OpenEXA Request access

The thinking behind the swarm.

Ten notes on why high-stakes work should be run by thousands of narrow agents behind one deterministic boundary.

Research notes

  1. 01

    Compounding Error and the Case for Decomposition. A Research Note on Agent Architecture

    6 min
  2. 02

    Specializing a Model to a Rulebook and Proving What It Did. A Research Note on Post-Training and Audit

    7 min

The series

  1. 03

    Not Tasks. Lifecycles. Why We Stopped Building Assistants

    3 min
  2. 04

    Why Five Thousand Small Agents Beat One Large One

    3 min
  3. 05

    Agents Decide. Code Executes. The One Boundary That Matters

    3 min
  4. 06

    Post-Training a Model on a Rulebook, Not on the Internet

    3 min
  5. 07

    The Execution Council. Governance in Milliseconds

    3 min
  6. 08

    A Ledger Nobody Can Edit. Hash Chains as Audit Infrastructure

    3 min
  7. 09

    Lifecycle 01. What Ten Sessions on Real Markets Taught Us

    3 min
  8. 10

    Master and Copy. How a Proven Agent Becomes a Platform

    3 min

Three collections, from market structure to AI agents.

All 29 posts The papers on Google Scholar

Three of these ideas run in your browser.

Open the swarm lab Change a record Run the desk