backers including Breyer Capital, Eric Schmidt, T. Rowe Price-advised funds, Paladin Capital and S32.

The federal government has since become a direct financial stakeholder in another part of Hidary's strategy. In June 2026, SandboxAQ signed a $500 million CHIPS research and development agreement that gives the Commerce Department a minority, non-voting equity stake. That award supports quantitative AI for semiconductor materials and is separate from venture funding.

SandboxAQ has been applying the same broad model strategy across navigation, materials, cybersecurity and drug discovery. In August, RuntimeWire reported that Hidary had put SandboxAQ's numerical drug-screening models behind an interface using Anthropic's Claude. AQNav follows the same commercial playbook in a less forgiving setting: package specialized scientific work so an established customer can add it to systems already in use.

Lumberjack gives Hidary a useful proving ground because the aircraft combines autonomous software, multiple navigation sources and the cost constraints of an expendable platform. The flight test shows that SandboxAQ can get AQNav aboard that class of drone quickly. Navigation accuracy, repeatability and a path from demonstration to procurement will determine whether the one-hour installation becomes a defense product rather than another successful flight test.