Everything in one place — courses, tutorials, articles, and AI tools.
Browse all 361 items, or start typing.
Machine Learning Fundamentals
beginner • 3 sections
AI Agents from Scratch
Understand what an AI agent really is and build one yourself. Covers tool use, the agent loop, memory, multi-agent patterns, and the guardrails that keep agents safe and affordable.
advanced • 3 sections
AI for Absolute Beginners
Start from zero. Understand what AI really is, how modern models work, and how to use them confidently in everyday life and work — no math or coding required.
beginner • 2 sections
AI Image Generation Studio
A complete beginner-friendly course for planning, prompting, generating, reviewing, and publishing AI images for lessons, tutorials, products, and websites — with a reusable visual workflow from brief to final asset.
beginner • 4 sections
Building LLM Apps with RAG
Go from prompts to products. Learn how to ground an LLM in your own data using Retrieval-Augmented Generation, including embeddings, vector search, chunking, and evaluation.
intermediate • 3 sections
Data Science Foundations
Learn the workflow that turns raw data into insight: asking questions, exploring data, visualizing it, and reasoning about uncertainty — the groundwork for all of ML.
intermediate • 3 sections
Deep Learning Essentials
Understand how neural networks learn and meet the architectures behind modern AI — from the perceptron to CNNs and transformers — with intuition over heavy math.
intermediate • 3 sections
Generative AI: Images, Audio & Video
A hands-on tour of creative AI. Learn how diffusion models generate images, how to prompt them well, and how to create voice, music, and video — with both no-code tools and APIs.
beginner • 3 sections
Machine Learning Foundations
Build a solid base in classical machine learning: how supervised learning works, how to train and evaluate models, and how to avoid the most common pitfalls — with hands-on Python.
intermediate • 3 sections
Prompt Engineering Mastery
Learn the repeatable techniques that turn vague prompts into reliable, high-quality results — from clear instructions to few-shot examples, structured output, and chain-of-thought.
beginner • 3 sections
Python for AI: A Gentle Start
New to coding? This course teaches just enough Python to start doing AI — variables, data structures, functions, and the data libraries you’ll use every day.
beginner • 3 sections
Write Your First Effective Prompt
Turn vague requests into prompts that get great results on the first try.
Prompt Engineering
Few-Shot Prompting: Teaching by Example
Examples in a prompt fix the output format, and quietly bias the answers. Flip the labels of four examples and watch a real model’s verdicts move, then calibrate it.
Prompt Engineering
Reliable JSON from an LLM: Four Ways to Ask
A real model turns café orders into JSON four ways, from a plain request to constrained decoding. Your browser parses every answer: valid is easy, right is not.
Prompt Engineering
Call an LLM API in Python
Make your first programmatic call to a language model and handle the response.
LLM Application Development
Build a Simple Chatbot with Memory
Create a command-line chatbot that remembers the conversation.
LLM Application Development
Streaming LLM Responses: Shorter Waits, Same Answer
Replay a real model’s answer token by token, streamed and not, and see why a long prompt delays the first word even when you stream.
LLM Application Development
Embeddings: Turning Meaning into Numbers
Explore 32 sentences through a real embedding model, measure “close” three ways, and find out what similarity scores miss.
LLM Application Development
Build a RAG System: Search, Then Answer
Follow 20 customer questions through a complete RAG pipeline, from embedding to answer, and find out whether each wrong answer came from the search or the reading.
LLM Application Development
Run LLMs Locally with Ollama
Run open models on your own machine — private, offline, and free.
AI Tools & Workflows
Fine-Tuning vs. RAG: Which Do You Need?
A practical decision guide for customizing model behavior.
LLM Application Development
Hallucinations: Why Models Make Things Up
Read 1,260 real answers from a small model, see where it invents names and dates, then try the fixes: a source, permission to say “I don’t know”, and asking more than once.
LLM Application Development
Tokenization: How Language Models Read Text
Train a real tokenizer in your browser, see the numbers a model actually receives, and find out why the same sentence costs more in some languages than others.
Deep Learning
How Transformers Work: Attention, Step by Step
The idea behind every chatbot: each word decides which other words matter. Tap a word in a real sentence and watch where its attention goes.
Deep Learning
Cleaning Data with pandas
Run a messy table through a cleaning script one line at a time, catch a filter quietly dropping a customer, and see what filling in blanks does to a survey’s answers.
