LOS ANGELES, CA / ACCESS Newswire / November 19, 2026 / While Silicon Valley's headline "world models" generate video pixels on power-hungry cloud GPU clusters, Qluu Lab Inc., a U.S. defense and dual-use autonomy software company backed by the Techstars New York Accelerator, is taking world models off the screen and onto real moving machines. Today, Qluu announced that its physics-grounded world-model and autonomy software has been demonstrated in live outdoor trials aboard flying drones and ground vehicles-running entirely offline on an off-the-shelf $80, 12-watt processor.

Qluu Lab is a pure-play software company: it does not manufacture drones, ground vehicles, sensors, or compute boards. Instead, domestic manufacturers and infrastructure operators run Qluu's software stack on inexpensive commercial ARM processors integrated onto their existing vehicles and perimeter sensors-avoiding the cost, battery drain, heat, and supply-chain bottlenecks of $2,000 to $5,000 onboard GPUs.
Aligned with federal and Department of War Super Intelligence (SI) priorities under the Pentagon's newly designated Super Intelligence Office (SIO), Qluu delivers sovereign edge physical intelligence with built-in human-on-the-loop controls. Rather than forcing hardware makers to spend years and millions of dollars building a custom AI stack for every vehicle hull, Qluu provides one reusable software brain.
Following closed-loop command-and-control field experimentation at Fort Hood, benchmark evaluation across nearly 700,000 physics-simulated trajectory windows, and live multi-vehicle outdoor trials in California this November, Qluu Lab highlighted four capabilities:
- World Models That Work in Physical Reality-Substantially Outperforming Physics-Only Filters: Qluu pairs compact learned models with hard physical bounds so vehicles predict how real objects move in three dimensions without hallucinating unsafe maneuvers. Across both held-out multi-domain simulations and live outdoor drone and ground-vehicle trials, Qluu's world models substantially outperformed standard physics-only tracking filters while enforcing human-governed safety rules.
- Nearly 700,000 Simulated Trajectories - With Near-Identical Sim-to-Real Execution: Before flying outdoors, Qluu's models were trained, tuned, and evaluated across 687,137 physics-simulated multi-domain trajectory windows (including more than 76,000 held-out test windows never seen during training) in simulation environments built on the IEEE 1278.1 Distributed Interactive Simulation (DIS) standard. Because Qluu runs the same software stack in simulation and on the vehicle, missions rehearsed in Qluu's digital-twin simulator produced nearly identical actions and flight behavior during live outdoor drone and ground-target trials in California-letting partners test and verify missions on a laptop before ever heading to a flight range, while undersea autonomy and ground-based orbital command-and-control capabilities continue advancing in simulation.
- Resilient Coordination Over 32-Byte Messages When Links Degrade: Rather than relying on fragile live video feeds that stall in contested or bandwidth-starved environments, Qluu-equipped vehicles share target custody and coordinate using compact 32-byte messages-cutting message payload data by more than 99.9% compared with standard compressed camera frames and recovering cleanly after communications blackouts.
- Passive Multi-Sensor Drone Awareness for AI Data Centers & Critical Infrastructure: Because federal FAA and FCC rules restrict private operators from jamming drones or transmitting disruptive radio signals, Qluu's software fuses a facility's existing cameras, acoustic arrays, and perimeter radars without adding radio-frequency jammers-dramatically reducing 3D tracking error to help protect AI data centers, power substations, and ports.
"Everyone in tech is talking about world models, but most of them stay trapped inside a data center generating video," said Michael Jacobs, Co-Founder of Qluu Lab Inc. "Real autonomy has to work outdoors when the wind is blowing, bandwidth is tight, and the entire compute budget is twelve watts on an $80 board. By proving physics-grounded world models on live vehicles in the field, we give American hardware makers and infrastructure operators practical physical intelligence they can license today."
Note: Participation in field experimentation at Fort Hood or Defense Innovation Unit (DIU) events does not constitute endorsement, certification, or procurement by the U.S. Army or the Department of War. Detailed benchmark reports are available to qualified U.S. defense and commercial partners under NDA.
About Qluu Lab Inc.
Qluu Lab Inc. is a U.S.-owned defense and dual-use autonomy software company based in Los Angeles, California, co-founded by Arkadiy Okhman and Michael Jacobs. Backed by the Techstars New York Accelerator, Qluu Lab licenses hardware-agnostic physical world models, passive multi-sensor fusion, and digital-twin simulation software across air, land, maritime, undersea, space command-and-control, and critical infrastructure-manufacturing zero vehicles or hardware so it never competes with its OEM partners. Learn more at https://qluulab.com.
Media & Partnership Contact: Credible PR (for Qluu Lab Inc.) · max@crediblepr.com · https://qluulab.com
SOURCE: Qluu Lab Inc.
View the original press release on ACCESS Newswire
