FOR IMMEDIATE RELEASE
WASHINGTON, January 22, 2026. The Heritage Foundation published TIDALWAVE on Tuesday. The Watch was the AI and tradecraft partner. I want to write briefly about how the work got done. This is the first of these notes. They will appear as the work warrants them.
Heritage's Allison Center led the project. Heritage's Center for Data Analysis collaborated internally. The University of Central Florida's Modeling and Simulation Center built the visualization layer. We were responsible, alongside the Allison Center's analysts, for compiling, vetting, and modeling the U.S. and PRC force structures, and for running the analytical work through our structured pipeline.
That pipeline is what the Heritage report calls our "structured rinse cycle." It covers five stages: sourcing, verification, tradecraft, audit, and formatting. Each stage is human-led and AI-enabled. None is free-form generation. The agents in our platform run narrowly-scoped procedures developed by former senior intelligence professionals on our team. The procedures enforce structured analytic techniques. Model behavior is constrained to reproducible steps. Every machine-produced output is reviewable, auditable, and subordinate to human tradecraft.
This is the part of the work that does not photograph well. It is also the part that determines whether intelligence-grade analysis holds up under the scrutiny it is supposed to face.
Within The Watch platform, TIDALWAVE ran through four of our analytical pipelines. Foundational Intelligence built the baseline assessments of the systems and forces under examination. Anticipatory Analysis projected sustainment trajectories under varying conflict scenarios. Actionable Analysis identified leverage points and vulnerabilities. Argumentative Analysis stress-tested findings against alternative hypotheses. The methodology is documented in chapter two of the Heritage report, and serious readers can examine it for themselves.
What I will say about the result is that more than 7,000 curated sources are not assembled by accident. They are the product of systematic harvesting, deduplication, source-category profiling, recency weighting, and entity-concentration checking. They are run against ICD 203's nine analytic standards and the related sourcing requirements. The sourcing was itself an object of analysis. Heritage's authors called this "analysis of the sourcing of the analysis." We agree with the framing. It is the discipline that distinguishes work that survives review from work that does not.
I want to name two of the people who made this work possible. Barclay Adams led the model development on our side, and his work shaped the simulation in ways most readers will not see but that are an engine of the report's contribution. Jim Fein, the Allison Center's Senior Research Associate for National Security and Defense Industrial Base, anchored the U.S. and PRC sustainment research that this project depended on. The named authors at the Allison Center wrote a report worth reading and the report's recommendations are theirs. I am not writing here to second any of them, only to describe the work we contributed.
The conclusion of TIDALWAVE that has drawn the most attention is the redaction request. Senior officials at the Department of War asked for findings to be removed from a public, all-open-source report before it was released. The methodology worked as intended. Open-source analysis, applied with discipline at scale, can produce findings that would have once required classified collection to reach. The data was always public. What changed is the discipline applied to it.
Two more notes worth making before closing.
The first is for the intelligence community. TIDALWAVE was an external project whose findings, drawn entirely from public sources, the Department of War asked Heritage to redact before release. The next such project will not announce itself in advance. The community has an interest in being on the inside of this methodology rather than reacting to its outputs.
The second is for organizations that have leaned on the assumption that public information is, by definition, low-value information. That assumption is wrong, and TIDALWAVE is the public proof. The question for any organization that depends on intelligence work is no longer "is this material classified?" The question is "is this material structured?"
TIDALWAVE demonstrates that the discipline works at scale. We expect to apply it elsewhere. This is the work The Watch was built for.