Once you’ve secured the sovereign infrastructure to train your own models, the next hurdle is deciding the integrity of the workloads populating that hardware. We are currently seeing a rush to build “Digital Twins”—replicating everything from fashion models for brands like Rainbow to 12th-century saints like St. Neophytos in Cyprus—but there is a massive disconnect between the marketing pitch and the production logs.
In the boardroom, a Digital Twin is sold as a living, breathing virtual counterpart. In the engineering room, we’re often shipping “Digital Puppets”: sophisticated UI layers built on top of asynchronous, stale data. Whether it’s a medical diagnostic tool in Aragon or a virtual influencer, these systems usually lack the closed-loop telemetry and real-time state synchronization that define a true twin. We aren’t building dynamic entities that evolve alongside their physical subjects; we are deploying sophisticated, static snapshots.
The “revolución” of AI is currently hitting a wall of data gravity. While cities like Córdoba boast world-class connectivity, the software layer is still struggling with “generative resemblance”—the tendency for an AI to look right while being functionally ungrounded. To move from “resemblance” to “synchronization,” we need to stop prioritizing the generative output and start solving for the real-time data fabric. Until we bridge this gap, we aren’t creating twins; we’re just haunting our hardware with expensive, high-fidelity ghosts.