AI & LLM Fundamentals
How AI actually works, without the hype or the math. Learn what large language models really do — tokens, next-token prediction, training, context and their limits — and remember it with spaced repetition.
- practice cards
- 90
- practice cards
- per day
- ~10 min
- per day
- level
- Beginner
- level
- modules
- 5
- modules
What is a large language model?
A large language model (LLM) is an AI system trained on enormous amounts of text to do one deceptively simple thing: predict the next token. From that single capability — repeated over and over — emerges the ability to answer questions, write code, summarize documents and hold a conversation.
Under the hood, an LLM does not "look things up". It breaks text into tokens, turns them into numbers, and uses billions of learned parameters to estimate what comes next. Understanding tokens, the context window, training and inference is what separates using these tools by superstition from using them with intent.
This track explains AI and LLMs from first principles in plain English — no advanced math required — and uses spaced repetition so the concepts stick. It is for anyone who wants to actually understand the tools they use every day, from curious beginners to engineers building on top of them.
5 modules, seed to bloom
Each module is a set of practice cards — 90 in total. Answer, review, and watch your knowledge grow from seed to full bloom.
What is an LLM
How LLMs work — tokens, next-token prediction, parameters, sampling, and the transformer
18 cardsTraining & Inference
How models are built and used — pretraining, fine-tuning, RLHF, and the training vs inference split
18 cardsContext, Tokens & Embeddings
What a model sees and remembers — context windows, statelessness, token counting, and embeddings
18 cardsCapabilities & Limits
What LLMs can and can't do — hallucinations, knowledge cutoff, bias, and why they make things up
18 cardsModel Types
Picking the right kind of model — chat, base, embedding, multimodal, reasoning, and foundation models
18 cardsSample questions
A taste of the real cards. Pick an answer, then reveal the explanation.
What is the core task an LLM performs to generate text?
- APredicting the next token — given the text so far, it estimates what comes next
- BSearching a knowledge base — it looks up the best matching stored answer
- CParsing grammar rules — it builds sentences from a fixed set of language rules
- DTranslating intent — it converts your meaning into a pre-written response
What is a model's "context window"?
- AThe maximum amount of text — measured in tokens — it can consider at once
- BThe maximum number of users — it can serve at the same moment in parallel
- CThe maximum size of a reply — the longest answer it is allowed to produce
- DThe maximum file it can open — the largest document it can store on a disk
What is a "hallucination" in the context of an LLM?
- AA confident answer that is false — fluent text the model simply made up
- BA request the model refuses — output it declines to produce on safety grounds
- CA pause during generation — a short delay while the model fetches more data
- DA warning the model prints — a notice it flags whenever unsure of the facts
What are the two main phases of an LLM's lifecycle?
- ATraining and inference — first it learns from data, then it generates output
- BEncoding and decoding — first it compresses text, then it expands it back
- CIndexing and retrieval — first it stores facts, then it looks them up later
- DCompiling and running — first it builds the code, then it executes the code
Learn it once, keep it for good
Answer a question
Each card is one practical concept with multiple options. Pick what you think is right.
Get the full answer
See the correct option plus a clear explanation, and a link to deeper docs when one is available.
Review at the right time
A spaced-repetition engine (SM-2 or FSRS) resurfaces each card just before you would forget it.
Why understanding AI is worth your time
AI is in everything now
From editors to search to customer support, LLMs are everywhere. Knowing how they work makes you better at using them.
Prompt with intent, not superstition
Once you understand tokens, context and how models predict, you stop guessing at what makes a prompt work.
Spot the limits
Knowing why models hallucinate and where the knowledge cutoff bites helps you trust output appropriately.
Foundation for building
These fundamentals are the groundwork for prompt engineering and building real apps on top of models.
Common questions
Do I need a technical or math background? +
No. The whole track is in plain English and explains every concept from first principles — no coding or advanced math required.
How long does it take? +
About 10 minutes a day. Spaced repetition means short, frequent sessions beat long cramming — most learners grasp the fundamentals within a couple of weeks.
Is it free? +
Yes, completely free. No registration or credit card is required, and all your progress is stored locally in your browser.
Will this teach me to write prompts or build AI apps? +
This track covers the fundamentals of how models work. Writing prompts and building apps are their own tracks (Prompt Engineering and Building with LLMs) that build on what you learn here.
Ready to understand AI for real?
Plant your first seed today. Ten minutes a day is all it takes to grow a real, lasting understanding of how AI works.