The Three Levels

Before we start clicking buttons, a simple map. AI usage roughly comes in three levels. Today we're going to move you from Level 1, where almost everyone is, up to Level 2. Level 3 is for much later, maybe months from now. It'll make more sense if you know where you are in the picture.

Here's what today looks like in one sentence: we're getting you set up on the tool you choose, ChatGPT, Claude, or Gemini, and building your first Project, which is Level 2.

That's the whole job for this guide. Everything from here to the end is just getting you there, one page at a time.

Level 1: The Chatbot

A chat website. You open chat.openai.com, or claude.ai, or gemini.google.com. You type a question in the box. You get an answer. You close the tab.

This is probably how you've been using AI so far. Maybe you've had a few good experiences: the AI helped you draft something, or explained a word, or cleaned up an email. Useful.

Good for one-off, self-contained tasks:

• A student asking it to explain a confusing concept in plain English, or to check whether an argument in a draft essay holds together.
• Someone job hunting tightening one paragraph of a cover letter, or rehearsing answers to a common interview question.
• A teacher brainstorming three ways to explain fractions, or turning a dense paragraph into a warm-up question.
• A freelancer rewriting a single email to sound warmer or shorter, or unsticking a headline they've been staring at.
• A hobbyist asking for a substitute for buttermilk, or how to fix a wobbly table leg.
• Anyone decoding a bit of jargon: a legal clause, a medical term, a news story, a word from a meeting.

What it can't do:

• It forgets everything the moment you close the tab. The next conversation starts from scratch, like you've never spoken before.
• It has no idea who you are or what you're working on. Every answer is built from guesses about a generic person.
• Because of that, the answers sound generic. Because they are.

This is where most people live. It's useful in the way that Google is useful, but it stops short of anything that would meaningfully change how you work.

Level 2: Projects (this is what we're setting up today)

Still a chat. Same interface. But now the chat lives inside a workspace that has a permanent memory of you and whatever you're working on.

Here's how it works. You write a few short files: about you, about the thing you're working on, samples of how you like to write. You upload them into a workspace. From that moment on, every single chat you start inside that workspace already knows all of it.

You don't re-explain yourself. You don't paste the same paragraph at the top of every conversation. The AI just knows.

Here's what's really happening under the hood. Levels 2 and 3 both work the same trick: they keep your real information sitting in the AI's working memory, the stuff actively in front of it, so it never has to fall back on vague guesses about a generic person. That's the entire difference. At Level 1 the AI is recalling averaged internet patterns; at Level 2 it's reading your facts. Same AI, far better answers, purely because the setup stopped making it guess.

What changes:

• Answers are tied to your real situation: your actual documents, your real numbers, your real voice.
• You can ask questions that only make sense in context. "Given everything above, what am I missing?" can now be answered instead of deflected.
• Writing in your voice stops being a chore. The AI already has samples of how you sound.

All three tools have a version of this. The names differ; the idea is identical:

Total time to set one up: under an hour, spread across a few pages. After that, your day-to-day use of AI looks completely different. This is the one step that explains almost all of the gap between "AI is kind of useful" and "AI actually changed how I work."

What Level 2 makes possible, across all kinds of work:

• A student loads their course notes, the assignment brief, and a marking rubric into a workspace, then asks it to check a draft against the rubric or quiz them before an exam, grounded in the actual course, not the internet's idea of it.
• A job hunter keeps their resume, a few target job descriptions, and stories from past roles in one place, and every tailored cover letter comes out sounding like them and pointed at that specific role.
• A teacher keeps their curriculum, their students' reading level, and their own worksheet style loaded, so lesson plans and handouts come back ready to use instead of generic.
• A freelancer or small shop owner loads their voice samples, service list, and past client emails, and gets newsletters, replies, and pitches that sound like them, plus a place to pressure-test a new idea against their own notes.
• A hobbyist building something, a garden plan, a D&D campaign, a home renovation, keeps all their notes in one workspace and asks questions that build on everything so far.
• Anyone with their own data. Drop in a budget and a few months of spending and ask where the money went; upload a report or a long PDF and get answers that quote the actual document, nothing invented.

Level 3: The AI Work Partner (come back in a few months)

Level 3 is when AI stops just answering questions and starts producing real work: files, documents, even working websites. This is where the words you may have heard, Claude Code, Cowork, CLI, codebase, come in. All three tools have a version of this; Claude's happens to have the clearest names, so we'll use them as the example.

Two shapes are worth knowing about:

• A side-by-side document mode (Claude calls it Cowork; ChatGPT and Gemini both have a "canvas"), a split screen where the AI drafts in a document pane while you edit alongside it, like two people in the same Google Doc. Good for longer writing where back-and-forth chat feels clunky.
• A tool that works on real files on your computer (Claude Code is the clearest example), a bigger leap. The AI runs on your machine and can create, edit, and organize actual files in a folder. You can say things like "take these text files and build me a simple website" and watch the files appear.

Both change what AI is: not just a thing you chat with, but a thing that makes stuff.

Why you're not doing this today: Level 3 takes a little more setup, a little more confidence, and a little experience with Level 2 first so you know what problems you want solved. People who skip straight to Level 3 tend to get overwhelmed. The people who get the most out of Level 3 are the ones who used Level 2 for a while first.

Later in the guide there's a gentle preview of the words you'll see at Level 3, codebase, IDE, CLI, so they're not intimidating when you come back to them. You can skim it, or even skip it. But it's there for when you're ready.

What Level 3 makes possible, across all kinds of work:

• A student hands over a messy CSV of survey results and asks for the actual statistics and a chart, real math, not a guess, or has it build a small study-tracker page.
• A hobbyist turns a folder of notes into a simple website: a recipe collection, a travel log, a portfolio, without hiring anyone or touching a template.
• Anyone with a mess of files, photos, downloads, scanned documents, has the AI rename them, sort them into folders, cull duplicates, or convert them in one pass.
• Someone doing their taxes turns a pile of receipts or a bank export into one clean yearly summary.
• A freelancer or small shop owner exports a year of sales or booking data and asks for the trends and the single most profitable thing they do, or bulk-rewrites forty old captions in a fresher voice in a single pass.
• Small automations. A quick tool that takes an intake form and drafts a standard reply, saving the same fifteen minutes every time.

Why the goal is to be fluent in both Level 2 and Level 3

Level 2 and Level 3 are not competitors. They do different jobs.

Level 2 is for thinking, writing, research, and strategy. When you have a question about your work or your life and you want a thoughtful, grounded answer, that's a Project. Most days, most of the time, this is where you'll live.

Level 3 is for making. When the job isn't "advise me" but "build it for me," real files, real websites, real analysis on real data, real automation, that's Level 3. Fewer days, but worth far more on the days you need it.

Together they cover almost everything. Level 2 on its own is transformative. Level 2 plus Level 3 is a different kind of life, one where large, intimidating tasks (building a website, analyzing a year of data, rewriting a pile of old work) stop being projects you put off and start being afternoons you actually finish.

You don't have to be fluent in both on day one. Start at Level 2 this week. Live there for a while. Come back to Level 3 when you hit something Level 2 can't do, and you will. That's the natural path.

At a glance

Level 1, Chatbot. Open a website, ask a question, close the tab. Forgets you between every conversation. Where most people stop.

Level 2, Projects. A chat that remembers you and your work. Set up once, use forever. Today's work.

Level 3, Work partner. AI that edits documents alongside you and makes real files on your computer. Come back in a few months.