Working with AI

27 terms in this category.

agent
An AI that doesn't just answer — it takes actions, like reading files, running commands, or browsing the web, usually within limits you set.
Anthropic
The company that makes Claude.
autonomy
How much the AI is allowed to do on its own without asking you first.
ChatGPT
The consumer chat app from OpenAI, built on their GPT models.
Claude
An AI assistant made by Anthropic. One of the main AI tools people use for writing, coding, and reasoning through problems.
Claude Code
also: CC
A version of Claude that lives in your terminal and can read, write, and run code on your computer.
context window
also: context
How much text the AI can hold in its attention at once in a single conversation.
embedding
A way of turning a piece of text into a list of numbers that captures its meaning, so a computer can compare it to other pieces of text.
evals
also: evaluations
Tests that check whether an AI system is actually doing its job well.
fine-tuning
Taking an existing AI model and training it a bit more on your own examples, so it behaves the way you want.
GPT
A family of AI models made by OpenAI. ChatGPT is the best-known product built on them.
guardrails
Rules or limits that stop an AI system from doing things it shouldn't.
hallucination
When an AI confidently says something that isn't true.
inference
What's happening when the AI is actually answering you, as opposed to when it was being trained.
LLM
also: large language model
The kind of AI that Claude, GPT, and Gemini all are — trained on huge amounts of text to predict what comes next.
MCP
also: Model Context Protocol, MCP server
A shared standard that lets AI tools plug into other software — so an AI can read your email, update your calendar, or open your files without being hand-wired into each one.
model
The specific AI doing the thinking in a given product — Claude, GPT-5, Gemini, and so on.
OpenAI
The company that makes ChatGPT and the GPT models.
prompt
What you type to an AI to tell it what you want.
prompt engineering
The craft of writing prompts that reliably get good results.
RAG
also: retrieval-augmented generation
A technique where the AI searches your documents for relevant bits and reads them before answering, so it can speak about your specific stuff — not just what it learned in training.
system prompt
A set of instructions that shape how the AI behaves for an entire conversation, before you've said anything.
temperature
A setting that controls how much the AI varies its answers — low temperature for consistent, predictable replies; high temperature for more variety and surprise.
token
A small piece of text that an AI reads and writes in — sometimes a whole short word, sometimes a few characters, sometimes a chunk of a longer word.
tool use
When an AI reaches for a specific tool — like a calculator, a search engine, or your file system — instead of just talking.
training
The long, expensive process of teaching an AI by feeding it huge amounts of text or data before anyone ever uses it.
vector database
also: vector DB
A database built to store embeddings and find the closest matches quickly.