Independent intelligence for building AI you control.
Models · Compute · Data · Agents · Edge · Physical AI. Research, testing, and perspectives from practitioners around the world.
What actually matters right now.
Models
Open-weight models are now credible enough to benchmark first for many bounded workloads. The useful question is no longer open versus closed in the abstract, but which option clears the requirement at an acceptable cost and level of control.
Compute
Large unified memory pools change which models a single workstation can hold. That expands the local option for small organisations, but capacity still has to be tested against speed, power, and the real workload.
Agents
A convincing agent run needs more than a fluent transcript. It needs a predeclared definition of done, evidence from the tools and systems it changed, and a record a human can audit.
Edge
Platform-level language models now run on the device with structured output and tool use. That makes local execution a real application boundary, while larger or more demanding work still belongs elsewhere.
Sovereignty
The European Commission has put sovereignty into a cloud procurement framework with scored criteria across jurisdiction, technology, operations, supply chain, data and AI. The term is moving from positioning into evaluation.
Six layers. Every one of them can be owned, run, and inspected by the organisation using it.
Open and controllable intelligence.
Local, edge, and private infrastructure.
Context you own and govern.
Systems that execute real work.
Intelligence where the work happens.
Sensors, machines, robotics, and real-world systems.
We test the stack, not the marketing claims.
PLANNED
Can a small organisation run serious AI locally?
PLANNED
Same workflow. Same workload. Same measurements.
PLANNED
Can an agent execute, verify, and document real work?
Short, dated observations from work in progress.
An edited selection, not a feed.
Sonya Huang frames sovereign AI as an own-versus-rent decision across cost, speed, performance, and proprietary data—and argues that teams should begin with evaluations.
In the room, in the field, with the machines running.
Featured
Builders, researchers, infrastructure companies, and enterprises exploring AI that can operate locally, privately, and independently.
Event detailsPeople building local AI, in the open.
Build in the open. Keep control of your tools.
A community for people building local AI, open infrastructure, agents, edge systems, and physical AI.
Join the communityDISCUSSIONS
What is working, what is not.
BUILDS
Rigs, racks, and running systems.
QUESTIONS
Answers from people who have shipped.
PROJECTS
Work in progress, in the open.
EVENTS
Meetups, workshops, and field labs.
CODE
Configs, scripts, and reproducible setups.
Sovereign AI Stack runs an agentic newsroom that monitors emerging signals, coordinates research, produces media, distributes stories, and learns from audience interest. Editorial judgement stays with people.
How the newsroom worksThe categories of technology and infrastructure this work runs on.
SILICON & SYSTEMS
Local compute we test on.
OPEN MODELS
Weights we run and evaluate.
INFRASTRUCTURE
Storage, networking, deployment.
RESEARCH & EDUCATION
Labs, programs, universities.
COMMUNITY
Meetups, builder groups, events.
PRODUCTION
Media, studio, and field systems.
PROPOSED · NOT YET FOR SALE
A black-base Sovereign AI Stack merchandise direction using the canonical stacked-S mark and Rigamesh identity.
View the collection concept →
One useful briefing on sovereign AI. No daily AI noise.
Email · Video · Audio · Build
Subscription options
