Sovereign AI Is Not a Flag
Sovereign AI is often discussed as ownership of a model. The more useful question is what dependencies remain when the model changes, is withheld, copied or orchestrated.
There is a threshold where engineering becomes governance and infrastructure becomes policy. It is rarely marked, often crossed without noticing. I write here to think at that boundary: the hidden mechanics of digital public systems, the friction between technical possibility and institutional reality, and the quiet decisions that shape both.
Start Here
Latest Articles
- The Infrastructure You Cannot Copy A companion to "Sovereign AI Is Not a Flag". At the model layer, sovereignty is fragile because everything copies. At the physical layer, dependence becomes harder to escape because almost nothing does.
- Reusing public data at the point of friction How cup-check turns OpenCUP data into a local-first operational control, and why the design choices matter for public-sector software.
- Data in the digital transition of Italian public administration Analysis of the persistent gap between digital governance normative frameworks and organizational implementation capacity in Italian public administration.
- No Final Diagram Every stage of this homelab simplified something the previous stage made too complex. The next one will do the same. Infrastructure does not converge on a final state. It converges on a clearer understanding of what you actually need.
- What Breaks, What Survives Backup and monitoring are not separate concerns. Monitoring tells you when something breaks; backup determines how far back you can go. Together, they form the operational contract that makes the rest of the infrastructure sustainable.