| name | exploring-generative-ai |
| description | Use when evaluating coding assistants, agent autonomy, AI-assisted development workflows, context engineering, local coding-model viability, or software-engineering trade-offs using Martin Fowler's Exploring Generative AI series. |
Exploring Generative AI
Series: Martin Fowler, Exploring Generative AI
Sources: 29 official articles, published 2023-07-26 through 2026-07-08
How to Use
- Start with the Topic Index for a decision; then read every named chapter before making an article-specific claim.
- Use
cheatsheet.md for a fast risk, autonomy, context, or local-model decision; read its cited chapters before acting.
- Use
patterns.md when designing a workflow or harness; use glossary.md when terminology is unclear.
- Treat each chapter's source URL, author, and date as its authoritative metadata. Say a claim is not covered when no chapter supports it.
Core Decision Rules
| Signal | Default action | Escalate when | Read |
|---|
| Output is plausible but unverified | Add the fastest trustworthy feedback: compiler, test, linter, review, or observable artifact. | Feedback is slow, ambiguous, or unavailable. | ch06, ch08, ch12 |
| Probability of error is high, impact is high, or detectability is low | Do not use unsupervised execution; reduce scope and require human review. | A public API, security boundary, data change, or production behavior is involved. | ch16, ch19, ch21, ch27 |
| Context is stale, broad, or causing guesses | Refresh the smallest relevant reference, specification, and workspace evidence; remove stale noise. | The agent cannot reliably decide what context to load. | ch09, ch22, ch25 |
| Repeated agent misstep has a growing impact radius | Encode the missing judgment or constraint in a reusable skill or harness control. | It reaches team iteration or codebase-lifetime impact. | ch13, ch24, ch26 |
| Considering a local model | Test the actual harness in sequence: memory fit, speed, tool calling, correctness, context, task size, and review burden. | Any earlier stage fails. | ch28, ch29 |
Chapter Index
| # | Article | Author |
|---|
| 1 | The toolchain | Birgitta Böckeler |
| 2 | Median - A tale in three functions | Birgitta Böckeler |
| 3 | In-line assistance - when is it more useful? | Birgitta Böckeler |
| 4 | In-line assistance - how can it get in the way? | Birgitta Böckeler |
| 5 | Coding assistants do not replace pair programming | Birgitta Böckeler |
| 6 | TDD with GitHub Copilot | Paul Sobocinski |
| 7 | How is GenAI different from other code generators? | Birgitta Böckeler |
| 8 | How to tackle unreliability of coding assistants | Birgitta Böckeler |
| 9 | Onboarding to a 'legacy' codebase with the help of AI | Birgitta Böckeler |
| 10 | Building an AI agent application to migrate a tech stack | Birgitta Böckeler |
| 11 | Expanding the solution size with multi-file editing | Birgitta Böckeler |
| 12 | What role does LLM reasoning play for software tasks? | Birgitta Böckeler |
| 13 | The role of developer skills in agentic coding | Birgitta Böckeler |
| 14 | Guiding an LLM for Robust Java ByteBuffer Code | Unmesh Joshi |
| 15 | Building TMT Mirror Visualization with LLM: A Step-by-Step Journey | Unmesh Joshi |
| 16 | Coding Assistants Threaten the Software Supply Chain | Jim Gumbley、Lilly Ryan |
| 17 | Building Custom Tooling with LLMs | Unmesh Joshi |
| 18 | Autonomous coding agents: A Codex example | Birgitta Böckeler |
| 19 | I still care about the code | Birgitta Böckeler |
| 20 | Partner with the AI, throw away the code | Matteo Vaccari |
| 21 | To vibe or not to vibe | Birgitta Böckeler |
| 22 | Anchoring AI to a reference application | Birgitta Böckeler |
| 23 | Understanding Spec-Driven-Development: Kiro, spec-kit, and Tessl | Birgitta Böckeler |
| 24 | Assessing internal quality while coding with an agent | Erik Doernenburg |
| 25 | Context Engineering for Coding Agents | Birgitta Böckeler |
| 26 | Harness Engineering - first thoughts | Birgitta Böckeler |
| 27 | Humans and Agents in Software Engineering Loops | Kief Morris |
| 28 | Viability of local models for coding | Birgitta Böckeler |
| 29 | Experiences with local models for coding | Birgitta Böckeler |
Topic Index
- AI output may be wrong / unreliable: ch04, ch06, ch08, ch12
- agentic coding / harness engineering: ch13, ch18, ch24, ch26, ch27
- autonomy, oversight, or high-risk execution: ch16, ch18, ch19, ch21, ch27
- code quality and review complacency: ch04, ch05, ch19, ch24
- context drift / reference application / context loading: ch09, ch22, ch25
- local-model viability: ch28, ch29
- prototype versus production: ch20, ch21, ch24
- risk assessment / vibe coding: ch19, ch21
- security and supply chain: ch16
- spec-driven development: ch23
- TDD: ch06, ch14
Supporting Files
- glossary.md — precise terms and source chapters.
- patterns.md — workflow patterns; read before designing a harness or repeated agent process.
- cheatsheet.md — decision tables; read before choosing autonomy, handling context drift, or evaluating a local model.
Scope and Limits
This skill covers only the 29 official URLs listed in the chapter files. It does not treat the series as a single-author book: chapter attribution remains article-specific. Do not infer claims beyond the cited article.
