the ai glossary

AI words, in plain English.

The AI world loves its jargon. Every headline, podcast, and sales pitch assumes you already speak it. Here's the translation: the words you'll actually run into, defined the way we'd explain them across a table.

Skim it, search it, or come back when a word stumps you. We add words as the AI world coins them.

The basics

AI (artificial intelligence): Software that can do things that used to need a human brain: writing, understanding language, recognizing images, making decisions. When people say AI today, they usually mean tools built on large language models.
LLM (large language model): The engine behind tools like ChatGPT and Claude. A huge piece of software trained on massive amounts of text until it got good at predicting words. That turns out to be enough to write, summarize, translate, and reason.
Model: One specific AI brain, released under a name and version. Different models have different strengths, speeds, and prices, which is why picking the right one matters.
Chatbot: AI you talk to in a chat window. You ask, it answers, the conversation goes back and forth. The help bubbles on websites are chatbots, and so is ChatGPT in its basic form.
AI agent: AI that does more than answer. You give it a goal and it takes real actions on a computer: organizing files, drafting documents, building things, checking its own work. The step beyond chatbots, and the reason AI suddenly matters for regular work.
Generative AI: AI that creates new things (text, images, music, video, code) instead of just sorting or ranking existing ones. The 'generative' is why it can write your email instead of only filtering your spam.
Machine learning: How AI gets good at things. Instead of a programmer writing rules, the software studies huge piles of examples until it finds the patterns itself.
Training: The months-long, expensive process of feeding a model examples until it learns. Done by AI companies on giant computer clusters, long before you ever type a question. You use the finished result.

Talking to AI

Prompt: Whatever you type or say to an AI. The clearer the prompt, the better the answer: 'write a friendly but firm follow-up to a customer who missed their appointment' beats 'write an email'.
Prompt engineering: The skill of writing prompts that get great results. Less magic than it sounds: mostly being specific, giving examples, and saying who the answer is for.
Context: Everything the AI can currently see: your conversation so far, files you shared, instructions it was given. AI only knows what's in its context, so the more relevant material you give it, the better it does.
Context window: The AI's short-term memory. It can only hold so much at once; go past the limit and the oldest parts fall out of view. A bigger context window means it can read longer documents and follow longer conversations.
Token: The small chunks AI reads and writes text in, usually a word or piece of a word. AI pricing and limits are counted in tokens, which is why the word shows up on every AI bill and plan.
Hallucination: When AI states something false with total confidence. It isn't lying on purpose; it's filling a gap with a plausible-sounding guess. The fix is simple: double-check anything that matters.
System prompt: Hidden instructions an AI is given before you start typing: who it is, how to behave, what to refuse. It's why the same model can act like a lawyer's assistant on one site and a cooking coach on another.
RAG (retrieval-augmented generation): A fancy phrase for a simple idea: let the AI look things up in your documents before it answers, so it responds from your facts instead of its memory.
Multimodal: AI that handles more than text: images, audio, or video, in or out. Snapping a photo of a broken part and asking what it is: that's multimodal.
Prompt injection: A trick where text hidden in a webpage, email, or document tries to slip new instructions to an AI. It's why well-built AI tools ask for your approval before doing anything risky.

