Chapter 37 Further Reading: Custom GPTs, Assistants, and Configured AI Systems

Resources are organized by theme. API documentation URLs in particular should be treated as starting points — verify current URLs at the provider's documentation home page.

Official Platform Documentation

OpenAI GPT Builder Guide https://help.openai.com/en/articles/8554397-creating-a-gpt

OpenAI's official guide to the GPT Builder interface, covering the configuration options, knowledge file formats, actions setup, and publishing. Updated as features change — the canonical reference for current capabilities and limitations.

OpenAI Assistants API Documentation https://platform.openai.com/docs/assistants/overview

Comprehensive documentation for the Assistants API, including thread management, run lifecycle, tool use (file search, code interpreter, function calling), and best practices. The quickstart guide is the fastest path to a working implementation.

OpenAI Assistants API Cookbook https://cookbook.openai.com/examples/assistants_api_overview_python

Worked Python examples demonstrating Assistants API patterns, including file search integration, multi-turn conversation management, and function calling. More practical than the reference documentation.

Anthropic Projects Documentation https://docs.anthropic.com/en/docs/about-claude/projects

Anthropic's official documentation for Claude Projects, covering setup, document upload, and usage best practices. Check this page regularly as Projects features are actively developing.

Anthropic System Prompts Documentation https://docs.anthropic.com/en/docs/build-with-claude/system-prompts

Anthropic's guidance on writing effective system prompts, including best practices specific to Claude's capabilities and how to structure instructions for reliable behavior.

System Prompt Design

"The Prompt Engineering Guide" (Comprehensive) https://www.promptingguide.ai

A community-maintained, frequently updated guide to prompt engineering techniques. The sections on role prompting, chain-of-thought, and few-shot prompting are directly applicable to system prompt design for configured systems. Free and open source.

Anthropic's Prompt Engineering Guide https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview

Anthropic's official prompt engineering guidance, with Claude-specific insights about how Claude interprets instructions, handles ambiguity, and responds to behavioral guidelines. The "be clear and direct" and "use XML tags" guidance is particularly relevant for system prompt construction.

"Prompting Is Programming" — Simon Willison's Blog https://simonwillison.net

Simon Willison is one of the most thoughtful practitioners writing about AI system design. His blog includes extensive analysis of prompt injection, system prompt design, and the practical challenges of building reliable AI-powered tools. The posts on prompt injection attacks are essential reading before deploying any configured system that will accept user-provided input.

Knowledge Management and Retrieval

"Working in Public: The Making and Maintenance of Open Source Software" by Nadia Eghbal While not AI-specific, Eghbal's analysis of how knowledge is documented, maintained, and made useful in collaborative contexts is directly applicable to knowledge base design for configured AI systems. The tension between comprehensiveness and accessibility maps directly to the challenges of knowledge file design.

"Building a Second Brain" by Tiago Forte Forte's personal knowledge management system (PARA — Projects, Areas, Resources, Archives) provides a principled framework for organizing the kind of contextual knowledge that goes into configured AI systems. The "progressive summarization" technique for distilling documents into retrievable formats is particularly relevant.

Qdrant Documentation: Vector Search for RAG Systems https://qdrant.tech/documentation

For practitioners who want to go beyond built-in knowledge file retrieval, Retrieval-Augmented Generation (RAG) systems allow more sophisticated document search with custom chunking, embedding, and retrieval logic. Qdrant is an accessible starting point for understanding how semantic search works — which helps you design better knowledge files even if you do not implement custom RAG.

Design and User Experience

"Don't Make Me Think" by Steve Krug Krug's classic UX book is directly applicable to configured AI system design. His principles for interface clarity — reducing cognitive load, making the right action obvious, designing for how people actually use things rather than how you expect them to — apply to AI system design as readily as to web design. The user testing protocol in chapter 9 is a model for the user simulation testing step recommended in this chapter.

"The Design of Everyday Things" by Don Norman Norman's framework of affordances, feedback, and mental models maps well to configured AI system design. An AI system that does not clearly signal what it can do (affordances) or explain why it declined a request (feedback) will be used poorly regardless of how well it is configured. This book develops the design vocabulary for thinking clearly about these problems.

Governance and Deployment

"Thinking in Systems: A Primer" by Donella Meadows For practitioners who will deploy configured AI systems to teams or organizations, Meadows' introduction to systems thinking provides conceptual tools for anticipating how systems behave in practice — including the feedback loops and unintended consequences that make organizational AI deployments succeed or fail. A short book with lasting relevance.

NIST AI Risk Management Framework https://airc.nist.gov/Home

The U.S. National Institute of Standards and Technology's framework for AI risk management. For practitioners in regulated industries deploying AI tools to teams or customers, the AI RMF provides a structured vocabulary and process for identifying, assessing, and managing AI-related risks. More formally applicable to enterprise deployments than to individual practice, but useful as a vocabulary source for anyone thinking about AI governance.

Case Studies and Practitioner Examples

OpenAI Usage Policies https://openai.com/policies/usage-policies

Understanding what Custom GPTs cannot do (prohibited use cases) is as important as understanding what they can do. Read the usage policies before deploying any Custom GPT, particularly ones that will be shared publicly or used in customer-facing contexts.

Anthropic Usage Policy https://www.anthropic.com/legal/aup

The equivalent for Claude and Claude Projects. Covers what configured Claude deployments may and may not be used for.

"Real-World AI Case Studies" — Various Sources Search for case studies on your specific industry or use case on Google Scholar, LinkedIn, and professional association publications. The most valuable practitioner learning comes from people in your own field who have deployed configured AI systems for similar purposes. Industry-specific communities (legal tech, marketing technology, software engineering) often have the most directly applicable examples.