The Agentic State

● Research

Research

Three directions grow out of the Agentic State framework. Each is described here at the stage it has actually reached: what is a concept, what runs as a prototype, and what has not been tested.

These are working texts by an independent research project. None has been peer reviewed, and none reports a pilot or a deployment with any government. Each page states its sources, its limits and how to cite it.

Directions

  1. 01State OSState OS is a research project on decision provenance: a public record, kept beside one real government and holding no authority, that binds each covered cabinet decision to the evidence available at the time, to authored alternatives, to dissent, and to forecasts registered before their outcomes.● Research prototype · no software release, pilot or deployment
  2. 02Virtual LeaderVirtual Leader is a research direction that asks which functions of political leadership can be separated from a human office-holder, and which cannot. Its own analysis rejects a machine that holds authority over people, and keeps a narrower design: software that applies values authored by people, in a non-binding shadow mode, under human custody.● Research and design · no system built
  3. 03Research ProgrammeThe research programme asks how counter-power can be built into the same architecture that makes a state capable of acting through AI agents. It consists of a prior-art audit, a research agenda of eight open questions, each stated with a method and with the result that would count against it, and a working paper on current government cases.● Research agenda and working paper · no study run

Publications

State OS

State OS: Concept and Architecture

A shadow-mode record of government decisions: what was decided, what was knowable, what was expected, and who changed the machine’s draft

Concept paper · 1.1-web · Research Programme

Keeping an Agentic State Answerable

A research agenda: eight open questions, with methods and with the results that would count against each

Research agenda · 0.2 · Research Programme

The Agentic State: Capability and Counter-Power as One Architecture

A field paper on what governments are building with AI agents, and what has to exist on the citizen’s side of it

Working paper · 1.0-web ·