Every one of these gets thrown around like you’re supposed to know it. Here they are in plain English, each with a real example, so they finally stick.
Everyone in AI talks like the words are obvious. They aren’t, and you don’t have to fake it anymore. Read this once, save it, and come back whenever a word trips you up.
01
LLM (Large Language Model)
The brain behind the chatbot. It read a huge amount of text and learned to do one thing really well: predict the next word. When you talk to Claude, ChatGPT or Gemini, you’re talking to an LLM.
Why it matters: think of it as the world’s most well-read autocomplete. Once that clicks, a lot of AI stops feeling like magic.
02
Prompt
What you type to the AI. The more specific, the better. "Write something about marketing" gets generic junk. "Write a 3-sentence Instagram caption for a life coach who helps burnt-out 9-to-5 workers, casual tone, end with a question" gets something you can post.
Why it matters: most bad AI answers come from vague prompts.
03
Tokens
Tokens are how AI reads. It sees chunks instead of whole words, so "basketball" might be split into "basket" and "ball." Usage limits and API prices are counted in tokens.
Rough math: a token is somewhere between half a word and three-quarters of a word, depending on the model. On Claude’s current models, 1 million tokens is about 555,000 words. (Anthropic)
Why it matters: when you hear "token limit" or "price per million tokens," this is what they mean.
04
Context Window
The AI’s short-term memory for one conversation: how much it can hold in its head at once. All of Claude’s current models (Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5) have a 1 million token window. On paid Claude plans with code execution turned on, when a chat gets close to the limit, Claude summarizes the older messages to make room. (Claude Help Center)
Why it matters: if a long chat starts feeling forgetful or off-track, start a fresh chat with a short recap. It’s sharp again.
05
Hallucination
When AI makes something up and says it like it’s 100% fact. It’s guessing, and it sounds confident while it guesses. Example: the Eiffel Tower was finished in 1889 for the Paris World’s Fair. If an AI confidently gives you 1895 or invents an engineer, that’s a hallucination.
Why it matters: AI is a great first draft and a bad final source. Double-check names, dates, numbers and quotes.
06
Machine Learning
How AI learns from examples instead of from rules someone typed in. Feed it thousands of emails labeled spam or not spam, and it figures out the signals on its own, like all-caps subject lines and weird links.
Why it matters: nobody wrote those rules. The AI found the patterns from the examples.
07
MCP / Connectors
The plugs that let AI reach into your other tools, like email, calendar, files and Slack. MCP (Model Context Protocol) is the open standard Anthropic created so AI apps can connect to tools the same way. In Claude, you add connectors from the Customize page in the sidebar, which is being renamed Plugins.
Why it matters: once AI can see your real inbox and calendar, it goes from chatbot to assistant.
08
Generative AI
The catch-all term for AI that makes new stuff, like text, images, video, music or code. If it creates something instead of only sorting what’s already there, it’s generative.
Why it matters: it’s the umbrella word for most of the AI you hear about.
09
AGI (Artificial General Intelligence)
The idea of an AI that can do basically any thinking task a human can, including things it was never trained for. There’s no agreed definition, and people argue a lot about whether and when it will happen.
Why it matters: when people argue online about AI "taking over," this is usually what they mean. It’s a debate about the future, and the tool you use today is something else.
10
Vibe Coding
Building software by telling AI what you want in plain English instead of writing the code yourself. You describe it, the AI writes it, you test it and ask for changes. Andrej Karpathy coined the term in February 2025. (Wikipedia)
Why it matters: it’s why total beginners are shipping real apps. You need a clear idea and the words to describe it.
New terms show up every week. Paste this into Claude or ChatGPT with the word and where you saw it, and you’ll get the like-you’re-10 version plus a quick quiz.