Shared tools, an OS and autonomy that make cars, trucks, drones and warships intelligent.

A bet against a single winner

Applied Intuition starts from an argument about how the car industry is built. Its founding deck, as Andreessen Horowitz described it in July 2026, had a slide called Ideas We Believe To Be True. One of them held that any single autonomous vehicle maker would stay a single-digit share of the global car market, because cars are made and sold by many manufacturers. In 2017 most money went the other way, into full-stack robotaxi programs that each hired engineers, built a fleet and rebuilt roughly the same development tools. If no one company took the market, the rest of the industry would need intelligence too, and most manufacturers could not rebuild every layer of software themselves. An independent supplier would have to provide the common infrastructure. That is the company Qasar Younis and Peter Ludwig set out to build.

What’s different about Applied Intuition versus other companies in our ecosystem is we don’t succeed if our customers don't succeed.

Qasar Younis · Co-founder and CEO · Isuzu partnership announcement · 2024-08-26

The second belief explains the shape of the business today: autonomy is one engineering problem whether the machine drives, flies or sails. The company sells three layers, tools and infrastructure, an operating system and autonomy, to automakers, truck makers, mining, construction and farm equipment makers, and to the U.S. military and its allies. It calls this dual-use. Why now has two answers. In cars, the vehicle became software-defined, and in 2026 Stellantis, Honda and Nissan announced work on the company's Vehicle OS. In defense, the company's own pages argue that the Department of War is fielding autonomous systems faster than it can test, integrate and sustain them, and in May 2026 the Pentagon's Chief Digital and AI Office chose Applied Intuition to build the platform under Autonomy Factory, a shared pipeline for every service.

One mission, followed through the toolchain

Take an illustrative program that wants drones from several vendors to search an area as a team. Today that is usually proven by putting real vehicles on a range, at a cost the company puts at hundreds of thousands of dollars per vehicle, a few times a year, and often the live event is the first time the vendors' systems talk to each other. In a digital proving ground the same mission first runs in a vendor-neutral simulation with hundreds to thousands of agents. Axion generates synthetic sensor data with ground-truth labels, so perception can be trained and tested before a flight. The same tests then run software-in-the-loop, with real hardware in the loop, and with a few live platforms flying beside virtual ones. When the team reaches the range, the company says a problem can be found, fixed and flown again within an hour instead of weeks. Afterwards a data engine brings field data back, retrains models and updates them over the air.

People's judgement stays in three places. Someone has to turn an operator's need into scenarios and pass criteria; the Seoul defense postings ask engineers for the step before it, drawing out Korean defense customers' operational requirements and turning them into product specifications. Someone has to decide where simulation fidelity matters and close the gap between simulated and real flight, a task the drone autonomy posting spells out. And a person supervises the machines in the field; in the 2026 AeroVironment deal, a team of drones works under the supervision of a single operator. The commercial loop is the same idea with cars: Toyota's Woven Planet group used the tools to run tens of thousands of scenarios at once in the cloud.

What a buyer is really paying for

Dual-use here is code reuse, not a slogan. The Scientific Systems case says many of the simulation features used for its maritime work were first built for on-road autonomy programs. The defense business therefore inherits years of automotive engineering, and a buyer is paying for fewer test events and faster fixes more than for any single feature.

Two Seoul roles build parts of this loop: defense tooling, and multi-agent drone autonomy tools that must hold up on real airframes. The Lead Customer Engineer and the new-grad FDE take the same toolchain into customers' programs. All four are on site in Seoul, and the two defense roles collaborate daily with US teams.

Four products on three layers

Applied Intuition's catalogue reads as three layers, each sold to commercial and defense customers, sometimes under different names. The tools layer is Tools for Vehicle Intelligence for commercial customers and Axion for defense. The commercial toolchain grew from the first simulators, Simian, Spectral and Orbis, into products for drive data, maps and validation, and now covers petabyte-scale data ingestion, a closed-loop flywheel that turns sensor logs into labeled training segments, neural simulation and an SDK that runs on cloud, on premises or in air-gapped networks. Axion, launched in May 2025, gives defense engineers and operators one environment to build, test and deploy, plus a data engine for fielded systems. The U.S. Navy's first large-scale data engine, the Data Edge Collection Kit, is part of that work, and in May 2026 Axion became the base of the Department of War's Autonomy Factory.

The autonomy layer is the Self-Driving System for commercial vehicles and Acuity for defense. SDS for Trucking drives Isuzu's heavy trucks in Japan, and SDS for Automotive, launched in 2025, reached Japan in June 2026 using production cameras and radar without HD maps or lidar. Acuity runs onboard ground, air and maritime vehicles and coordinates several of them over mesh networks in jammed or denied environments. It builds on EpiSci, a defense autonomy company bought in February 2025 whose software was used in the U.S. Air Force's first successful dogfight between an AI-piloted jet and a human pilot, and it has been deployed on the X-62A VISTA jet.

The middle layer is Vehicle OS, an operating system and code-first toolchain that runs across a vehicle's computers. The company says it can reduce integration effort by up to five times. TRATON is building TRATON ONE OS on it for all new vehicles across Scania, MAN, International and Volkswagen Truck & Bus, with rollout targeted for 2028; Stellantis expanded its work into the STLA Brain platform in May 2026; Honda and Nissan followed in October; and in April HII signed an MOU with the company on Warship OS. Above all three layers sits Dana, launched in July 2026, which lets AI agents drive the toolchain.

Named customers, and what each case measures

The company publishes named cases on both sides. Scientific Systems, which builds collaborative autonomy for the U.S. military, used to spend 6 to 12 months and millions of dollars arranging a field trial that often yielded dozens of usable logs. With Axion it now runs up to 20 virtual tests for the effort of one physical trial, and vessel setups that took a day run in minutes. For the U.S. Army, engineers turned a bare-bones Infantry Squad Vehicle autonomous in 10 days, then spent two weeks with soldiers of the 101st Airborne at the Joint Readiness Training Center, pushing software updates live. Isuzu's trucks ran two seven-day demonstrations for the Japanese government in 2025, each covering 1,800 km of autonomous driving with 92 terminal maneuvers. In Korea, the Police Science Institute of the Korean National Police University used Simian and Basis to turn the Road Traffic Act into test scenarios.

Live testing simply can’t keep pace with the modern battlefield. Digital-first development and integration are essential to scaling unmanned systems.

Peter Ludwig · Co-founder and CTO · Axion and Acuity launch · 2025-05-20

How to read these cases

Every result here is published by Applied Intuition, and the measures differ: test effort for Scientific Systems, days to autonomy for the Army, kilometers driven for Isuzu. The defense results describe demonstrations and exercises. In an interview, ask how a program moves from a demonstration to a production contract, and which measure the company tracks across customers.