The architecture I wrote in 2021, validated by 2026.

One body of work in two movements. First, the Synchronicity papers, a decentralized AI governance platform sketched from 2018 to 2022 and independently validated by the 2026 agent landscape. Second, OrgOps, the buildable architecture that follows. The papers were right about the architecture and wrong about the substrate; both lessons live here.

Essays

The Ontology-Driven Control series. Two posts are live now; more are coming.

• The Through-Line: Ontology as Operational Substrate

  An ontology is the runtime substrate an AI control system stands on: state as a typed graph, actions as functions, and the rule that the LLM proposes while the logic disposes.

• The Semantic Stack: RDF, OWL, SHACL as Engineering Materials

  RDF, OWL, and SHACL treated as materials with real performance trade-offs and failure modes, not academic formalisms.

• Decentralized Org Control: First, a Decentralized Ontology

  Decentralized control is downstream. Before an org can be run by an evolving ontology, you need a system that creates that ontology without a central owner - and that is the hard part.

Interactive exhibits

Standalone dark visual guides. Each opens in its own page.

• Natural Games: a visual guide

  An interactive walk through the Anttila & Annila paper that grounds economics in thermodynamics, with entropy as the universal payoff.

• Pressure Field to Discrete: a visual explorer

  Five ways to bridge a continuous pressure field and a discrete representation that agents can read and write, with live animations.

• ERPBench / Anchor: a homage

  A love letter to ERPBench: 300 verifiable long-horizon tasks in a real ERP, generated from one constraint program so instruction, environment, solution, and grader cannot drift.

Papers

The original three. Full text is coming soon.

• Synchronicity (2021) Decentralized economics platform
• Cooperative AI (2021) Process mining for RL environments
• Twitter DAO (2022) Governance via Coase