Learn AI.
Understand what's under the hood.
Every product you use now claims to have AI. Very few people understand what that actually means. This track teaches the fundamentals — machine learning, neural networks, NLP, and LLMs — with hands-on browser demos that run entirely on your machine. No API keys. No cloud bills. No black boxes. Just the concepts, made concrete.
Prerequisite: Python. The AI track assumes you can read Python code — variables, functions, loops, dicts. If you haven't completed the Python track, do that first. AI is applied Python. Without the foundation, every lesson becomes a struggle to decode syntax instead of understanding concepts.
What you'll learn: What machine learning actually is (and isn't). How neural networks learn. What NLP does. How LLMs work. How to train a small model in your browser. What you can — and can't — build with AI today.
Why AI, and why now
AI is the defining technology of the decade. It's in your search results, your recommendations, your email spam filter, your phone's camera, your writing assistant. The people who understand how it works — not just how to prompt it — will have a structural advantage for the next 20 years. Not because they'll build the next GPT, but because they can evaluate claims, spot hype, and build small useful things.
The barrier to entry has never been lower. Ten years ago, training a neural network required a GPU cluster and a PhD. Today, TensorFlow.js runs a small model in a browser tab. The concepts are the same — the access has changed. You can learn the fundamentals on a laptop with no setup, which is exactly what this track does.
The trap: AI has an enormous amount of noise. Influencers selling courses. People claiming AGI is here. People claiming AI is useless. The way through the noise is understanding the mechanism. Once you know what a transformer actually does, you can read a paper, evaluate a product, and form your own opinion. That's the goal of this track.
🗺️ The 6-week path
Follow these in order. The concepts build on each other.
What AI actually is
The difference between AI, machine learning, and deep learning. What "training a model" means. Why AI works now when it didn't in the 1980s. The vocabulary you need to read any AI article without being confused.
Machine learning basics
Linear regression, logistic regression, decision trees, and clustering. Not the math — the intuition. What each algorithm does, when to use it, and why "machine learning" is mostly just statistics with a marketing budget. A live demo trains a model in the browser.
Neural networks
What a neuron is. How layers work. Backpropagation — the algorithm that makes training possible, explained without calculus. Why deep learning beat everything else once GPUs arrived. An interactive demo lets you train a tiny neural network in the browser.
NLP — how machines read
Tokenization, embeddings, attention — the three ideas behind modern NLP. How text becomes numbers. Why "king - man + woman = queen" works in embedding space. The mechanics behind sentiment analysis and text classification.
LLMs and transformers
What GPT actually does. Why it "hallucinates." What a token is. What temperature does. How prompt engineering works (and doesn't). The transformer architecture explained without a math PhD — enough to understand the mechanism.
Build something
Pick a project from the list below. Build it. Ship it. The whole point of learning AI is making something with it — not collecting concepts. A sentiment analyzer, an image classifier, a chatbot. Small, focused, real.
🎯 The 5 concepts that matter most
If you only truly understand five things about AI, make it these.
1. AI is not magic
Every "AI" feature you use is a model trained on data to find patterns. It's statistics at scale. Once you see it as pattern-matching, the mystery disappears — and so does the hype.
2. Training is finding numbers
A neural network is millions of numbers. Training adjusts those numbers so the network outputs the right answer for the training data. That's the entire process. Gradient descent, backpropagation — all just ways to adjust numbers.
3. Embeddings are the bridge
Text becomes numbers through embeddings. Words with similar meanings have similar numbers. This is how machines "understand" language — not by reading, but by comparing coordinates in a high-dimensional space.
4. LLMs predict the next token
GPT doesn't "think." It predicts the most likely next word based on everything before it. That single mechanism — next-token prediction — scaled up to billions of parameters, produces something that feels like intelligence.
5. Hallucinations are features, not bugs
An LLM that always told the truth would be impossible with the current architecture — it doesn't know what truth is, it knows what sounds likely. Understanding this is what separates informed users from confused ones.
🎬 See it in action
AI that runs in your browser — no server, no API, no data leaves your machine.
Input layer Hidden layers Output ───────────── ───────────── ────── ┌───┐ ┌───┐ [feature 1] ──→ │ │ │ │ │ │──→│ │ [feature 2] ──→ │ │ │ │ ──→ [prediction] │ │ │ │ [feature 3] ──→ │ │──→│ │ └───┘ └───┘ (weights) (weights) Each connection has a "weight" — a number. Training adjusts the weights to reduce error. That's the entire neural network.