Python for AI
Splitting Data the Right Way
Get 82% accuracy out of pure noise by choosing columns before splitting, watch a model recognise people instead of learning, and learn the checks that keep a test score honest.
Machine Learning
Evaluating Classifiers: ROC, Precision-Recall and Calibration
Three real models score 3,000 bank customers. Slide the threshold along ROC and precision-recall curves, make the yeses rare, and see whether “99% likely” means 99%.
Machine Learning
Prompt Injection: Hidden Instructions in Your Data
A real model summarises a café’s inbox while twelve emails try to hijack it. Test four defences and a separate check, and see which attacks get through everything.
LLM Application Development
Evaluating LLM Outputs: Scores You Can Trust
Grade a support bot’s answers with BLEU, ROUGE and a quick check, catch them rewarding a wrong answer, and see how many test questions it takes to spot a real improvement.
LLM Application Development
Function Calling: How an AI Uses Your Tools
A model can’t check the weather, but it can ask your code to. Step through the real message exchange, then try to break a tool call.
LLM Application Development
AI Agents: A Model, Tools and a Loop
Watch an agent think, act and observe its way through a task, then stop it making a confident mistake. The two guardrails every agent needs.
LLM Application Development
Caching LLM Calls: Stop Paying Twice
Run a support bot’s traffic through prompt caching and a semantic cache, and find the settings that save most of the bill and the ones that quietly cost more.
LLM Application Development
Sentiment Analysis: What the Model Sees, and Where It Breaks
Two real Hugging Face models read café reviews. Delete words to see what each answer rests on, then watch them fail, confidently, on sarcasm, slang and plain facts.
Machine Learning
Transcribe Audio with Whisper: Size, Speed and Noise
Listen to real recordings with noise mixed in, read what three sizes of Whisper wrote, and score them word by word. On clean audio they tie; in noise, size pays.
AI Tools & Workflows
LoRA: Fine-Tune a Model by Training a Tiny One
Full fine-tuning adjusts billions of numbers. LoRA trains two thin matrices instead. Count what it saves on real model sizes, then train one yourself.
Deep Learning
System Prompts That Work: Rules, Facts and Tests
Five system prompts, twelve awkward customer messages and a real model, with a test on every reply. One line passes 2 tests, clear rules pass 10, and only facts fix the last two.
Prompt Engineering
Chunking Strategies for Better RAG
How you cut documents into pieces quietly decides what your AI can answer. Cut one yourself and watch an answer get sliced in half.
LLM Application Development
Measuring Retrieval Quality for RAG
Move answers up and down a results list to see what recall, MRR and nDCG reward, then compare four searches on a real test set.
LLM Application Development
pandas GroupBy: Split, Apply, Combine
Step a table through split, apply and combine, see how missing values are counted, and watch a grouping reverse which of two bots looks better.
Python for AI
Feature Engineering: Giving Models Better Inputs
Give a simple model the same hour of the day three ways and watch its accuracy nearly quadruple, then see how units and useless columns mislead a nearest-neighbour model.
Machine Learning
Cross-Validation Explained
One test split can score the same model anywhere from 67% to 100%. Watch the luck, let cross-validation steady it, and split rare classes and time safely.
Machine Learning
Hyperparameter Tuning Without Fooling Yourself
Search a real model’s settings by grid and at random, then watch the winner of a hundred tries fall apart on new data.
Machine Learning
Random Forests, Explained
One decision tree memorises its data; a hundred slightly different trees, voting, beat it. Grow both in your browser and see why the randomness is the trick.
Machine Learning
Gradient Boosting and XGBoost, Step by Step
The method behind countless winning models on table-shaped data. Add small trees one at a time, each fixing the last one’s mistakes, and find out when to stop.
Machine Learning
K-Means Clustering, Step by Step
Find groups in data nobody labelled. Run k-means by hand, trap it in a bad answer, pick k with the elbow, and see why scaling your data matters.
Machine Learning
Time Series Forecasting: Beat the Baseline
Forecast a café’s coffee sales four ways, backtest them honestly, and watch a model ace a test it should have failed.
Machine Learning
Recommendation Systems: How “You Might Also Like” Works
Rate a few dishes and get recommendations from a model trained in your browser, then watch a recommender trap itself in a loop of its own making.
Machine Learning
Data Visualization: Plot It Before You Trust It
Four datasets with identical statistics and nothing else in common, a geyser whose histogram changes with every bin width, and a bar chart that turns 7% into sixfold. Drag, slide and see.
