By Lymi

AI Terms, Plainly Explained

Learn 25 everyday AI terms first, then explore 11 deeper ideas about models and connected tools.

Technology36 cards

Add to LymiTry a few cards

Lymi is a flashcard app: it brings each card back right before you’d forget it. Free to use.

  • Start here

    Machine learning

    A way to teach software to find patterns in data instead of giving it every rule.

  • Using AI

    A small piece of text that a language model reads or writes.

    Token

  • Tools and agents

    Memory

    Information an AI app saves to use again later.

  • Go further: how models work

    Showing a model a few examples before asking it to do a similar task.

    Few-shot prompting

  • Go further: building with AI

    Evals

    Tests that check how well an AI system does specific tasks.

What’s inside

See all 36 cards

Every section is open, so any card can come up in review.

  1. Section 1: Start here

    7 cards

  2. Section 2: Using AI

    6 cards

  3. Section 3: When to trust an answer

    6 cards

  4. Section 4: Tools and agents

    6 cards

  5. Section 5: Go further: how models work

    5 cards

  6. Section 6: Go further: building with AI

    6 cards
Add to Lymi

All 36 cards

In the order you learn them, one section at a time.

1Start here7 cards

Artificial intelligence (AI)
Software that can do tasks usually linked to human thinking, such as recognizing patterns or using language.
Machine learning
A way to teach software to find patterns in data instead of giving it every rule.
Model
A computer system trained on data to produce answers or predictions.
Generative AI
AI that creates text, images, sound, code, or other content from a request.
Large language model (LLM)
A model trained on lots of data to work with language.
Training
Teaching a model by changing it as it learns from data.
Inference
Using a trained model to answer a new question or make a prediction.

2Using AI6 cards

Prompt
What you give a model to tell it what to do.
Prompt engineering
Writing and improving prompts to get more useful answers.
Response
What a model gives back after receiving an input.
Token
A small piece of text that a language model reads or writes.
Context
The information a model has for its current answer.
Context window
The maximum amount of information a model can use in one request, measured in tokens.

3When to trust an answer6 cards

Hallucination
An AI answer that sounds believable but states something false.
Grounding
Using supplied or found information as the basis for an answer.
Knowledge cutoff
A date that roughly marks how recent a model’s training information is.
Bias
A pattern of skewed or unfair results from an AI system.
Prompt injection
An attempt to trick an AI assistant into following instructions it should not follow.
Human in the loop
A person checking, guiding, or approving part of an AI system’s work.

4Tools and agents6 cards

Multimodal
Able to work with different kinds of information, such as text, images, and sound.
Tool
A separate service or function an AI system can use to get information or do something.
Tool call
A request from a model to use a tool.
Agent
An AI system that can choose steps and use tools to complete a task.
Workflow
A sequence of steps a system follows to complete a task.
Memory
Information an AI app saves to use again later.

5Go further: how models work5 cards

System instruction
A direction from an app that guides how its AI assistant responds.
Few-shot prompting
Showing a model a few examples before asking it to do a similar task.
Temperature
A setting that changes how varied a model’s answers can be.
Dataset
A collection of examples used to train or test a model.
Parameter
A value inside a model that it learns during training.

6Go further: building with AI6 cards

Model Context Protocol (MCP)
A standard way for AI apps to connect to outside data and tools.
Embedding
A list of numbers that represents content so software can find related things.
Semantic search
Search that finds results by meaning, even when the words differ.
Retrieval-augmented generation (RAG)
Finding relevant information and giving it to a model before it answers.
Fine-tuning
Training an existing model further for a particular kind of task.
Evals
Tests that check how well an AI system does specific tasks.

How it works

  1. Add the deck

    Sign in with Google and it joins your Library, free.

  2. Recall, then turn it over

    Try to remember the answer before you look. Try it on the cards here: they’re from all over the deck.

  3. Review a few minutes a day

    Choose how many cards a day. Lymi brings each one back right before you’d forget it.

Start here

Large language model (LLM)

Published September 25, 2026