Thoughtware
Studio

The primitive

The Cognitive Unit

A foundational theory of the smallest unit of intelligence that you can name, measure, price, trust and reuse, at scale.

Every codebase full of language-model calls already has the shape of a cognitive unit, but without a name, you cannot measure, share, or improve it. This book names the cognitive unit and works out what follows: open and closed decisions, purity, trust, cost, and composition into agents.

Who should read it

Builders shipping model-backed features who need checkable claims about the smallest piece of thinking.

Download PDF

Contents

  1. Preface

    Why This Book Exists

  2. Introduction

    Thoughtware, Agents, and Cognitive Units

  3. Part I · The Cognitive Unit

    What a Name Buys

  4. The Cognitive Unit

  5. The Principles

  6. Async Functions, and How a Cognitive Unit Fails Differently

  7. Part II · Decisions

    Open and Closed Decisions

  8. How Decisions Close

  9. Purity and the Five Levels

  10. The Shapes of Cognitive Units

  11. Part III · Trust

    Why You Cannot Read a Cognitive Unit

  12. Reliability Is Not a Number

  13. Sharing, Substitution, and Libraries

  14. Part IV · Economics

    Cost Per Decision

  15. Tuning by Wrapping

  16. Paying for Reliability, and Why Unit Cost Falls

  17. Part V · Composition

    Where the Cognitive Unit Ends and the Agent Begins

  18. Composing Agents from Cognitive Units

  19. The Library Is What Unlocks Autonomy

  20. Inside the Call

  21. Finding the Weak Decision, and How Systems Fail

  22. Adopting Cognitive Units in Existing Code

  23. The Principles, Earned

  24. Epilogue

    If the Word Survives