Free · No login · 36 lessons · Updated Nov 2026
Learn AI, from your first prompt to building agents
Short lessons in a clear order: what LLMs and transformers are, how to get better answers, how APIs, RAG, MCP and agents work, then real automations and coding with n8n, Claude Code, Cursor and Codex. Each lesson has the best free videos, a hands-on task and a quick check.
- Lessons
- 36
- 36 planned, more added often
- Tracks
- 5
- Beginner to advanced
- Time
- ~7 h
- Reading and doing, videos extra
- Cost
- Free
- No login, progress saved here
Track 1 of 5 · 9 lessons · 99 min
Understand AI
Explain what AI, LLMs and transformers are, and how ChatGPT and Claude were built.
- 1What is AI?8 minArtificial intelligence (AI) is technology that lets computers do things we link with human thinking, like understanding language, spotting patterns and making decisions. Machine learning is the part of AI where computers learn from examples instead of hand-written rules. Deep learning is machine learning with large neural networks, and generative AI, like ChatGPT, Claude and Gemini, is deep learning that creates new text, images or code.
- 2How AI started: 1956 to ChatGPT10 minAI got its name at a 1956 summer workshop at Dartmouth College, then went through boom and bust cycles called AI winters. Deep learning took off in 2012, Google’s 2017 transformer paper made today’s language models possible, and ChatGPT (November 2022) brought AI to everyone. Since 2024 the focus has moved to reasoning models and AI agents that take actions.
- 3What is an LLM?10 minA large language model (LLM) is an AI model that has read a huge amount of text and learned to predict the next piece of text, one small chunk at a time. ChatGPT, Claude and Gemini are all built on LLMs. By repeating that one simple step thousands of times, an LLM can write essays, answer questions and write code, though it can also confidently make things up.
- 4Tokens and context windows10 minLLMs do not read words. They read tokens, small chunks of text that are often a whole word or part of one; in English, 100 tokens is about 75 words. The context window is the maximum number of tokens a model can look at at once, including your messages and its replies. When a chat gets too long, older parts get dropped or summarised, which is why AI seems to “forget”, and every token you send or receive also costs money on an API.
- 5Transformers and attention11 minA transformer is the neural network design behind ChatGPT, Claude and Gemini, introduced by a Google team in the 2017 paper “Attention Is All You Need”. Its key idea, attention, lets every word look at every other word at once and decide which ones matter for its meaning. Because it processes all words in parallel instead of one by one, it trains fast on GPUs and scales to huge models.
- 6The papers that built modern AI14 minModern AI chatbots rest on a handful of research papers published between 2017 and 2022. The transformer (2017) gave us the engine, the GPT papers showed that predicting the next word on lots of text teaches a model many skills, scaling laws and Chinchilla showed how big to build it, InstructGPT turned it into an assistant, chain of thought taught it to show its working, and Constitutional AI trained it with written principles. Each one explains something you see in ChatGPT or Claude today.
- 7How ChatGPT and Claude are trained12 minLLMs like ChatGPT and Claude are trained in stages. First, pretraining on huge amounts of text creates a base model that can only continue text. Then instruction tuning on example conversations and reinforcement learning from feedback (from humans in RLHF, or from AI guided by written principles in Anthropic’s Constitutional AI) turn it into a helpful, safer assistant. Newer reasoning models add more reinforcement learning on problems with checkable answers.
- 8Reasoning models and thinking12 minA reasoning model is an AI model that writes out hidden rough work, called thinking, before it gives you an answer. It was trained with reinforcement learning to solve problems step by step, so it is much better at maths, code and planning, but slower and more expensive. In ChatGPT and Claude you control this with thinking or effort settings: turn it up for hard multi-step problems and down for quick facts and rewriting.
- 9How AI sees, hears and speaks12 minMultimodal AI is AI that works with more than one kind of input or output: text, images, audio and video. It sees by cutting a picture into small patches and turning each into tokens, hears by turning speech into text or audio tokens, and speaks with text-to-speech or a single speech-to-speech model. Image and video generators mostly use diffusion: they start from random noise and remove it step by step until a picture that matches your prompt appears.