Building with AI

Vibe coding: Building software by describing what you want in plain English and letting AI write the code. You steer by reacting: ask, look, adjust, repeat. People with zero programming background ship real websites and tools this way.
Iterate: To improve something in rounds instead of nailing it on the first try. Working with AI is iteration by nature: ask, check the result, ask for changes, repeat until it's right. It's the normal workflow, not a failure.
API: How software talks to other software. When an app quietly asks an AI model for an answer behind the scenes, it's 'calling the API'. The phrase just means 'our tool asks their tool'.
API key: The secret password a tool uses to prove it's allowed to use an API, and to get billed for it. Treat it like a card number: never share it, never post it.
MCP (Model Context Protocol): A standard way to plug AI assistants into other tools and data: your calendar, your files, a database. Think of it as a universal connector, USB for AI.
Terminal: The text-only window where you type commands to your computer instead of clicking buttons. It looks like hacker television, but it's really just texting with your computer, and many AI tools live there.
GitHub: The website where the world stores and shares code. Think Google Drive, but for software projects, with a saved history of every change ever made. 'It's on GitHub' means the code is there to see.
Repository (repo): One project's home on GitHub or a similar service: all its code, files, and complete change history in one place.
Open source: Software whose code is public for anyone to use, inspect, or improve, usually free. A huge share of the AI world is built on it.
Automation: Making a task run by itself: the invoice goes out, the reminder texts, the report builds, without you touching anything. AI made automations far easier to set up than they used to be.
Workflow: A chain of automated steps: a form comes in, the data gets logged, a confirmation goes out. Tools like Zapier, Make, and n8n exist to wire these together.
No-code: Building apps and automations with visual tools instead of programming. AI pushed it further: now you can often just describe what you want.
Deploy: To put a website or app live on the internet. 'We deployed it' means real people can use it now.
The cloud: Other people's computers, rented over the internet. When your files or website live 'in the cloud', they sit in a professional data center instead of on your machine.
Bug: A mistake in software that makes it misbehave. Finding and fixing them is half of building anything, and AI is now good at both causing and catching them.

Words you'll hear online

AI slop: The flood of low-effort, AI-generated content clogging the internet: generic articles, six-fingered images, spam videos. It's why 'made with AI' sometimes has a bad name. The difference between slop and good work is human taste and editing.
Deepfake: An AI-made photo, voice, or video of a real person doing or saying something they never did. Getting easier to make and harder to spot, so treat shocking clips with healthy suspicion.
AGI (artificial general intelligence): Hypothetical AI as capable as a human across the board, not just at specific tasks. It doesn't exist yet, and smart people disagree loudly about when or whether it will.
GPT: Short for 'generative pre-trained transformer', the technology family behind ChatGPT. In everyday talk, people use it to mean ChatGPT-style AI in general.
Bot: Any program or account that acts on its own. Some are helpful (booking bots, reminder bots), some are pests (spam bots). AI made both kinds smarter.
Jailbreak: Tricking an AI into ignoring its safety rules. AI companies patch against it constantly; you'll mostly meet the word in news stories.

Names you'll keep hearing

ChatGPT: OpenAI's chatbot, the product that kicked off the current AI wave in late 2022 and became the household name for AI.
Claude: Anthropic's AI assistant, widely known for natural writing and strong coding ability.
Gemini: Google's AI assistant, woven into Search, Gmail, Android, and the rest of the Google world.
Copilot: Microsoft's AI assistant, built into Windows, Office, and its coding tools.
Perplexity: An AI search engine: ask a question, get a direct answer with sources cited, instead of a page of links.
Midjourney: One of the best-known AI image generators, famous for striking, artistic results.
ElevenLabs: AI voice generation: turn text into realistic speech, or clone a voice (with permission) for narration.
NotebookLM: Google's research tool: feed it your documents and it summarizes, answers questions, and can even turn them into a podcast-style audio discussion.
Suno: AI that makes complete songs, vocals included, from a text description.

Words are the easy part. Seeing these tools work on your own screen is what makes it click, and that's exactly what the free hour is for.

ready when you are

The first hour is free. Bring a real problem.

Two minutes of quick questions, then you pick the time. No pitch, no jargon: the hour happens on your real stuff, and you keep whatever we set up.

Why free? We'd rather prove it than pitch it. And it won't be free forever: as client work fills the calendar, the free hour goes away (a session you've already booked stays free). After that it's $50 an hour online, or $100 for a 90-minute in-person visit. That's low for the Bay Area on purpose: the AI does the heavy lifting, and we want the work to speak first. Either way you keep everything we set up, most tools we recommend are free or cheap, and you'll know the price of anything before we touch it.

Book your free 1-hour AI setup session

Prefer to talk to a person first? The first 30-minute phone call is free. Text us: +1 650 679 0216

Want to know who you're meeting first? Meet your consultant