Experience-led AI & LLM literacy course

The Downward Journey

27 modules, built around a real condominium solar hot-water system conversation and a Kitchen Inventory Agent build. Start with what you already experience with AI Tool, then travel downward into the mechanics and mathematics that make it possible — stopping at whichever depth (Everyday, Curious, Technical, Mathematical) suits you on any given lesson.

27 modules · L00–L262 case studies · Condominium Solar System · Kitchen Inventory AgentNo backend · all state is local to your browser

Phase 1

Recognising What You Already Experience

L00

One module. Your own AI habits become the syllabus — every later lesson answers a question raised here.

L00The AI You Already Use“Why does AI Tool finish my sentence, change its answer when I change one word, or sometimes sound confidently wrong?”ABCNone mathEither10 min

Phase 2

From Words to Meaning

L01–L03

Text becomes tokens, tokens become numbers, and numbers get a way of being compared. The condominium vocabulary anchors every example.

L01From Words to Tokens“Why does a longer conversation with AI Tool cost more, and why does it sometimes cut a word in a strange place?”ABCLight mathCondo20 min L02How Words Become Numbers“Computers calculate with numbers — how can they understand 'boiler', 'solar', 'tank' or 'failure'?”ABCModerate mathCondo20 min L03Similarity and the Dot Product“If words are arrows on a meaning map, how might a system tell whether two of them point the same way?”ABCModerate mathCondo20 min

Phase 3

How the Model Computes

L04–L09

Neurons, training, next-token prediction, transformers, and attention — the mechanical core, built diagram-first and reused across two attention lessons.

L04Neural Networks“How does the system turn these plain numerical representations into something useful?”ABCModerate mathCondo25 min L05How a Network Learns“Who chooses the billions of weights inside a model like this?”ABCDeep mathCondo30 min L06Language Models“AI Tool wrote 'The solar system contains a 300 litre…' — what happens next inside the model?”ABCLight mathCondo20 min L07Transformers“If the model predicts one token at a time, how does it keep track of relationships across a sentence or document?”ABCModerate mathCondo25 min L08Attention“In '…a 300 L tank with two coils and a 2 kW electric immersion heater,' what actually matters when I ask how long heating will take?”ABCDeep mathCondo30 min L09Why Multiple Attention Heads?“The heater's power and the tank's volume are both relevant, but in different ways — can one attention pass capture that?”BCDeep mathCondo15 min

Phase 4

Context, Training & Reasoning

L10–L15

Where the condominium conversation's knowledge actually comes from, how memorization/generalization/reasoning differ, and a maths consolidation checkpoint with an opt-in deeper pathway.

L10Context“AI Tool remembered the tank size I mentioned three messages ago — is that the same as it 'knowing' about tanks in general?”ABCNone mathCondo20 min L11Where Did the Knowledge Come From?“How did the model know what an immersion heater or a solar coil was, when I never explained it?”ABCLight mathCondo20 min L12Post-Training: From Predictor to Assistant“If the model only learned to predict the next likely word, why doesn't it ramble forever, argue with itself, or answer a different question than the one I asked?”ABCNone mathCondo20 min L13Memorization vs Generalization vs Reasoning“AI Tool told me the tank will take about 6 hours to heat — did it look that up, guess from a pattern, or actually work it out?”ABCModerate mathCondo30 min L14Mathematics of Reasoning“I've now seen several small pieces of maths — how do they fit together, and where would I look for the full picture?”BCDeep mathCondo20–35 min, self-paced L15External Mathematics Pathway“I want to actually be able to derive this, not just follow along — where do I go?”BCDeep math—Variable, self-paced

Phase 5

Going Deeper, Going Beyond the Model

L16–L20

Five parallel extensions past the model itself: tools, retrieval, multimodal input, memory and data handling. Both case studies appear here, split by which one demonstrates the idea best.

L16Tools: When the Model Is Not Enough“I asked about current Portuguese regulations for this installation — and separately, I told my kitchen agent 'I used two eggs.' Neither answer came purely from what the model already knew.”ABCNone mathBoth20 min L17Retrieval and RAG“I uploaded a 100-page technical manual — does AI Tool have to permanently learn the whole thing to answer questions about it?”ABCLight mathCondo25 min L18Multimodal AI“How can AI Tool look at my photo of the boiler room and discuss what's actually in it?”ABCLight mathCondo20 min L19Memory“AI Tool remembered something about me from a completely different conversation last week — is that the same as it remembering what I said five minutes ago?”ABCNone mathEither15 min L20Data and Privacy: What Happens to What You Share“I uploaded photos of my boiler and typed details about my apartment into this chat — where does that go? Is it used to train the model? Who else can see it?”ABCNone mathCondo20 min

Phase 6

Trust, Systems & the Full Picture

L21–L26

Hallucination, prompting, the model/product distinction, agents, responsible use, and a two-lane capstone reconstructing both case studies end to end.

L21Hallucination“AI Tool said the tank would take exactly 6.1 hours to heat. Is the calculation wrong?”ABCLight mathCondo25 min L22Why Prompts Matter“'Tell me about boilers' and 'Analyse this specific system…' get very different answers — why, and how do I write the second kind on purpose?”ABCNone mathCondo20 min L23From Chatbot to AI System: The Model vs. the Product“When I chat with AI Tool, am I talking directly to 'the model', or to something built around it?”ABCNone mathKitchen25 min L24Agents“Instead of asking 'how do I log that I used two eggs', I just said 'I used two eggs' and it updated my inventory — what changed?”ABCNone mathKitchen25 min L25Trust and Responsible Use“When does an AI Tool answer need a second look before I act on it — and does that change when the AI is taking an action, not just answering?”ABCNone mathBoth20 min L26The Complete Journey“Looking back across both the condominium conversation and the kitchen agent build — what actually happened, start to finish?”ABCNone mathBoth25 min

Dependency map

The curriculum isn’t one straight line. It’s linear through Context (L10) — everything up to there builds directly on what came before. Past that point, five threads branch off in parallel (tools, multimodal input, memory/privacy, hallucination, and prompting) and only reconverge once the systems/agent view needs all of them. External Mathematics (L15) sits outside the sequence entirely — a side panel reachable from four specific points, not a numbered stop. Click any box to jump to its first lesson.

L00 →L01–L09 →L10 →L11–L14 →L15 →L16–L17 →L18 →L19–L20 →L22 →L21 →L23–L24 →L25 →L26 →