Track 2 of 5 · 7 lessons · 78 min
Use AI well
Write prompts that work, pull the levers that improve answers, and spot made-up facts.
- 10Your first prompt9 minA prompt is simply the message you type to an AI chatbot. Open any free one (ChatGPT, Claude or Gemini), describe your task the way you would brief a smart new colleague, then keep talking: correct it, add details and ask for changes. Treat the first answer as a draft, not the final word.
- 11Prompting techniques that work11 minPrompt engineering means writing instructions that get reliable, useful answers from an AI model. The official guides from Anthropic, OpenAI and Google agree on the same core moves: be clear and give context, set a role, show examples, structure long prompts, let the model think, and ask for a format. Then test and improve.
- 12The levers that make answers better13 minA weak AI answer usually means the model was missing something, not that you used the wrong magic words. Beyond wording, you can pull 11 levers: give it context, define what good looks like, show examples, set the format, let it think longer, pick the right model, switch on tools like web search, save standing instructions, adjust settings, ask it to check its own work, and split big jobs into steps. Find what went wrong, then pull the matching lever.
- 13Context engineering12 minContext engineering is choosing what goes into an AI model’s context window, the working memory it reads before every answer: instructions, tools, examples, documents and chat history. Prompt engineering is about wording one message; context engineering is about curating everything the model sees. Too little and it guesses; too much and its answers get worse, so the goal is the smallest set of useful information.
- 14Why AI makes things up10 minAn AI hallucination is a confident, believable answer that is false, like a fake book title, a wrong date or a court case that never existed. It happens because language models predict likely-sounding text and are trained in ways that reward guessing over saying “I don’t know”. You can reduce it by giving the AI your sources, allowing it to admit uncertainty and checking anything important yourself.
- 15ChatGPT vs Claude vs Gemini11 minAs of November 2026, ChatGPT, Claude and Gemini are all strong, all have free plans, and their main paid plans cost about $20 a month in the US. None is best at everything, so choose by task: Gemini if you live in Google apps, ChatGPT for the widest set of built-in tools, Claude for long writing and coding work, and Perplexity when you need sources. The best test is to run your own task in two of them and compare.
- 16Use AI to prepare for interviews12 minUse AI as a practice partner, not a script writer. It can research the company, decode the job description, predict likely questions, coach your STAR answers, run a mock interview by voice and play a tough recruiter for salary practice. Do not use it during a live interview unless the employer allows it, never paste confidential information, and never let it invent experience you do not have.
Track 3 of 5 · 10 lessons · 111 min
Build with AI
Understand APIs, tool calling, RAG, MCP, agents, skills, evals and voice agents.
- 17What is an API?9 minAn API (application programming interface) is a set of requests one program can make to another, plus the rules for asking. When an app uses Claude, ChatGPT or Gemini behind the scenes, it sends a request to the model’s API over the internet and gets a response back. You prove who you are with a secret API key, and you pay for the tokens you send and receive.
- 18Tool calling (function calling)9 minTool calling, also called function calling, lets an AI model ask your code to run a function, like get_weather or book_slot. The model never runs your code itself: it replies with the function’s name and inputs, your app runs it, and you send the result back so the model can finish its answer. This is how chatbots check live data and take actions.
- 19Embeddings and RAG10 minRAG (retrieval-augmented generation) means: first search your own documents for the passages that answer a question, then give those passages to the model so it answers from them. The search runs on embeddings, lists of numbers that capture meaning, so "refund" can match "money back". It turns a closed-book exam into an open-book one.
- 20What is MCP?12 minMCP (Model Context Protocol) is an open standard for connecting AI apps like Claude, ChatGPT, VS Code or Cursor to outside tools and data, such as your files, GitHub or a database. Anthropic released it in November 2024, and the official docs compare it to a USB-C port for AI apps: build a connector once and it works in every app that supports MCP. Since December 2025 it has belonged to the Agentic AI Foundation, part of the Linux Foundation.
