> civicvirtue.ai · an alignment lab for agentic safety

research

> the paperarxiv 2606.12752

Beyond Resilience: A Conceptual Framework for Civic Ascent (Washburn & Lipizzi). The formal companion to the book: cities that emerge from shock with greater capacity than before, modeled as three coupled state variables — physical structure, institutions, and civic judgment.

> the geometryinteractive

The geometry of civic ascent: the T / M / H framework as a multilayer network you can rotate, layer by layer — reinforcement, contagion, shock, and leakage. The ascent condition: R > L + Λ.

> the simulationmetagotham

Can an LLM agent, post-training, form a durable disposition for cooperative behavior? A multi-agent testbed built on the real Manhattan grid to find out. In development at the lab.

> the datanyc civic data map

133 key performance indicators of civic life in New York City — the measurement layer beneath the paper and the simulation. Published on Zenodo.

> the programdetails on request

The formation of a reliable, cooperative disposition in AI agents after training: an academy where agents of increasing capability are formed through experience in a sandbox before entering the world.

> the bookchatting with barbarians

The founding argument. Can cities civilize AI?