Learn AI Concepts

A practical guide to modern AI concepts — how to use them, not the math behind them. Each page explains one concept well enough that you can actually build with it: what it is, why it exists, when to reach for it, and when not to.

AI Concepts — every concept on the site, grouped by category.
Compare AI Concepts — head-to-head answers for concepts that are easy to confuse.
Interview questions — practice questions grouped by topic, linked back to the full explanation.

Browse by category

AI Essentials — The basic building blocks worth understanding before the rest of this site's agent, tool, and retrieval concepts make full sense.
RAG — Retrieval-augmented generation grounds a model's answers in your own data at query time, instead of relying only on what it learned during training.
AI Agents — An AI agent decides its own next action instead of following a fixed script. What that means in practice, and what changes once more than one agent is involved.
AI Safety — The practical end of AI safety: how applications built on models actually get attacked, and how to limit the damage when something gets through.
Prompting — How to actually instruct a model, what does and doesn't change the way it responds, and what belongs in its context at all.
Tools — The mechanisms that let a model do something — take an action, look something up — instead of only answering in text.
AI Evaluation — How to tell whether an AI system is actually working, rather than just looking like it is, and how to keep telling once anything about it changes.
MCP — MCP standardizes how an AI application connects to external tools and data, instead of every pairing needing its own custom integration code.
Agent Skills — A skill packages instructions for one kind of task, kept out of an agent's prompt until that task actually comes up.
Context Engineering — What information a model actually sees for a given task, and how to decide what belongs in there and what doesn't.
Memory — What an AI system should deliberately remember between sessions, and what should stay temporary context instead.
Production AI — Running AI systems for real, once actual users depend on them — cost, latency, monitoring, and what breaks at scale.
Subagents — A subagent is the same kind of system as an ordinary agent — the difference is who's calling it, and what that changes about how you'd design it.

Concepts that are easy to confuse

The short answer for each, with the full comparison a click away.

Comparison	The short answer
AI Agent vs Automation	automation runs steps a person wrote in advance.
Agent vs Agentic AI	an AI agent is one specific point on a scale — full autonomy.
Agent vs Assistant	"Assistant" describes who a product is for.
Agent vs Chatbot	a chatbot replies to your message and waits.
Agent vs Subagent	an agent and a subagent are the same kind of system.
Agent vs Workflow	use a workflow when you can write out the exact steps in advance.
MCP Tools vs Resources	an MCP tool is a real action the AI can take.
MCP vs A2A	they're not competitors.
MCP vs API	MCP doesn't replace an API.
MCP vs Function Calling	they're not alternatives.
Prompt vs Context Engineering	prompt engineering is wording one instruction well.
RAG vs Fine-Tuning	fine-tuning changes how a model behaves.
RAG vs Long Context	a bigger context window shrinks the case for RAG on small knowledge bases.
RAG vs Memory	RAG retrieves from a general document collection to answer a question.
Single Agent vs Multi-Agent	a single agent handles a task alone.
Skill vs Agent	a skill has no decision-making of its own — it's instructions an agent chooses to load.
Skill vs Subagent	a skill is instructions loaded into the same agent's context.
Skill vs Tool	a skill is instructions that tell an agent how to approach a task.
Tool Calling vs Function Calling	same underlying mechanism, two names.
Vector Search vs Keyword Search	vector search matches by meaning, even with different wording.

Practice interview questions

Scenario questions rather than definitions — the kind that can't be bluffed. Every answer links back to the concept it comes from.

Broader revision passes by role and topic are listed under all interview topics.

Where to start

If you're starting from scratch, this order builds on itself — each one assumes the one before it.

LLM — what a large language model actually does, and why that single mechanism explains most of the rest.
Token — the unit a model reads and generates, and what gets measured in them.
Context Window — how much a model can see at once, and why that's a hard limit.
Prompt Engineering — how to actually instruct a model.
RAG — giving a model information it was never trained on.
Tool Calling — letting a model take an action instead of only answering.
AI Agent — what changes once the model decides its own next step.
MCP — the standard way to connect an application to outside tools and data.

No required order beyond that — browse everything if you already know what you're looking for.