Agent Teams: A Theoretically Grounded Approach¶
James L. Caton
Early Draft — Experimental
This is an early draft of a working manuscript. Content is incomplete and subject to substantial revision. This manuscript is an experiment in AI-assisted academic authorship: it was written in active cooperation with a agent team, with the author directing the research agenda, theoretical arguments, and editorial decisions while agents assisted with drafting, revision, and consistency auditing.
This working manuscript develops the theoretical foundations of multi-agent AI systems, grounding the design of agent teams in Austrian capital theory, transaction cost economics, and constitutional constraints. Drawing on Hayek, Coase, Lachmann, and Nonaka, it analyzes how specialized agents coordinate under an authority hierarchy, how knowledge is generated and propagated, and how constitutional rules prevent scope creep.
Contents¶
| Chapter | Title |
|---|---|
| 1 | The Coding Agent: What It Is and What It Isn't |
| 2 | The Problem of Context |
| 3 | Organizations, Capital, and the Division of Labor |
| 4 | Constitutional Constraints and Authority Hierarchies |
| 5 | The Orchestrator |
| 6 | The Navigator |
| 7 | The Security Agent |
| 8 | The Code-Hygiene Agent |
| 9 | The Conflict Auditor |
| 10 | Conflict Resolution |
| 11 | The Adversarial Agent |
| 12 | The Agent Updater |
| 13 | The Agent Refactorer |
| 14 | The Cleanup Agent |
| 15 | The Domain Agents |
| 16 | The Repo Liaison: Cross-Repository Governance |
| 17 | The Abstract Agent |
| 18 | The Agent Investigator |
| 19 | The Daily Pipeline |
| 20 | Investigation and Remediation |
| 21 | Drift, Contradiction, and Self-Repair |
| 22 | Agent Teams as Spontaneous Orders |
The manuscript reader supports sidebar navigation, chapter-to-chapter links, and in-text citation popups linked to the full bibliography.