liveBook · Manning

What you will learn in Self-Improving Agents

How to engineer adaptive harnesses

by Micheal Lanham · Manning Early Access Program (MEAP)

Chapter by chapter

Chapter 1: Foundations of self-improving agents

The introduction explains why production agents need ongoing improvement and presents a self-run improvement loop that does not retrain the underlying model.

1.1 What a self-improving agent is · 1.1.1 Defining the agent and the scaffolding · 1.1.2 The self-improving agent · 1.1.3 Why the model never changes · 1.2 The improvement loop · 1.2.1 The loop, walked through one task · 1.2.2 The three dials · 1.2.3 Online and offline: where the loop runs · 1.3 The four kinds of things the loop can improve · 1.3.1 What an artifact is · 1.3.2 The four layers and the toolkit that runs them · 1.4 The signal and search families behind the loop · 1.4.1 The signal family · 1.4.2 The search family · 1.5 Summary

What you learn in chapter 1 →

Chapter 2: Getting started improving agents

The chapter introduces an offline improvement loop for a RAG agent’s prompt instruction and previews the tools used to build it.

2.1 Building the agent and text prompt artifact · 2.1.1 Introducing the RAG agent · 2.1.2 Running the code and picking LLMs · 2.1.3 Discovering layer 0 – the helix framework · 2.1.4 Connecting the prompt artifact archive · 2.2 Measuring the signal · 2.2.1 Establishing what good is · 2.2.2 Producing the signal gap measurement · 2.2.3 Implementing LLM-as-Judge for signal measurement · 2.3 Searching for artifact variants · 2.3.1 Selecting the search method · 2.3.2 Search cost versus reach · 2.3.3 Inside the search · 2.3.4 Understanding SPO (Self-Supervised Prompt Optimization) · 2.3.5 Implementing SPO search · 2.4 Putting it together with Helix · 2.4.1 Understanding the helix components · 2.4.2 The OfflineImprover · 2.4.3 The ImproverPolicy · 2.4.4 The code · 2.5 Exercises · 2.6 Summary

What you learn in chapter 2 →

Chapter 3: The Gödel Machine, and the importance of signal

The introduction contrasts the Gödel machine’s proof-based self-improvement with the practical approach of measuring changes and searching through candidates.

3.1 Measuring the gap · 3.1.1 The signal step of self-improvement · 3.1.2 The Gödel machine and the proof gate · 3.1.3 Why we measure instead of prove · 3.1.4 The GapMeasurement union: one shape for every signal · 3.2 The agent revenue task · 3.2.1 One vague question, six hidden rules · 3.2.2 Ground truth: precise, and only as good as the checker · 3.2.3 The proof versus ground truth · 3.3 The family of signals · 3.3.1 Reflection: a critique with no score · 3.3.2 The signal ceiling · 3.4 Applying signals to improvement · 3.4.1 Consuming signals with search · 3.4.2 Hill-climbing: the empirical Gödel ratchet · 3.4.3 SPO: climbing without a ground truth · 3.4.4 Two climbers, one wall · 3.5 Exercises · 3.6 Summary

What you learn in chapter 3 →

Chapter 4: Darwin Gödel Machine, improving search

The introduction frames evolutionary search as a way to improve candidate generation beyond a simple hill-climber, while noting its added cost.

4.1 The oldest search algorithm · 4.1.1 Vary, select, remember: natural selection in a tiny script · 4.1.2 From fitness to Signal: evolution in the framework's words · 4.1.3 Why a population beats a single climber · 4.2 GEPA: a population that reads its own failures · 4.2.1 Introducing Genetic-Pareto evolutionary search · 4.2.2 Reflection: mutating from the whole run, not just the outcome · 4.2.3 The Pareto front: keeping the best-at-something · 4.2.4 Population, crossover, and the climb · 4.3 DGM: the archive that never forgets · 4.3.1 The Darwin Gödel machine · 4.3.2 Proof relaxed to evidence · 4.3.3 Quality-diversity: why you go back into the archive · 4.3.4 A deceptive landscape: greedy sticks, an archive crosses · 4.3.5 The recovery · 4.4 Seeing the evolution: the lineage tree · 4.4.1 The Helix Observatory: running the dashboard · 4.4.2 The Generations view: watching selection happen · 4.4.3 Reading GEPA against DGM, generation by generation · 4.5 Exercises · 4.6 Summary

What you learn in chapter 4 →

Concepts explained in this book

These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.