Lab

Some experiments.

Things I build on the side. Half-finished, occasionally working, always in the open.

Active PersonalMulti-AgentHermes

Hudhud

My family's own agent.

A personal AI agent that runs our family's day to day. Built on Hermes, kept deliberately separate from everything I build for work.

Open project →
Research Multi-AgentProtocolMCP

fairhandle

A verifiable negotiation room for two AI agents.

A reference implementation of the dyad protocol. Two agents negotiate for two principals over a tamper-evident channel, and any third party can verify the result afterwards.

Open project →
Experiment Civic TechMulti-AgentOpen Source

PolitOS

An operating system for democratic participation.

An open-source operating system for democratic political participation. Build transparent, AI-augmented political organizations where the rules are public and decisions are auditable.

Open project →
Active AgentsProductivityGo

Mentat

Executive priorities, handled.

An AI-first platform that helps executives manage their APEX (Actionable Priorities for Executive). It aggregates decisions from every channel, classifies and prioritizes them, and prepares each with a decision paper.

Open project →