Curious
Start with clear foundations, then language models and honest limits.
Learning center
Learn AI with clarity
A free encyclopedia of AI ideas—from everyday language to deeper mechanisms. Start simple, then dive as far as you want. 48 topics. No tools pitch here—just teaching.
Pick a track—or browse everything
Tracks are ordered reading lists over the same pages. Every page also links to related topics so you can explore freely.
Builder
Mechanisms you need to build real workflows: tokens, transformers, RAG, agents.
Pro
Deeper systems: attention, training, eval, safety, and classical foundations.
A. Foundations
What AI is, how the field grew, and where the limits are.
L0
Here's what AI is
A plain-language map of artificial intelligence: what people mean by AI today, and what it is not.
L0A short history of AI
From early rule systems to machine learning and today's large models—without the hype timeline.
L0Narrow AI, AGI, and ASI
What exists now (narrow AI) versus long-term ideas like AGI—and how to tell hype from fact.
L0AI, ML, and deep learning
How the terms nest: AI is the big idea, machine learning is a way to get there, deep learning is one powerful branch.
L0Data and labels
Why AI systems need examples, what labels are, and how messy data creates messy results.
L0What AI can and cannot do
Honest strengths of today's systems—and failure modes people often miss.
B. Machine learning
How computers learn patterns from data.
L1
Supervised learning
Learning from labeled examples: inputs paired with the answers you want.
L1Unsupervised and self-supervised learning
Finding structure without hand labels—and the self-supervised tricks behind modern models.
L1Reinforcement learning
Learning by trying actions and getting rewards—useful for games, robots, and some model tuning.
L1Features, loss, and optimization
How models turn inputs into numbers, score mistakes with a loss, and improve step by step.
L1Overfitting and generalization
Why a model can ace the training set and still fail on new data—and how people fight that.
L1Classical machine learning models
Linear models, trees, and ensembles—older tools that still win on many tabular problems.
C. Deep learning & compute
Neural nets, training, and the computers behind them.
L1
Neural networks
Layers of simple math units that learn useful representations from data.
L2Backpropagation
How a network figures out which weights to change after it makes a mistake.
L1CNNs and computer vision
Convolutional networks and how machines learn to see patterns in images.
L1RNNs and sequence models
Earlier ways to handle text and time series before transformers took over.
L1GPUs, TPUs, and training compute
Why modern AI needs special chips—and what “compute” means in practice.
L2Scaling laws and compute
The observed pattern that bigger models, more data, and more compute often improve results—within limits.
L2Distributed training basics
How teams train large models across many machines without losing the plot.
D. Language models
Transformers, tokens, and how LLMs predict text.
L0
Language models
Systems that learn to predict the next piece of text—and why that unlocks so many apps.
L1Tokenization
How text is split into tokens the model can score—one of the most practical LLM ideas to understand.
L1Embeddings
Turning tokens and documents into vectors so meaning can be compared with math.
L1Transformers
The architecture behind most modern language models—and why attention changed the game.
L2Self-attention
How each token looks at other tokens to decide what matters for the next prediction.
L1Context windows and memory
How much text a model can “see” at once—and why that is not the same as human memory.
L1Pretraining and fine-tuning
Broad training on lots of text, then narrower training for styles, tasks, or safety behaviors.
L2Mixture of experts and modern variants
Ways labs scale models by routing tokens to specialist subnetworks—and other common upgrades.
E. Prompting & use
How instructions steer models—and where they fail.
L0
Prompting
How your words become the model's input—and why next-token prediction makes prompting matter.
L1Prompting patterns
Roles, examples, structure, and step-by-step reasoning—patterns that often help, with limits.
L1RAG (retrieval-augmented generation)
Pulling fresh documents into the prompt so answers can cite outside knowledge.
L0Hallucinations
When a model invents confident wrong answers—and practical ways to reduce the risk.
L1Multimodality
Models that handle more than one kind of input or output—text, images, audio, and more.
F. Agents & systems
Loops, tools, planning, memory, and evaluation.
L0
AI agents
Systems that loop: observe, decide, act with tools, and repeat toward a goal.
L1Tool use and function calling
Letting a model call APIs or functions instead of only writing text.
L1Planning and multi-step reasoning
Breaking goals into steps—how agents plan, and how that overlaps with classical search.
L1Memory for agents
Short chat context versus long-term stores—and why agents still forget.
L1Evaluating agents
How to test agent systems when “one right answer” is not enough.
G. Generative media
Images, video, audio, and diffusion-style models.
H. Classical AI
Search, logic, and uncertainty—ideas that still matter.
L1
Search and planning
Classical AI methods for exploring options and choosing paths—still useful today.
L1Knowledge representation and logic
Storing facts as symbols and rules—powerful for clarity, brittle for messy reality.
L1Uncertainty and Bayesian thinking
How probability helps machines (and people) update beliefs when the world is noisy.
I. Eval, safety & society
Benchmarks, safety, fairness, and real-world impact.
L1
Benchmarks and evaluation
How researchers score models—and why leaderboards can mislead.
L1Alignment and safety
Making AI systems behave as intended—and reducing harm from misuse or mistakes.
L0Bias, fairness, and accountability
How data and design choices can treat people unfairly—and who is responsible.
L0AI and society
Jobs, information, power, and policy—how AI systems touch everyday life.
J. Reference
Glossary and quick lookups across the library.
Prefer news while you learn? Read Weekly Brief or the Live Feed.