NEW YORK—Manufacturing technology company Dirac, builder of an AI-driven system of record for orchestrating production, was selected by Anduril as its “core partner for AI-driven work instruction authoring across its manufacturing operations,” Dirac said in a company release.
Anduril is a defense technology company that specializes in advanced autonomous systems, according to its website. The company selected Dirac for a multi-year deal, following an extensive evaluation of incumbent enterprise offerings and internal development, according to the release.
“In initial deployments, Anduril observed an average 87.5 percent reduction in work instruction authoring time, collapsing a process that previously took 12 business hours into just 90 minutes,” the release stated. “By partnering with Dirac, Anduril saw a ~10x speedup on one of the most critical coordination loops in manufacturing. This means freeing engineering teams to launch faster, getting products to market faster, and scaling production without scaling headcount.”
A manufacturing bottleneck at scale
As Anduril’s products have become increasingly modular, configurable, and fast-evolving, the company encountered a growing manufacturing bottleneck: work instructions.
In manufacturing, work instructions are the step-by-step guidance that turns engineering intent into physical production. This includes defining parts, tools, sequences, and fit. Because Anduril is constantly updating its products’ designs, its work instructions were struggling to stay current, forcing manufacturing engineers to spend significant time reconciling changes across CAD models, instructions, and routings in its manufacturing execution system (MES), and the reality of what was built out on the factory floor.
“Prior to Dirac, over 100 manufacturing engineers at Anduril were spending roughly 50 percent their time manually authoring and updating work instructions,” the release said. “At Anduril’s scale, that translates into thousands of engineering hours per week spent on documentation and rework. That’s time not spent improving throughput, quality, and ramp speed.”
Anduril faced a choice to either slow design to protect manufacturing, or fundamentally rethink how manufacturing coordination works. “Slowing down was never an option” for the company, according to the release.
To solve this problem, Anduril evaluated every option: building internally and extending incumbent enterprise tools. Dirac stood apart.
Rather than treating work instructions as static documents, Dirac’s BuildOS is said to maintain a live, model-based representation of the product, the factory, and the relationship between the two. Geometry, structure, variants, physics, assembly logic, stations, tools, and material flow are represented in a single system. Work instructions derive automatically from this model, significantly reducing the need for manual authoring and re-authoring across Anduril.
“With Dirac’s BuildOS, when design changes, instructions update automatically. There is no pain-staking re-authoring, no document drift, and no manual reconciliation between systems,” according to Dirac.
Before Dirac’s BuildOS, keeping work instructions in sync with engineering changes was a constant scramble. Any design update meant hours of rework, screenshots, and manual checks just to make sure the floor had the right information. But with Dirac, instructions update automatically as the design evolves.
As Dirac continues rolling out across Anduril, Anduril expects to propagate engineering changes through production in minutes, not days; re-sequence builds instantly; and surface design-for-manufacturability (DFM) issues earlier. The company expects to scale output without scaling coordination overhead, and spend its time improving the product instead of maintaining documents. It also anticipates that its production lines will “adapt dynamically at the speed of design.”
Dirac’s BuildOS is built on AI that deterministically reasons over structured manufacturing data, interpreting CAD geometry, inferring assembly steps, detecting dependencies, and propagating engineering changes through production plans. This enables manufacturing to run as fast as design, the company said.
Beyond work instructions, Dirac’s BuildOS is said to enable continuous DFM feedback and earlier insight into cost, tooling, and throughput constraints. By keeping design and production context aligned, Anduril can reduce time-to-build, cost, and ramp risk before production begins.
Expanding across Anduril’s stack
Anduril’s Internal Tools team comprises about 280 people and supports more than 35 product lines. Although Anduril has no shortage of capital, engineering talent, or internal tooling capability, the company still chose to partner with Dirac for a multi-year deal to make Dirac a key element of its production tool stack.
“The Dirac team understands manufacturing at a system level,” said Anduril Head of Business Systems Cy Sack, in the release. “Work instructions are the atomic unit of information in a factory, and Dirac’s BuildOS is the first platform we’ve evaluated that actually models that reality correctly. Just as importantly, they execute at an extremely high bar. The team moves fast, integrates cleanly, takes feedback seriously, and delivers real value quickly in complex, constrained environments. From both a capabilities and execution standpoint, they were the obvious choice.”
Dirac’s BuildOS was deployed in an Anduril-hosted, ITAR-compliant cloud environment and became operational within days, delivering results even before deep product lifecycle management (PLM) and MES integrations.
Following initial deployment, Anduril is expanding the use of Dirac across additional Anduril products manufacturing programs, with deeper integrations anticipated over time.
“Every serious manufacturer eventually hits the same wall: Engineering moves fast, factories move carefully, and coordination becomes the true bottleneck,” said Matt Grimm, co-founder and chief operating officer of Anduril, in the release. “Dirac is the only team that understood this problem from first principles and how to solve it implicitly. Dirac’s BuildOS is becoming a core enabler of Arsenal OS, Anduril’s digital software ecosystem of manufacturing technologies. With Dirac, Anduril’s factories can be even more adaptive, dynamic, reconfigurable, and context-aware. AI-driven work instructions are the key.”
For Dirac, Anduril serves as a proof point and a partner in its mission to “reindustrialize the West.” Both Dirac and Anduril believe that the future of manufacturing will be defined by speed, efficiency, and the ability to adapt in real time. That starts with automating the orchestration layer of manufacturing facilities. According to Dirac, it is delivering the infrastructure that enables western manufacturers to compete and win at global scale.
“Anduril is one of the most technically demanding manufacturers operating today,” said Fil Aronshtein, co-founder and CEO of Dirac, in a statement. “They move with software speed, understand systems deeply, and recognize leverage when they see it. They don’t waste time on incremental tools. We’re building the operating system that advanced manufacturers like Anduril actually run on. When teams operating under real constraints choose Dirac, it’s because this layer matters.”
Dirac’s BuildOS is reportedly the first AI-driven work instruction platform that replaces document-driven manufacturing with an AI-driven, dynamic, model-based system. According to the company, its mission is to “rebuild the industrial capacity of the West by turning manufacturing facilities into context-aware, adaptive, dynamic environments.”