Learn how AI works,from the basics to building agents

They nest. Artificial intelligence is the broad field, including rule-based systems that never learn. Machine learning is the part that learns patterns from data. Deep learning is machine learning with many-layered neural networks. Generative AI produces new content — text, images, code — and today is built on deep learning. Most AI running in production is not generative.

A language model splits text into tokens and predicts the next token, one at a time, based on everything before it. It learns this by pre-training on a very large body of text, then further training on instructions and on feedback about which answers are better — the step that turns a text predictor into a usable assistant. It produces plausible text, which is not the same as verified fact.

Writing the input to a language model so it reliably produces the output you need. The techniques the major model providers all recommend are: clear and specific instructions, a role, the relevant context, one or two examples, clear structure, a defined output format, and breaking complex tasks into steps. Reasoning models need simpler prompts that state the goal rather than the steps.

Try them in that order. Improve the prompt first — it is the cheapest fix and the most common problem. Add retrieval (RAG) when the model lacks facts, especially private or current information. Consider fine-tuning only when a consistent style or behaviour cannot be held with good prompts at your volume, and only with an evaluation set to prove it helped.

Start from your language and the hard part of your problem. Graph frameworks such as LangGraph, Microsoft Agent Framework and Google ADK suit long-running, stateful work; LlamaIndex and Haystack suit retrieval-heavy work; a model vendor’s own SDK is the thinnest path on that vendor’s models. Many production systems use no framework and call the model API directly.

An agent loop is a single AI agent repeating its own work toward a goal: acting, observing the real result, reasoning about the gap, and retrying until a testable stop condition is met — for example, the test suite passing. Some practitioners call designing these loops “loop engineering”, an emerging term rather than a standard.

The hub is organised into 5 topics: AI Basics (The history of AI, the difference between AI, machine learning and generative AI, and a glossary of the terms that get used loosely); Large Language Models (How large language models are trained and produce answers, what tokens and context windows are, why they hallucinate, and how to choose one); Prompt Engineering (Prompt engineering from the fundamentals to few-shot and chain-of-thought prompting, prompt chaining, and the CO-STAR, CRISPE, RTF, RACE and RISEN frameworks); RAG & Fine-Tuning (When to use RAG or fine-tuning, and how retrieval-augmented generation works in depth: chunking, embeddings, hybrid search, reranking and evaluation); AI Agents (AI agents vs agentic AI, chatbots, copilots and RPA; how to build an agent with or without code; tool use, MCP and memory; loops, graphs and frameworks).