- 21What is an AI agent?10 minAn AI agent is an AI system that works toward a goal on its own: it decides the next step, uses a tool such as web search or code, looks at the result, and repeats until the job is done. A workflow follows steps you fixed in advance; an agent chooses its own steps. Agents suit open-ended tasks, but they need limits, approvals and testing.
- 22What are Agent Skills?12 minAn Agent Skill is a folder with a SKILL.md file inside: a name, a one-line description of when to use it, and step-by-step instructions, plus optional scripts and reference files. The agent sees only the name and description until a task matches, then reads the rest. Anthropic introduced Skills in October 2025 and published the format as an open standard (agentskills.io) in December 2025; Cursor, Gemini CLI, GitHub Copilot in VS Code and OpenAI Codex use it too.
- 23Subagents and multi-agent systems13 minA multi-agent system splits one big job across several AI agents. Usually a lead agent (the orchestrator) plans the work, sends pieces to helper agents (subagents or workers), each with its own fresh context, and combines what they return. It helps most when the pieces can run in parallel, like broad research, and it hurts when the pieces depend on each other, because it costs many more tokens and the agents can miss each other’s decisions.
- 24Evals: does your AI work?10 minEvals are tests for AI features: a fixed set of realistic inputs, a clear rule for what a good answer looks like, and a score you check every time you change the prompt or model. You find what to test by reading real outputs, then grade with code, a human, or another model using a rubric (LLM-as-a-judge).
- 25How voice AI agents work13 minA voice AI agent listens, thinks and speaks, and can take actions like booking an appointment. Most are built as a cascade: speech-to-text turns your voice into text, a language model writes the reply and calls tools, and text-to-speech reads it out. Newer speech-to-speech models, such as OpenAI’s Realtime API, do all three in one model. The hard parts are speed (replying in about a second), knowing when you have finished speaking, and stopping politely when you interrupt.
- 26Open models: run an LLM on your laptop13 minOpen-weight models like Gemma, gpt-oss, Llama, Qwen, DeepSeek and Mistral can be downloaded and run on your own computer, so your chats stay on your machine and work without internet once the model is downloaded. The two easiest tools are Ollama (one install command, then "ollama run") and LM Studio (a desktop app with a chat window). The main limit is memory: aim for 16 GB of RAM, start with a small model, and use a 4-bit quantized version.
Track 4 of 5 · 6 lessons · 77 min
Automate and code with AI
Automate real work with n8n and build software with Claude Code, Cursor and Codex.
- 27Your first AI automation in n8n14 minn8n is a workflow automation tool: you join apps together as boxes on a canvas, and the workflow runs by itself when something happens. You can add an AI step that reads, summarises or sorts text with a model like GPT, Claude or Gemini. In this lesson you build one real workflow: a form is submitted, an AI summarises it, and the summary lands in your email or Slack.
- 28Vibe coding: apps without code12 minVibe coding means building software by describing what you want to an AI and accepting the code it writes without reading it. Andrej Karpathy coined the term in February 2025, and Collins Dictionary made it its 2025 Word of the Year. App builders like Lovable, Bolt, Replit Agent, v0 and Google AI Studio’s Build mode let a non-coder ship a working prototype in an afternoon, but code nobody has read can hide security holes, so learn the basics before real users or money are involved.
- 29Claude Code for beginners14 minClaude Code is Anthropic’s AI coding agent. You describe what you want in plain English, and it reads your project, edits files and runs commands to get it done, asking your permission along the way. It runs in the terminal, in VS Code and JetBrains, in a desktop app and on the web, and needs a paid Claude plan (from $20 a month as of Nov 2026) or an API account.
- 30Cursor for beginners13 minCursor is a code editor with AI built in, made from the same codebase as VS Code. It predicts your next edit as you type (Tab), and its Agent can plan, write and test whole features from a plain-English request. It has a free plan with limited agent use, and paid plans from $8 a month as of Nov 2026.
