Get Started With AI — Open the Big Three and Learn How They Actually Work
What this project is
Open free accounts on the three main AI tools, switch off the settings that let them train on you, and have one short guided conversation that leaves you actually understanding what these tools are and why they can do different things. This is the foundation for every other mission.
Capabilities gained
- Getting set up on the major AI tools with a single Google login each
- Controlling your privacy — turning off model training and knowing what that does and doesn't do
- Understanding the mental model behind every AI tool — a model plus a harness — so you can tell what any platform can really do
Where else this applies
- Deciding which AI tool to reach for on a given job
- Judging whether a new "AI app" is actually special or just a familiar model in a nicer wrapper
- Keeping sensitive work out of tools that would train on it
Copy-paste build prompt
You are my AI tutor and I'm a complete beginner. Teach me the basics in a quick, guided way — one idea at a time, in plain language, no jargon, just a few sentences each. After each idea, ask me one simple question to check I got it before moving on. Don't dump everything at once and don't go off on tangents. Cover these in order: 1. What a large language model (LLM) actually is — the "brain": I feed it words, it predicts words back, and it only knows what it was trained on plus what I tell it in the chat. 2. What a "harness" is — the platform wrapped around the model (Claude, ChatGPT, the Gemini app). The same kind of model can live in different harnesses, and the harness is where the extra powers get added. 3. The extras a harness can add — one line each: tools/capabilities (like web search or image generation), connectors (linking it to your Gmail, Drive, etc.), memory (remembering things across chats), and skills (reusable instruction packs that give it a repeatable ability). 4. Why the same underlying model can do more on one platform than another — because of the harness, not the brain. Keep the whole thing short and beginner-friendly. Start with idea 1.
Copy-paste explanation prompt
Now quiz me. Ask me four quick questions, one at a time, to check I understood: what an LLM is, what a harness is, why the same model can do different things on different platforms, and one example of an "extra" a harness can add. After each answer, tell me if I got it and briefly fix anything I got wrong.
What’s going on
- You open free accounts on Claude, ChatGPT, and Gemini with your Google login
- You try a quick image generation in Gemini to see a "capability" in action
- You turn off the setting that lets each tool train on your chats
- You skim the settings so you've at least seen what's there
- You run one guided conversation that teaches you what these tools actually are, then quiz yourself
Step-by-step instructions
Step 1 — Open the three tools and sign in with Google (one login each): Claude at claude.ai, ChatGPT at chatgpt.com, Gemini at gemini.google.com. Click "Sign in with Google" (or "Continue with Google") on each and pick your account.
You should now see three AI chat windows, each logged in and ready.
Step 2 — See a capability in action. In Gemini, type "Generate an image of a golden retriever wearing sunglasses on a beach" and send.
You should now see an actual image appear. (Try the same request in Claude and it'll tell you it can't make images — that's your first lesson that different tools have different powers.)
Step 3 — Lock down privacy. The three tools handle this differently (and these settings move — if a label differs, look for the "improve the model" or "activity" wording):
Claude → Settings → Privacy → turn off "Help improve Claude." A dedicated training switch.
ChatGPT → Settings → Data Controls → turn off "Improve the model for everyone." Also a dedicated training switch. (Heads-up: giving a thumbs-up or thumbs-down on a reply can still send that one chat in.)
Gemini → tap your profile picture → "Gemini Apps Activity" (being renamed "Keep Activity"). This one's different: it's really a "save my chats" switch. Turning it off stops Google saving your chats, and as a result they're not used to train future models. But it has a cost — you lose your Gemini chat history and Gemini's links to Gmail, Drive, and Calendar stop working. So this one's your call: turn it off for maximum privacy, or leave it on to keep those features.
You should now see training off in Claude and ChatGPT, and a clear choice made in Gemini. Two honest caveats for all three: this only affects chats going forward — it doesn't erase or "untrain" what's already there — and turning it off isn't the same as deleting your data.
Step 4 — Skim the settings in each tool for about a minute. Don't try to understand everything — just scroll through so you've seen what's there.
You should now have a rough sense of what each tool lets you control — history, connectors, memory, account.
Step 5 — The important part: open Claude, paste the build prompt above, and go through the guided lesson one idea at a time.
You should now see Claude teaching you a single idea at a time and checking you followed — by the end you'll have the model-plus-harness picture.
Step 6 — Paste the explanation prompt and quiz yourself.
You should now be able to say, in your own words, what an LLM is and why the same model can do different things on different platforms. That's the whole foundation — you're ready for the other missions.