The demo on this page is a trained model running locally. The sentiment analyzer (linked below) uses a small model in your browser to classify text. No text is uploaded — the computation happens on your device. That's the difference between "AI" as a marketing word and AI as a technology you can actually inspect. Every lesson in this track includes a working demo so you can see the concept, not just read about it.
⚠️ Common mistakes
The misconceptions that keep people confused about AI.
LLMs don't understand. They predict. A model that answers "what is 2+2" isn't doing math — it's generating the most likely response based on training data. This distinction matters for every AI product you evaluate.
You can't learn AI without Python. Every tutorial, every library, every paper uses Python. Trying to learn AI without it is like trying to learn literature without reading. Do the Python track first.
LLMs are the top of the stack. Starting there is like starting physics with quantum mechanics. Learn what a model is, how training works, what a neural network does — then LLMs make sense.
"AGI by 2027," "AI will replace all developers," "AI is useless." All headlines, no nuance. The truth: AI is a powerful pattern-matching tool with real limitations. Learn the mechanism, ignore the noise.
Watching 50 hours of AI videos without building anything is the #1 reason beginners fail. Watch one concept, build something small, move on. The building is where the learning happens.
You don't need a math degree, but you do need to understand what a derivative means and what a matrix is. Skipping math entirely means you'll hit a wall at backpropagation and never understand why neural networks work.
TensorFlow.js runs in a browser. A small model runs on a laptop. You can build useful AI tools with no cloud, no budget, no team. The barrier is understanding, not infrastructure.
If you can't explain what a token is to a friend, you don't understand it yet. Every lesson in this track has a challenge. Do them. The testing is the learning.
🔨 Projects by difficulty
Build these in order. Each one uses only concepts from the phases before it.
Sentiment analyzer
Classify text as positive/negative using a pre-trained model. Runs in the browser. Uses phase 1 + 2.
Linear regression demo
Train a model to predict values from a small dataset. Visualize the line of best fit. Uses phase 2.
Image classifier
Use a pre-trained model to classify images. Upload a photo, get a label. Uses phase 3.
Text summarizer
Summarize long text using an LLM or extractive method. Uses phase 4 + 5.
Simple chatbot
Rule-based first, then LLM-powered. Compare the two. Uses phase 5.
RAG demo
Connect an LLM to your own documents. Answer questions about PDFs. Uses phase 5.
❓ Frequently asked questions
Do I need Python to learn AI?
Yes. Python is the undisputed language of AI and machine learning. Every major AI library — PyTorch, TensorFlow, Scikit-learn, Hugging Face — uses Python. Complete the Python track first, or at minimum be comfortable with variables, functions, lists, dicts, and loops.
How long does it take to learn AI?
For the basics — what machine learning is, how neural networks work, how to train a simple model — 4-6 weeks. To be job-ready as an AI engineer, 12-18 months of consistent effort including Python mastery, math fundamentals, classical ML, deep learning, and LLM engineering.
Do I need a math degree to learn AI?
No. You need to understand basic linear algebra, statistics, and a little calculus. You don't need a math degree. Most AI engineering work involves applying existing models, not deriving new math.
Should I learn PyTorch or TensorFlow?
PyTorch is the industry standard for research and modern LLM development. TensorFlow is still used in production but PyTorch is the safer bet for 2026. This course teaches concepts first, frameworks second.
What is the difference between AI, machine learning, and deep learning?
AI is the broad field. Machine learning is a subset — systems that learn from data. Deep learning is a subset of machine learning — neural networks with many layers. All deep learning is machine learning, all machine learning is AI, but not the reverse.
Can I run AI models in the browser?
Yes. TensorFlow.js lets you train and run small models entirely in the browser. This course uses it for interactive demos so you can see machine learning working without installing anything. For production or larger models, you'd use Python with PyTorch or TensorFlow.
What can I build after this course?
A sentiment analyzer, an image classifier, a simple chatbot, a text summarizer. Small, focused AI projects that demonstrate understanding. Not production-grade systems — those take months more. But the foundation: you'll understand what's happening under the hood of every AI product you use.
Is this AI course really free?
Yes. Every lesson, every demo, every code example is free. No signup, no email gate, no paywall. The browser-based demos run entirely on your machine — nothing is uploaded.