You are here:Appendices IndexCourse Index][ Appendix A ][Workshops The main course follows a sequence. The appendices are reference material designed to be revisited.
Consult them when a term becomes unclear, a tool decision needs structure, you are preparing to delegate work to an AI agent, or an AI-generated answer looks plausible but wrong.
Together, the appendices provide the course’s working vocabulary, decision frameworks, checklists, and failure references.
Read the appendices as needed rather than working through all of them before starting the course.
Their purpose is to make professional AI use repeatable without adding process for its own sake.
AI discussions become messy quickly because many important terms are overloaded. “Agent” can mean a tool-using model, a scripted workflow, a coding assistant, or a product feature. “Memory” can mean conversation history, persistent product memory, retrieved documents, or context still available within the current session. “Reasoning” can refer to inference-time computation, product branding, explicit problem decomposition, or simply an answer that sounds coherent.
This appendix defines the course’s terms in practical, technical language. It provides a working vocabulary for people who need to discuss generative AI, LLMs, tools, retrieval, agents, and AI-assisted programming precisely, without relying on vendor language or academic shorthand.
Use it when
What it contains
Best read with
Tool selection becomes unreliable when people compare models through leaderboard screenshots, tools through social media momentum, and coding agents through the most impressive demonstration they have seen that week. Those signals can inform an evaluation, but they cannot replace requirements and evidence from the intended workflow.
This appendix provides a structured way to choose models and tools. It considers task shape, risk, data sensitivity, workflow fit, review burden, latency tolerance, cost profile, repository context, and operational overhead.
The framework is deliberately vendor-neutral so that it remains useful as products and model names change.
Use it when
What it contains
Best read with
Checklists become most valuable when work is moving quickly and shortcuts are tempting.
This appendix turns the course’s principles into practical review gates for technical work. Before asking AI to explain a codebase, change a file, carry out a refactor, generate tests, review a pull request, or act as an agent, establish clear answers to these questions:
These questions establish the boundary between supervised AI-assisted engineering and vibe-coding.
Use it when
What it contains
Best read with
Recurring AI failures are easier to manage when they have clear names.
These are recognizable failure patterns with causes that can be investigated and controls that can be improved.
This appendix catalogues common failures in AI-assisted technical work and provides a practical basis for diagnosing, supervising, correcting, and preventing them.
Use it when
What it contains
Best read with
Professional AI use depends on durable reference material outside any single conversation.
The course explains the ideas. The appendices help readers apply them after they close the article.
AI without illusions: Appendices was originally published in Stackademic on Medium, where people are continuing the conversation by highlighting and responding to this story.