The Translation Capture

Translation is the gap between discovery and deployment — the institutional process by which a scientific finding becomes a product, a system, a technology in use. It is the stage where the question "what is true?" transitions to the question "who benefits?" — and where the governance of the Knowledge Architecture is most complete and most invisible.

The translation stage is where the patent thicket captures commercializable discoveries. It is also where the classification system captures strategically sensitive ones, where the equity transfer captures the researchers themselves, and where the revolving door between public research institutions and private deployment entities ensures that the most valuable knowledge flows in the most institutionally captured direction. The Bayh-Dole Act structured the commercialization end of translation. DARPA structures the strategic end. Together they constitute the complete Sovereign Knowledge Pipeline.

The Defense Advanced Research Projects Agency funds high-risk, high-reward basic research at universities and national laboratories. Its funding model is designed to produce transformative breakthroughs — not incremental improvements to existing technologies, but the foundational advances that make new categories of technology possible. The internet. Stealth aircraft. GPS. The semiconductor revolution. Modern AI. Each traces significant portions of its foundational research to DARPA funding.

The DARPA model is not secret: it is a deliberate public investment strategy, and many of its results are published openly. What is less visible is the translation architecture that determines where DARPA's results go after the research phase. The most strategically sensitive results are classified — withdrawn from the public record, transferred to cleared facilities, and developed through defense contractors under security agreements that prevent public disclosure. The commercially valuable results are licensed through Bayh-Dole mechanisms to university spinouts in which DARPA program managers and university researchers frequently hold equity. The results that are neither classified nor commercially capturable are published and enter the public domain — the residual category after the higher-value capture mechanisms have operated.

The revolving door between public research institutions and private deployment entities is not merely an ethics problem of the kind documented in the Accountability Firewall series. In the Knowledge Architecture, it is a translation mechanism: the movement of researchers from public research roles to private deployment roles is the physical pipeline through which tacit knowledge — the knowledge that cannot be captured in patents or papers, that exists only in the heads of the researchers who developed it — transfers from public to private.

A DARPA program manager who leaves to found a startup brings not just her expertise but her relationships with the researchers she funded, her access to pre-publication results, her understanding of which research directions are most promising, and her credibility with the defense procurement system that will be her company's first customer. The translation capture at this level is not about patents or licensing — it is about the embodied knowledge that public investment created and private employment captures. The patent system at least requires disclosure. The revolving door requires nothing.

Contemporary AI development represents the Knowledge Architecture operating at its most complete: all four mechanisms of the KA series operating simultaneously and reinforcing each other. Bayh-Dole: university AI research is patented and exclusively licensed to companies in which researchers hold equity. The Journal Capture: AI research papers are published in venues whose impact factor is partly controlled by the same companies whose researchers dominate the program committees. The Patent Thicket: foundation model architectures and training methods are being aggressively patented by incumbents. The Translation Capture: DARPA AI programs transfer their most valuable outputs to defense contractors through classified programs; the revolving door between leading AI labs and government AI initiatives runs in both directions; and the training data — the accumulated written knowledge of human civilization — was consumed without public return to produce systems whose value accrues to private shareholders.

The cognitive sovereignty implication is direct: the technology that will most determine the cognitive landscape of the next century — AI systems that will mediate information access, assist (or replace) human reasoning, and shape the epistemic environment — is being developed by entities whose governance is not democratic, whose alignment with public interest is not required, and whose accountability to the populations whose knowledge funded their development is nonexistent.