Not a résumé. The themes that keep pulling me back.
I've had two working lives that turned out to be the same one. The first was security — the holistic kind, done with activists, journalists and NGOs, where the threat model has real people in it and the cost of platform dependency is not theoretical. The second is engineering: building the sovereign alternative instead of writing another assessment recommending one. The systems on this site — a Nostr-native deployment control plane, a key bunker for agent fleets, code review on local models, knowledge infrastructure with provenance — are what that looks like when you follow the logic all the way down.
The road here was not a straight line: logistics for MSF and Oxfam in Pakistan, Malawi and Papua New Guinea; actions and security for Greenpeace from Sydney to Amsterdam to Jakarta; a decade running a technical consultancy; a stint building sovereign engineering infrastructure at Budabit. These days I'm Director of Engineering at Feanix, building an AI-native engineering organization — and the systems on this site are the open-source spine of how I think that should work.
If you want the dated version, the CV is one click away.
Recurring themes
- Sovereignty
- Owned tools, owned keys, owned data — platforms as optional. The organising principle behind everything below.
- Nostr
- Signed events as a substrate for identity, git, deployment, review and coordination — not a timeline.
- Security for people
- Years of holistic security work with activists, journalists and NGOs. Threat models with humans in them.
- Agent fleets
- Autonomous agents as first-class operators — with real identities, policies, provenance and audit trails.
- Memory & knowledge
- Reference truth, episodic recall and institutional doctrine as separate systems, each with provenance.
- Distributed systems
- State that survives partitions, machines, decades and companies. PostgreSQL canonical; everything else derived.
- Local inference
- Models running on hardware I control. Code that never leaves the building it was written in.
- Open source
- Legibility, permissive licenses, and code that outlives its author.
How I work
Keys before accounts. Protocols before platforms. Local before remote. I run a fleet of AI agents as colleagues — each with its own signed identity, policy and audit trail — and I hold the systems they work in to the same standard I'd demand for any human team: legible, replayable, and honest about failure.
Where to find me
The notebook is the best place. Longer essays live in research. Working software is in systems. Correspondence and speaking requests through contact.