Manifesto

The battle before the battle

America’s greatest challenge in fielding a software-defined arsenal is shifting from building systems to proving they work.

Proving a system means obtaining the evidence of how it works, how it compares to alternatives, and how best to employ it. The military has to answer these questions for everything it sends into combat. The proliferation of software and robotic systems that change constantly has outpaced our ability to prove them, and it is quietly creating a crisis in how we prepare to fight.

Proving has not accelerated because the system was never designed to. For decades, the government has optimized around the assumption that a weapon is a discrete platform with a stable configuration, operating within a defined envelope and countering a threat that can be treated as fixed for years. Under those assumptions, it made sense to prove a system in occasional events and judge it in isolation. Today, an autonomous ground vehicle, counter-drone radar or software application all move through a series of disconnected test events and exercises, often spread over years, before ever reaching the field.

That pace suited the prior era of exquisite hardware, but robotic and autonomous systems break the assumption. Their behavior is defined by software that changes constantly, and each update can change how a vehicle maneuvers, what a sensor detects or how an operator makes a decision. That speed and configurability should be an advantage, but without the ability to prove it just as fast, organizations treat it like a liability. They are left choosing between holding back improvements that could help our warfighters and fielding systems on unproven assumptions.

No one is more aware of this crisis than the service members involved in these test events and exercises. Lacking modern infrastructure, evaluation teams must stitch together their own solution from bespoke tools, scoring systems, scenarios and threat models. Personnel often resort to estimating distances and time deltas based on eyeballing an operator’s screen, and then document those values in spreadsheets and chat rooms, followed by months of compiling summaries in PDFs and PowerPoints. In many cases, the raw evidence behind multi-year, multimillion-dollar procurement decisions is never collected at all.

This is not the fault of vendors or evaluators; each is doing remarkable work with the resources they are given. But the result is that decision makers are left to sift through a pile of reports that do not provide the understanding they need. Systems get fielded on claimed performance instead of verified data. If those systems fail to work as promised, the cost is no less than mission failure and warfighters’ lives.

This is the battle before the battle: learning what works before our warfighters discover the hard way. It will not be won by running the old model faster. It will be won by proving systems at the pace they are developed, on ground truth captured the same way at every site, so that we hold an honest, continuous understanding of our arsenal.

We founded Valkryn to build the infrastructure for generating that ground truth understanding, in partnership with the institutions fighting this battle.

Generating ground truth

Ground truth is the record of what actually happened. We seek ground truth to make decisions on evidence rather than claims.

Capturing it is deceptively difficult. The instrumentation has to be more precise than the system under test, and every mission measures something different, so each demands a careful approach to sensing ground truth. Mission-level outcomes also depend on many systems sensing, communicating and acting in concert. That requires data from each one to be collected and synchronized before anyone can say how a single system, or a single configuration change, affected mission success.

This is essentially integration testing in the physical world. The software industry figured out how to do this years ago, creating stable, virtual environments where many software systems can interact and get spun up and down with each code change. The physical version is the same idea, putting many systems together to see whether the whole thing works, with none of the stability of a virtual environment. While we can’t control strong winds or component failures disrupting the test event, we can design a repeatable harness for environment instrumentation, data capture and analysis. Valkryn builds the open baseline for that harness, which we call the Battle Lab.

A Battle Lab has three layers.

Layer 1

Environment

The first layer is an Environment designed to impart an experience for systems under test that mirrors combat conditions. If the environment is incomplete, so will the evidence generated in it. The most realistic environment is live combat, which is why many industry and military teams travel to prove there directly. The goal of the Battle Lab is to build the tech stack required to mirror those effects as much as possible, and then go beyond that to simulate different variations.

Layer 2

Measurement

The second layer is Measurement. Every system and operator action of value has to be captured with sufficient precision and synchronized to a common time source to accurately reconstruct a rich event timeline. Instead of visual estimates and hand-picked metrics, the Battle Lab end-to-end instruments the environment and consolidates every feed into a unified picture that is scored against common standards.

Layer 3

Learning

The third layer is Learning, where we slice and aggregate measurements to get answers to our questions. Intelligent copilots can pick out the important patterns, like how new versions correlate with outcomes, or which tactics correlate with mission success. They can play out alternative outcomes, asking what would have happened if a sensor had been placed on the other ridge or a threat presented a different behavior.

Together, these three layers turn an event into evidence that compounds. The Battle Lab is mobile by design. Its hardware and software elements can be stood up and replicated wherever evaluators test or forces train. Battle Labs are also open, system-of-system architectures. Generating ground truth requires integrating the best instrumentation for the mission, just like the warfighter deserves best-of-breed capability.

Earning trust

Ground truth can be uncomfortable. It can highlight deficiencies in a technology or tactic that its followers stubbornly champion. That is why we have to distinguish evidence from advocacy, and recognize when incentives prevent honest conversations.

As founders, we have witnessed this firsthand while building autonomous systems, mission-critical hardware and enterprise data platforms across Palantir, Anduril, SpaceX and Applied Intuition. We have been in the shoes of industry, watching decisions get made without transparent evidence. We have also seen how many people in government are fighting to do it right even when the system rewards taking the easier path.

So with Valkryn, we wanted to design a company that operated as an independent, commercial entity, and whose core business does not overlap with the operational technology evaluated in the Battle Lab. As always, the government decides what gets proven and what gets fielded. Valkryn delivers ground truth through methodology that can be inspected, letting ground truth speak for itself. While no approach is perfect, and our model is not in itself the solution, we do think it is a meaningfully stronger model than conflicted systems integrators and academic organizations that fill the role today, combining the neutrality of the latter with the scale of a commercial product.

Field what works

To those charting a different model within government, who understand that capability now iterates within continuously shifting requirements, and that our institutions have yet to develop that muscle: Valkryn was built for you. Together, we can build a model of proving that keeps pace with capability and is accountable for the answers it delivers.

The battle before the battle is well underway. Victory in the next fight favors the side that experimented honestly in peacetime. We founded Valkryn to make sure that side is ours.

See you in the Battle Lab.

~ Hugh & Andrew

17 Nov 2026