Physics Gets a Vote: Nominal Cofounders on Hardware Development in an AI World

The New Age of Hardware and the Broken Status Quo

We are entering a new age of hardware development, marked by a rapid re-industrialization and a compression of development timelines. This shift is driven by a new generation of founders, trained at innovative companies like SpaceX, Tesla, and Anduril, who are excited to build real things in the physical world. However, the tools and workflows for hardware engineering have not kept pace, creating a significant bottleneck.

The current state of hardware test data management is shockingly outdated. Unlike software development, which benefits from mature tools like GitHub and established CI/CD and DevOps practices, hardware engineering is still largely stuck in a pre-cloud era.

• Fragmented & Localized Data: Data is often stored on network-accessible drives, forcing individual engineers (avionics, GNC, thermal, etc.) to download massive files to their local machines.
• Siloed Analysis: Each engineer runs their own analysis using disparate tools like MATLAB or Python, leading to independent, uncoordinated results.
• Manual Collaboration: Insights are shared manually, often through screenshots pasted into PDFs and PowerPoint presentations. This “PowerPoint engineering” is still the bleeding edge for many organizations.

This fragmented approach means there is no centralized system of record, making it incredibly difficult to collaborate, maintain traceability, and build the foundational data sets required for modern AI.

Bridging the Simulation-Reality Gap

A significant challenge in hardware development is the gap between simulation and real-world performance. While there has been a push to simulate everything early in the design lifecycle, the reality is that “physics gets a vote.” The real power comes from continuously and iteratively blending simulation model outputs with real-world telemetry and sensor data from physical systems.

Nominal strategically focuses on hardware testing as its initial wedge because this is the critical juncture where software-defined hardware first touches reality. It is the tip of the spear for how software and AI will impact the development of physical systems. By creating a system of record for test data, Nominal provides the infrastructure to:

• Unify Disparate Teams: Connects the simulation team, the prototyping/R&D team, and the manufacturing/in-field monitoring teams onto a common data platform.
• Enable Continuous Validation: Allows validation logic developed during initial testing to be seamlessly applied throughout the lifecycle, from end-of-line manufacturing quality checks to in-field monitoring at the edge. The core Nominal product becomes the authoring hub for all logic governing the performance of a physical system.

The Future is Physical AI

The ultimate vision is a future where all hardware companies become “physical AI companies.” This requires a fundamental shift from the current state to one where data is a core, structured asset. Nominal provides the semantic layer necessary to prepare organizations for this transition.

The Foundational Verification Layer

Before AI can design, build, and test hardware, a robust verification layer is essential. In the software world, agents for coding have become powerful because the environment is verifiable, allowing for rapid learning loops. Hardware lacks this infrastructure.

The physical world will just always be different because it’s what we live in. If you have AGI designing a video game for your child, you might let them play it without it being rigorously tested. But if you had AGI building a toy for your child, you would really want to make sure that it wasn’t physically dangerous.

Nominal is building this necessary test and validation infrastructure. By cataloging and structuring vast amounts of test data—from millions of sensor data points per second to the audio of operators talking during a test—the platform creates the high-fidelity, human-enriched datasets required for AI research and development. It helps ensure data quality by automatically identifying anomalies in the physical hardware (e.g., a faulty servo in a robotic arm), allowing teams to filter out corrupted data before it pollutes a training set.

From Deterministic to Agentic Testing

The paradigm of testing itself is evolving. The traditional approach involves sequentially working through a giant, deterministic matrix of test points, a process that can take years. A new approach, which Nominal is exploring with DARPA, involves using AI agents paired with digital twins to recommend the next, knowledge-maximizing test condition in faster-than-real-time. This shifts testing from a discrete, linear process to a continuous, optimized gradient curve, dramatically accelerating the development cycle.

While the “AI Nirvana” of vibe-coding an airplane is still distant, the immediate future involves a massive increase in hardware testing to collect the necessary data to build these models. Nominal is positioned as the critical data and AI platform to enable this, turning the tedious, fragmented process of today into the automated, intelligent, and continuous hardware development of tomorrow.

• Tags:
• Simulation
• Data Infrastructure
• Hardware
• Verification
• Physical AI
• Engineering
• Hardware Testing