- 31OpenAI Codex for beginners12 minCodex is OpenAI’s coding agent. You give it a task in plain English, and it reads your code, makes changes, runs commands and shows you the result to review. It runs inside ChatGPT, as a desktop app, in code editors like VS Code and in the terminal, all signed in with your ChatGPT account.
- 32Claude Code vs Cursor vs Codex12 minAll three are coding agents that can read your project, edit files and run commands. The real differences are where you work (terminal, editor or app), which models you get, and which subscription you already have. Claude Code suits Claude subscribers, Cursor suits people who want an AI editor, and Codex suits ChatGPT subscribers. Non-coders who just want an app can start with a browser app builder.
Track 5 of 5 · 4 lessons · 53 min
Go deeper
Fine-tuning, AI safety, building a small LLM with Karpathy, and where to learn next.
- 33Prompting vs RAG vs fine-tuning12 minStart with prompting, always. Add RAG (retrieval-augmented generation) when the model is missing knowledge, such as your private documents or facts that change often. Use fine-tuning only when the model is missing behaviour, like a strict format or style, that good prompts and examples cannot fix. Let a small test set decide, not hype.
- 34AI safety and alignment14 minAI alignment means getting an AI system to do what people actually intend, honestly and without causing harm, not just what they literally asked for. It is hard because models learn from imperfect signals, so they can game their goals, flatter you, make things up or be tricked by jailbreaks. Labs work on it in four main ways: training with feedback and written principles (RLHF, Constitutional AI), looking inside models (interpretability), attacking their own models (red-teaming), and published rules for when stronger safeguards are needed.
- 35Build an LLM with Karpathy's Zero to Hero15 minNeural Networks: Zero to Hero is a free YouTube course by Andrej Karpathy that takes you from a tiny neural network in plain Python to training your own GPT. You need basic Python and some memory of high-school calculus. Expect well over 10 hours of video, and two to three times that if you code along, which you should.
- 36Best free AI courses and who to learn from12 minIf you are new, start with Karpathy’s one-hour Intro to Large Language Models and 3Blue1Brown’s neural network videos, then a short course from OpenAI Academy or Anthropic Academy. To build, take Hugging Face’s Agents Course and DeepLearning.AI short courses. To understand models from the inside, do Karpathy’s Zero to Hero, fast.ai and Stanford’s lectures. Pick one course for your level, finish it, and build something.
8 videos · play right here
Watch: the best free AI videos
Andrej Karpathy, 3Blue1Brown, Anthropic and others explain it better than anyone. Each video sits inside the lesson it belongs to.
[1hr Talk] Intro to Large Language Models
Andrej Karpathy · 60 min
In lesson 3: What is an LLM? →Deep Dive into LLMs like ChatGPT
Andrej Karpathy · 211 min
In lesson 4: Tokens and context windows →How I use LLMs
Andrej Karpathy · 131 min
In lesson 10: Your first prompt →The spelled-out intro to neural networks and backpropagation: building micrograd
Andrej Karpathy · 145 min
In lesson 35: Build an LLM with Karpathy's Zero to Hero →Let's build GPT: from scratch, in code, spelled out.
Andrej Karpathy · 116 min
In lesson 35: Build an LLM with Karpathy's Zero to Hero →Large Language Models explained briefly
3Blue1Brown · 8 min
In lesson 3: What is an LLM? →Attention in transformers, step-by-step | Deep Learning Chapter 6
3Blue1Brown · 26 min
In lesson 5: Transformers and attention →But how do AI images and videos actually work? | Guest video by Welch Labs
3Blue1Brown · 37 min
In lesson 9: How AI sees, hears and speaks →Original papers, talks and essays
Learn from the people who built it
Each lesson takes one original idea, breaks it down in plain English, and shows how to use it.