Python for AI
Prompt Chaining: Small Steps You Can Check
Count the complaints in twelve café reviews with one big prompt, then with a chain of small steps. Shuffle the reviews, find the misread ones, and see which jobs belong in code.
Prompt Engineering
Validate LLM Output with Pydantic: Valid Isn’t Right
Validate a real model’s JSON as you type, against a Pydantic model with business rules, then see what happens when its errors are sent back for a second try.
LLM Application Development
Discover Topics with Embeddings and Clustering
Cluster 120 support tickets by meaning, live, read each cluster’s words and typical tickets, and find a problem that the ticket categories split four ways.
LLM Application Development
Token Budgeting: What to Keep When the Chat Gets Long
Squeeze a real support chat into less and less room. Drop or summarise old turns, read what a real model answers each time, and watch the bill grow with every message.
LLM Application Development
Temperature and Sampling: How an AI Picks Its Next Word
Every reply is written one word at a time, by a roll of loaded dice. Turn the dial that loads them and watch the writing change.
LLM Application Development
Citations in RAG: Answers You Can Check
Number the sources and make the model cite them, then check every citation in your browser. A real model cites the right article every time it has it, and is still wrong six times.
LLM Application Development
Hybrid Search: Keyword Plus Vector
Keyword search misses meaning; vector search misses exact codes. Search a help centre three ways and tune the blend that finds every answer.
LLM Application Development
Neural Networks from Scratch
One neuron can only draw a straight line. Set its dials by hand, then train a real network in your browser and see why hidden layers need a bend.
Deep Learning
Activation Functions: Why Neural Networks Need a Bend
Without one, a network of any size can only draw a straight line. Train a tiny network with and without, then watch a deep one stop learning.
Deep Learning
Gradient Descent, Intuitively
How a model learns, shown as a ball rolling downhill. Set the step size, drop the ball, and see why one number decides whether training works.
Deep Learning
Overfitting and Regularization
Why a model that aces its training data can fail on everything else. Bend a curve until it memorises the noise, then fix it with more data or a penalty.
Machine Learning
Reading a Confusion Matrix
A model can be 99% accurate and useless. Drag a threshold through 40 emails, watch four numbers trade places, and learn which mistake you can live with.
Machine Learning
SQL for Data Science: Queries, Groups and Joins
Run real SQL in your browser: step through a query in the order the database runs it, join tables four ways, and catch the join that quietly doubles your totals.
Python for AI
NumPy Essentials: Shapes, Axes and Broadcasting
Squash a table along each axis, line up array shapes until they fit or refuse to, and time a Python loop against NumPy.
Python for AI
Rate Limits and Retries: Back Off and Spread Out
Watch a token bucket refuse bursts under a per-minute limit, then race five retry strategies against the same API and see which one wastes thousands of calls.
LLM Application Development
Choosing an Embedding Model: Test It on Your Own Questions
Five free embedding models answer a help centre’s questions in English, Spanish and German. Build your own leaderboard, change the rule, and watch the benchmark favourite lose.
LLM Application Development
Bias in Machine Learning: Measuring What’s Fair
Watch a model fail a group it barely saw, then try to make a loan score fair four different ways, and find out why you can’t have them all.
Machine Learning
Data Drift: Catching a Model Going Stale
Run the standard drift checks on a shop’s orders, then watch a fraud model fail in a way no input check can see.
Machine Learning
Object Detection: Boxes, Overlap and Cut-offs
Score a box by its overlap with the truth, then tune a detector’s confidence and non-maximum suppression thresholds and watch objects appear, vanish and double up.
Deep Learning
Image Segmentation: Labelling Every Pixel
Make masks for a microscope picture with thresholds, see why pixel accuracy flatters them, then split touching cells apart and count them.
Deep Learning
PCA for Dimensionality Reduction
Keep most of what your data says with far fewer numbers. Turn a line through a cloud by hand, compress a picture, and meet the units mistake that breaks it.
Machine Learning
t-SNE and UMAP: Maps of High-Dimensional Data
Run t-SNE live on 500 handwritten digits, then feed it data with a known shape and see which parts of the map you can trust.
Machine Learning
Imbalanced Data: When What You Want Is Rare
Try class weights, oversampling, undersampling and SMOTE on the same factory line, see what they really change, and watch a resampling mistake fake a perfect score.
Machine Learning
SHAP: Why Did the Model Say That?
