AI, in plain English.
A quick decoder for the words everyone keeps throwing around. No PhDs. No jargon.
AI (Artificial Intelligence)
Software that can do things that used to need a human brain.
An umbrella term for any computer system that can do tasks like understanding language, recognizing images, making decisions, or generating content. Most of what people call “AI” today is one specific flavor: large language models (see LLM).
Example: Your email's spam filter is AI. So is ChatGPT. So is Netflix's recommendations.
LLM (Large Language Model)
The brain behind ChatGPT, Claude, Gemini, etc.
An AI trained on massive amounts of text that learned to predict what word comes next — really, really well. That simple trick is what powers conversation, writing, coding, summarizing, and everything in between.
Example: GPT-5, Claude, Gemini, and Llama are all LLMs.
AI Assistant
An AI you chat with to get stuff done.
A general-purpose helper. You type or speak, it answers. It waits for your next message. It does what you ask, then stops. ChatGPT, Claude, and Gemini are assistants.
Example: “Help me write a polite email to cancel a meeting” → it writes one. Done.
AI Agent
An AI that takes actions on its own to finish a goal.
Same brain as an assistant, but with hands. You give it a goal and it plans steps, uses tools (browser, email, calendar, code), and keeps going until the job is done — without you babysitting every step.
Example: “Book me a dentist next Tuesday afternoon and add it to my calendar” — and it actually does it.
Prompt
What you type into the AI.
Your instructions. The clearer and more specific the prompt, the better the output. 80% of getting good results is writing a good prompt — which is exactly what this course teaches you.
Example: “Write a 3-line birthday card for my dad who loves fishing” is a prompt.
Context
The background info you give the AI before asking.
Who you are, who it's for, what's happened so far, what tone you want. AI without context guesses. AI with context delivers.
Example: Telling it “I'm a busy mom of two and I hate spicy food” before asking for dinner ideas.
Context Window
How much the AI can remember in one conversation.
Measured in tokens (roughly chunks of words). Bigger window = you can paste more (long docs, whole emails, entire chapters). When you exceed it, the AI starts forgetting earlier parts.
Example: Claude can hold a 500-page book. ChatGPT free tier holds much less.
Token
How AI counts words.
A token is a chunk of text — sometimes a whole word, sometimes part of one. Roughly 1 token ≈ ¾ of a word in English. Pricing and limits are usually measured in tokens.
Hallucination
When AI confidently makes stuff up.
Because LLMs predict plausible-sounding words, they can invent fake facts, fake quotes, fake citations — and sound 100% sure about it. Always sanity-check important facts.
Example: Asking for a book recommendation and getting a real author paired with a book they never wrote.
Model
A specific version of an AI brain.
Like an iPhone has different models, AI does too. Different models have different strengths (writing, reasoning, speed, cost). You'll hear names like GPT-5, Claude Opus 4, Gemini 2.5 Pro.
Multimodal
AI that handles more than just text.
It can see images, hear audio, read PDFs, analyze video. You can show it a photo of your fridge and ask what to cook.
Example: Snap a picture of a confusing form, ask AI what it means.
Generative AI
AI that creates new stuff.
Text, images, video, audio, code — generated from your prompt. The opposite is AI that just classifies or analyzes. ChatGPT, Midjourney, Sora, and ElevenLabs are all generative AI.
Vibe Coding
Building software by describing the vibe — not writing code.
You tell an AI what you want (“a website where my friends vote on weekend plans, make it feel cozy”), and it writes the actual code, runs it, and fixes its own bugs. You stay in the driver's seat by reacting to what you see, not by knowing syntax.
Example: Lovable, v0, and Bolt are vibe-coding tools. You're using one right now.
Custom GPT / Project
A reusable AI helper trained on your stuff.
Save instructions, files, and a personality once. Use it forever. Like having a specialist assistant for one job (your resume, your business, your writing voice).
Example: A “Reply Like Me” GPT that knows your tone and writes emails that sound like you.
Fine-tuning
Teaching an existing AI your specific style or knowledge.
More technical than a Custom GPT — you actually retrain a model on your data. Usually overkill for everyday users; Custom GPTs and Projects are easier and free.
RAG (Retrieval-Augmented Generation)
AI that looks things up before answering.
Instead of relying only on its memory, the AI searches a set of documents (or the web) and uses what it finds to answer. That's how Perplexity and NotebookLM stay grounded in real sources.
API
A way for apps to talk to AI directly.
Stands for Application Programming Interface. You won't touch APIs as a user, but it's how every AI tool you love (and the next one you'll build) connects to models behind the scenes.
Open Source AI
AI models anyone can download and run.
Companies like Meta (Llama) and Mistral release the model itself, free. Companies like OpenAI keep theirs locked up and charge per use.
Chain of Thought
Asking AI to “think step-by-step” before answering.
A simple trick that dramatically improves answers on hard problems. New models (like reasoning models) do this automatically.
Automation
Wiring AI into apps so things happen without you.
Tools like Zapier, Make, and n8n let you connect AI to email, calendars, spreadsheets, and 7,000+ other apps. We get into this in Day 27.
