ARQIV / AI topics

One place to understand, try and question AI

Every explainer, lab, game and vocabulary entry on ARQIV, organized by what it helps you understand. Start from zero, look up one topic, or go deeper. Nothing needs an account.

All ages57 pagesUpdated October 2026

The foundations route

Nine stops, in order

Each stop answers one question with an example you can check, a small activity and three questions at the end. Go in order or jump in.

  1. 01Map of AIIs every AI a chatbot?
  2. 02Machine learningHow can examples become a rule?
  3. 03DataWhat did we teach it by accident?
  4. 04Numbers & tensorsHow does a picture become numbers?
  5. 05Training vs usingLearning now, or using what it learned?
  6. 06Evaluating AIGood at what, for whom, compared with what?
  7. 07Classical modelsDo you always need a neural network?
  8. 08Finding groupsCan a computer find groups nobody labeled?
  9. 09Retrieval & RAGHow can an AI answer from documents it was given?

Stops 07–09 are optional depth. The five-stop AI journey (how AI works, transformers, history, today, what’s next) is still the quickest introduction to modern AI.

Directory

Everything, by topic

Filter by format if you are looking for something to do rather than read.

Start here

The big picture, and the words you will keep meeting.

Data and numbers

What AI learns from, and how pictures and words become numbers.

How machines learn

Learning settings, training, testing and the models beyond neural networks.

Language, media and agents

Chatbots, retrieval, creative AI and AI that uses tools.

AI in the world

The same ideas at work in science, farming, health and robots.

Use it, question it, teach it

Checking answers, testing models and classroom guides.

Advanced track (optional)

For older students and adults who finished the foundations. Each page lists what to read first.

Vocabulary

Stuck on a word?

One shared vocabulary serves the whole site. On ARQIV pages, AI words have a dotted underline: tap one for a short definition.

The full list

AI words in plain language

Every term with a short definition, a fuller explanation, a line to say to students and where to see it on ARQIV.

Open the vocabulary →

Learning paths

Routes for different readers

Proposed sequences to try. They have not yet been tested in classrooms; tell us what works.

Suggested learning routes (proposals to pilot, not proven sequences)
If you are…Try this routeYou’ll know it worked when you can…
Brand new to AIMap of AI → Next Word lab → Talking to AI → Verify It → one “AI in the world” pagesay whether a tool is predicting text, sorting things or taking actions, and spot when a confident answer needs checking
Using chatbots for school, work or daily lifeTalking to AI → Careful What You Wish For → Verify It → How Sure Are You? → Retrieval and RAGwrite a clear request, notice when a question pushes toward one answer, and know what never to paste into a chat box
Teaching grades K–2Teacher-led sorting with real objects (“did a person write the rule, or did it learn from examples?”), then the classroom starter packgive a reason for a grouping and say when more information is needed (no tensor words needed)
Teaching grades 3–5Train the Machine and Next Word → the grades 3–5 activity on Map of AIexplain why a rule that fits practice examples can fail on a new one
Teaching grades 6–8Map → Data → Machine learning → Tensors (visual part) → Evaluating AIdraw a pipeline and read the shape of a small image array
Grades 9–12, or adults who want the inner workingsMap of AI → Machine learning → Data → Training vs using → Transformers → Evaluating AI → Verify Itexplain how examples become a trained model, how bias gets in through the data, and why one accuracy number can mislead
Curious how chatbots and agents work insideMachine learning → Transformers → Embeddings → Retrieval and RAG → Prompting → Agents and permissions → How agents are builttell apart what a model learned in training, what is in its current context and what it retrieved
A teacher preparing a lessonPick an explainer, use its predict-first questions and lab in class, then its three-question check as an exit ticket. Classroom-ready lessons: For educatorsrun a complete activity with clear preparation and device needs
Building something yourselfBuild a small model → robust splits → evaluation → monitoring over time → responsible userun a reproducible experiment with a baseline, a fair split, a model card and a plan for when it goes wrong

About this map

How this map is organized. Topics are grouped by what they help you understand, not by product or brand. Some pages are dated (model rankings, research news); the explainers aim to stay true for years and say when they were last checked. Something missing or wrong? Tell us.