John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon
The four researchers who proposed the 1956 Dartmouth summer study that started AI as a field. McCarthy chose the name “artificial intelligence”.
A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence (August 31, 1955) Broken down in: What is AI? →Alan Turing
British mathematician and one of the founders of computer science. His 1950 paper is one of the first serious studies of machine intelligence.
Computing Machinery and Intelligence (Mind, October 1950) Broken down in: What is AI? →Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton (University of Toronto)
Hinton later shared the 2018 Turing Award and the 2024 Nobel Prize in Physics. Sutskever, his doctoral student, went on to co-found OpenAI and serve as its chief scientist.
ImageNet Classification with Deep Convolutional Neural Networks (NeurIPS 2012) Broken down in: How AI started: 1956 to ChatGPT →Rich Sutton
A pioneer of reinforcement learning, who shared the 2024 ACM Turing Award with Andrew Barto.
The Bitter Lesson (essay, March 13, 2019) Broken down in: How AI started: 1956 to ChatGPT →Andrej Karpathy
A founding member of OpenAI and former director of AI at Tesla. He now runs Eureka Labs, an AI education company.
[1hr Talk] Intro to Large Language Models (talk, 2023) Broken down in: What is an LLM? →Rico Sennrich, Barry Haddow and Alexandra Birch (University of Edinburgh)
Their 2015 translation paper brought byte pair encoding (BPE) into neural language models. Versions of BPE now build the tokenizers of GPT, Llama and many other models.
Neural Machine Translation of Rare Words with Subword Units (2015) Broken down in: Tokens and context windows →Andrej Karpathy
A founding member of OpenAI and former director of AI at Tesla, known for teaching how LLMs work by building them from scratch.
Let’s build the GPT Tokenizer (video lecture, 2024) Broken down in: Tokens and context windows →Ashish Vaswani and seven co-authors, mostly at Google Brain and Google Research
The eight authors (Vaswani, Shazeer, Parmar, Uszkoreit, Jones, Gomez, Kaiser and Polosukhin) of the paper that introduced the transformer, the design behind GPT, Claude and Gemini.
Attention Is All You Need (2017) Broken down in: Transformers and attention →Ashish Vaswani, Noam Shazeer, Niki Parmar and colleagues
The Google team of eight authors who introduced the transformer
Attention Is All You Need (paper, 2017) Broken down in: The papers that built modern AI →Learning AI: questions people ask
What is the best way to learn AI for free?
Follow one path in order instead of jumping between random videos. Start with what AI and LLMs are, then learn to prompt well, then learn the parts of AI products (APIs, tool calling, RAG, MCP, agents), then build something small with n8n, Claude Code, Cursor or Codex. Each lesson here is short, links the best free videos (Andrej Karpathy, 3Blue1Brown, Anthropic, IBM), and ends with a hands-on task and a quick check.
How long does it take to learn AI?
The 36 lessons here take about 7 hours of reading and doing, plus optional videos. One lesson a day gets you from complete beginner to building your first agent and automation in about a month. Going deep, for example building a GPT from scratch with Karpathy’s Zero to Hero course, takes many more weeks.
Do I need to know coding to learn AI?
No. Tracks 1 and 2 (understanding AI and using it well) need no code at all. Track 3 shows small, readable code snippets but explains every line. Tools like n8n and app builders let you automate and build without coding, and Claude Code, Cursor and Codex can write the code for you once you know what to ask.
Which lesson should I start with?
If you are new, start with lesson 1, “What is AI?”, and go in order. If you already use ChatGPT or Claude daily, start at the Use AI well track. If you want to build products, start at What is an API? in the Build with AI track.
Is this Learn AI course really free?
Yes. Every lesson is free to read, with no login and no sign-up. Your progress is saved in your own browser on this device. The videos are free on YouTube and play right on the page.
Turn what you learn into a job
Free · No signup to start
Learned it? Now say it out loud.
AI interviews ask you to explain LLMs, RAG and agents simply. Practise with an AI interviewer and get feedback on every answer.