From Zero to Agents: Applied Artificial Intelligence
A complete programme to understand how AI works, use it thoughtfully in your work, and explore the frontier of autonomous agents. No unnecessary jargon, with clear order.
Foundations
What AI really is and how it works
Understand what lies behind the term artificial intelligence, how models learn, and what types of AI exist — with enough grounding to distinguish marketing from reality.
- 1.1Read →
What AI is: beyond the hype and the fear
Neither magic nor apocalypse. AI is a family of statistical techniques that learn patterns from data. Understanding this changes everything else.
- 1.2Read →
How machines learn: the concept of training
What actually happens when a model 'learns': data, parameters, loss and gradient. No formulas, just logic.
- 1.3Read →
Types of AI: the map nobody gave you
Supervised, unsupervised, reinforcement learning, generative. The full taxonomy with real examples for each type.
- 1.4Read →
Three waves and a revolution: a brief history of AI
From the first perceptrons to transformers. Why AI failed twice before it worked, and what changed in 2017.
Language Models
The revolution that changed everything
Understand what an LLM is, how it processes language, what context and memory mean, and why hallucinations are a structural consequence of the design — not a correctable bug.
- 2.1Read →
What an LLM is and how it predicts the next word
Large language models do not understand: they predict. The difference has enormous practical implications for knowing when to trust them.
- 2.2Read →
Tokens, context and memory: what AI can and cannot retain
A token is not a word. Context is not memory. Understanding these distinctions prevents half the mistakes people make when working with AI.
- 2.3Read →
Alignment: how AI is taught to be helpful and safe
RLHF, instruction tuning and values. How you go from a model that predicts text to one that follows instructions without causing harm.
- 2.4Read →
Why AI hallucinates (and how to catch it before it matters)
Hallucinations are not technical glitches: they are consequences of the prediction mechanism. Warning signs and verification strategies.
The Art of the Prompt
Communicating well with AI
Develop the ability to write clear instructions, apply advanced prompting techniques, and refine results through dialogue rather than hoping AI will guess what you want.
- 3.1Read →
The principles of effective prompting
Role, context, task and format. The four elements of a prompt that works, with comparative examples.
- 3.2Read →
Advanced techniques: chain-of-thought, few-shot and role prompting
Three techniques that multiply output quality. When to use each and how to combine them for complex problems.
- 3.3Read →
Structuring outputs: get exactly what you need
Asking for JSON, tables, numbered lists or specific formats. How to stop AI from improvising structure when you need precision.
- 3.4Read →
Iterate and refine: dialogue as a working method
The first response is rarely the best. How to build on what AI produces to get closer to the desired result without starting over.
AI at Work
Real cases, concrete workflows
Identify where AI generates real value in daily work, build workflows with concrete tools, and distinguish useful automation from apparent automation.
- 4.1Read →
The tools map: text, image, audio and code
The AI ecosystem in 2025–2026. Which tool is best for which task, and why using one well-chosen tool beats five mediocre ones.
- 4.2Read →
AI for writing: from draft to edit
AI does not write for you: it accelerates the parts that consume the most time. How to use it at each stage without losing your voice.
- 4.3Read →
AI for research, summarising and synthesis
How to process large volumes of information with AI without swallowing the errors. The verification method that saves you surprises.
- 4.4Read →
Building your personal AI workflow
From a list of tools to a system. How to integrate AI into real work without the tool becoming another source of friction.
Beyond Text
Multimodal AI, images and code
Understand how generative AI works beyond text — images, audio, video and code — and learn to connect models with your own data using RAG.
- 5.1Read →
Generative images: how diffusion models work
Stable Diffusion, Midjourney, DALL·E. The logic of reverse noise that turns text into images, and why visual prompting is different.
- 5.2Read →
Audio, voice and video: the current state and its limits
Transcription, voice synthesis, cloning and video generation. What works, what fails and what raises serious questions.
- 5.3Read →
Coding with AI: programming without knowing how (and better if you do)
Copilot, Claude, Cursor. How AI changes software development for experts and non-experts alike, and where the real limits are.
- 5.4Read →
RAG: connecting AI to your own documents
Retrieval-Augmented Generation. How to make a model answer questions about your internal documentation without retraining it or sharing sensitive data.
AI Agents
The current frontier
Understand what makes an AI agent different from a chatbot, how it is built with tools, memory and reasoning, and when it makes sense — and when it does not — to trust it with a task.
- 6.1Read →
What an AI agent is: from chatbot to autonomous actor
A chatbot responds. An agent decides, acts and observes the result. The architectural difference that makes it possible — and the risks it introduces.
- 6.2Read →
Tools, memory and reasoning: the three pillars of an agent
How an agent uses external tools, maintains context between actions and chains reasoning to complete complex tasks.
- 6.3Read →
Agents in practice: when they work and when they fail
Agents are fragile in real-world environments. Cases where they add value, common design mistakes and how to supervise them without losing control.
AI with Judgement
Ethics, limits and future vision
Develop your own framework for using AI responsibly, understand its implications for privacy and work, and build an informed view of the future that depends neither on naive optimism nor unfounded fear.
- 7.1Read →
Biases and structural limits: what models do not do well
Biases are not accidents: they reflect training data. How to identify them, when they matter and what to do about it.
- 7.2Read →
Privacy and security: what goes in, what comes out, what stays
What providers do with your conversations. How to work with sensitive information without compromising your own data or that of third parties.
- 7.3Read →
The future of work with AI: what changes, what remains
Final reflection. AI does not replace people: it replaces tasks. How to build a productive relationship with the technology without depending on it blindly.