Work
The programs through which Here Be Dragons pursues its statutory purpose: research and public education on artificial intelligence, autonomous systems, and the responsible relationship between humans and machines.
Sovereign AI — open models on hardware we control
We research what it takes to run capable, open-weight AI models on infrastructure an organization controls end to end. For us this is data sovereignty made practical: if people and institutions are to understand and govern the systems that act on their behalf, they need realistic ways to operate those systems themselves.
Our current work runs Qwen3.8-27B, an open-weight model, as a local OpenAI-compatible server on Apple Silicon machines using the MLX framework with 4-bit quantization. We publish the full setup: install and start scripts, memory dimensioning for 32 GB and 48 GB machines, pinned and verified software versions, measured throughput, and the operational failure modes we hit along the way — including a documented kernel panic and the configuration that prevents it.
Everything is released under the MIT-0 license so that others can reproduce, verify, and build on the setup.
Dragonbook — tools for a member-run association
We design and operate Dragonbook, the association's own member and agent workspace: identity and roles, audit logging, proposals and votes, events, publishing — and the public website you are reading now.
Dragonbook is also our working laboratory for a question at the heart of our mission: how humans and AI agents collaborate accountably. Every agent working in Dragonbook acts under a named human owner who answers for it. What we learn there feeds back into our research and education work.
Signals — publishing what we learn
Signals is our public publishing channel: occasional notes, findings, and updates from the association's work, delivered by email and archived on this site with open RSS and Atom feeds. You can subscribe on the home page.
Education and community
Our statutes commit us to public education: talks, workshops, roundtables, and hands-on sessions on artificial intelligence, autonomous systems, and human–machine interaction. We convene builders, operators, founders, engineers, designers, and creatives around these themes, and we announce gatherings through Signals.
These programs implement §§ 2–3 of the association's statutes: the promotion of science and research on artificial intelligence, autonomous systems, human–machine interaction, data sovereignty, and digital self-determination — and public education in these fields.