SIAP
Structure and Interpretation of Agentic Patterns: a book in progress that teaches agent systems by building interpreters for reasoning and tool-use loops.
- Status
- active · manuscript in progress
- Role
- Author. I write the manuscript and maintain its build scripts.
- Stack
- agentic-systems, technical-writing, decision-theory, multi-agent-systems
- Repository
- Source code
- Dates
Problem
Agent engineering is often held together by framework defaults and prompt folklore, leaving objectives, memory, tool contracts, and failure modes under-specified.
What I built
Markdown chapters connect decision theory, planning, retrieval, safety, and multi-agent coordination to small interpreters. Python scripts build a static web reader and PDF, and audit citations and visuals.
Evidence
- 4 parts: Foundations of Agency, Reasoning Architectures, Tool Use and Embodiment, Multi-Agent Systems
- 36 chapter files, about 57,000 words of draft text
- Web reader and PDF built by the repository's scripts
One hard decision
Teach by construction: each pattern is introduced as a small interpreter the reader can run and break, rather than as a framework recipe.
Where it stands
The draft is readable online and the build checks citation and visual rules on every change.
Limits
This is an unfinished draft with no external review or publisher; chapters vary in completeness and the interpreters have not been evaluated with readers.