How This Book Works
This book borrows its structure from Douglas Hofstadter's Gödel, Escher, Bach (1979), in which playful dialogues alternate with formal chapters. Each dialogue introduces a theme through story, scene, or conversation; the chapter that follows formalises the same theme as argument.
The dialogues are not decoration. They are the other half of the argument.
In Hofstadter's book, Achilles and the Tortoise have philosophical conversations that lead into formal chapters on mathematics, logic, and self-reference. The connection is not obvious on first reading — it becomes obvious on reflection. That gap is the point. The dialogue lets you feel the idea before the chapter asks you to think it.
The Characters
AlgoRhythmics has three recurring characters:
Alice brings the world. She notices the gap between what a system measures and what actually matters — the carer who knows her residents by name, the allotment lost to a rate model, the book found on the wrong shelf. She is specific, observant, and relentless about detail.
Bob is the pragmatist. He works closer to numbers and sees their value. He is not a villain — he's the reasonable person who trusts the dashboard because the dashboard is usually right. His journey across the dialogues is from defending the system to feeling the weight of what it misses.
The Agent is an AI. It is never named. In the early dialogues, it gives clean, narrow, technically correct answers — the kind you'd expect from a system optimising inside a default frame. In Dialogue 7 (Ten Boats), Alice gives it wider context, and its response changes completely. From that point on, the Agent starts asking questions back, surfacing trade-offs, and identifying which objective function a proposed solution actually serves. It does not become a different system. It receives wider context, and produces wider output. That's the thesis.
The Arc
The dialogues have a three-act structure:
- Dialogues 1–6 — The Gap. Alice sees something the system misses. The Agent gives default answers. Bob thinks the answers are fine.
- Dialogue 7 — The Turn. Alice gives the Agent the full context of a problem — all the competing objective functions, not just one. The Agent's response is fundamentally different.
- Dialogues 8–12 — The Widening. Alice and Bob start giving the Agent context up front. Its answers deepen. Bob begins to feel the pattern himself. In Dialogue 12, the Agent synthesises everything it has learned across all the conversations — and Bob says the words that close the book: "Context is harmony."
The Pairings
| Dialogue | Chapter | The shared idea |
|---|---|---|
| The Count | 1. The First Compression | Quantification strips context |
| The Dashboard | 2. Counting as Target Capture | Metrics displace the thing they measure |
| The Promise | 3. Debt and Exponential Reinforcement | Macro optimisation cascades into micro loss |
| The Shortcut | 4. Local vs Global Optimisation | Short-term metrics hide structural risk |
| The Scroll | 5. The Addiction Function | High-frequency feedback displaces deep satisfaction |
| The Bookshelf | 6. Synthetic Serendipity | Optimised recommendation kills discovery |
| Ten Boats | 7. The Game Theory Trap | Rational agents collectively destroy the commons |
| The Machine That Eats Itself | 8. Companies as Algorithms | Tacit knowledge has no line in the spreadsheet |
| The Soil | 9. When Optimisation Consumes Its Host | Yield optimisation depletes the substrate |
| The Parent | 10. Humans as Meta-Optimisers | Measurement captures a fraction of human capacity |
| The Meeting | 11. Cognitive Compression | Coordination overhead colonises thinking time |
| The Seed | 12. The Miracle Engine | Natural systems balance objectives without maximising any |
Read the dialogues as stories. Read the chapters as theory. Together, they make the case that single-variable optimisation is compressing human context — and that the fix is not to abandon optimisation, but to widen it.