Split a prediction into fair shares by revealing its inputs in every order, then use SHAP to catch a model that is quietly cheating.
Machine Learning
Reinforcement Learning: Learning from Rewards
No answers, just rewards. Play the explore-or-exploit game yourself, then train a mouse to find the cheese and watch what it learns spread backwards.
Deep Learning
How RLHF Turns a Model into an Assistant
People compare answers, a reward model learns their taste, and the chatbot chases its score. Be the rater, then loosen the leash and watch it game the reward.
Deep Learning
Word Embeddings and word2vec
How words become points on a map of meaning. Train word2vec live, watch cats and dogs find each other unaided, then do arithmetic with meaning.
Deep Learning
Zero-Shot Classification: Labels Without Examples
Sort 50 support tickets into five categories with a small open model and no training, see what the prompt changes, and how few labelled examples a fast model needs to catch up.
LLM Application Development
Reranking for RAG: A Second, Closer Read
Watch a reranker read a search’s shortlist properly, fix its mistakes and make a few of its own, and find how many results are worth reranking.
LLM Application Development
Query Rewriting: Fixing the Question Before You Search
Follow-ups like “how long does it take?” break RAG search. Rewrite 22 help-desk chats four ways with a small open model and see where the answer lands, and where rewriting makes it worse.
LLM Application Development
Synthetic Data: When an LLM Writes Your Training Set
A small model writes training messages for a bank’s support classifier. Read them, measure how alike they are, then train on them in your browser and test on real customers.
LLM Application Development
Multi-Domain Research Agent Tutorial
A comprehensive guide to building and deploying a multi-agent research application using OpenAI's Agents SDK.
A/B Testing: Telling a Real Win from Luck
Identical pages can look 20% apart by chance. Run thousands of tests in seconds and learn the three mistakes behind most false wins.
Machine Learning
Outlier Detection: Finding What Doesn’t Fit
Watch late deliveries hide from a textbook rule, then compare four outlier detectors on data with more than one kind of normal, and see an isolation forest at work.
Machine Learning
Autoencoders: Learning by Squeezing
Train an autoencoder on handwritten digits in your browser, draw new digits from its codes, and use it to spot the odd ones out.
Deep Learning
Fine-Tuning a Text Classifier: When a Few Examples Beat Zero-Shot
Train a classifier in your browser on real bank customer queries, from 1 to 48 examples per intent, and compare it with zero-shot and a fully fine-tuned model.
Machine Learning
GANs: Two Networks in a Duel
Train a generator and a discriminator against each other in your browser, watch them reach a stalemate, and see how a small imbalance makes the generator collapse.
Deep Learning
Data Labelling: Agreement, Guidelines and Noisy Labels
See why two labellers agreeing 90% of the time can mean nothing, measure agreement beyond chance with kappa, and watch random and systematic label mistakes mislead a model and its score.
Machine Learning
What Is a Large Language Model, Really?
A jargon-free explanation of the technology behind ChatGPT, Claude, and Gemini.
Tutorials
Open vs. Closed AI Models: How to Choose
Open-weight models have caught up fast. Here is how to decide which path fits your project.
Opinion
Getting Started with AI Without Writing Code
You do not need to be a programmer to get real value from AI today.
Tutorials
RAG, Explained Simply
Why "retrieval-augmented generation" is the most important pattern in applied AI.
Tutorials
AI Ethics for Builders: A Practical Checklist
Responsible AI is not abstract. Here are concrete questions to ask before you ship.
Opinion
AI Agents: Hype vs. Reality
Autonomous agents are promising and overhyped. Here is where they actually work today.
Industry
How to Learn AI in 2026: A Roadmap
A practical, no-nonsense path from curious beginner to capable builder.
Tutorials
Does Prompt Engineering Still Matter?
As models get smarter, is prompting becoming obsolete? Not quite.
Opinion
Context Windows, Explained
What a "200k context window" really means — and why bigger isn’t always better.
Tutorials
Why Models Hallucinate (and What Helps)
Hallucination isn’t a glitch — it’s a side effect of how models work.
Research
When Fine-Tuning Is Worth It
Most teams reach for fine-tuning too early. Here’s when it actually pays off.
Tutorials
Evaluation Is the Real Moat
Anyone can call an API. The teams that win are the ones who can measure quality.
Opinion
Small Models Are Underrated
You probably don’t need a frontier model for that task.
Opinion
Anatomy of a Modern AI App
The common building blocks behind most production AI features today.
Tutorials