Appendix A — Setting Up Your AI Toolkit

“Prepare your work outside; get everything ready for yourself in the field, and after that build your house.” — Proverbs 24:27 (ESV)

Before the first lab, you need a working toolkit. Here is the good news, and I want you to feel it before you read another line: you do not need to buy anything, install anything, or ask your IT department for anything. If you have a web browser and an email address, you already have almost everything this whole course requires. A phone works too.

I know some of you arrived nervous. You’ve heard AI is expensive, or that it needs a powerful computer, or that you have to be technical to touch it. None of that is true for this class. We are going to walk the cheapest path that works, on purpose, and I’ll be honest with you at every step about what — if anything — costs money.

One rule governs everything you’re about to set up. Write it where you’ll see it:

You choose the tool. You own the verdict. AI drafts, generates, and accelerates; you decide, verify, and are accountable. AI is an assistant, not an authority.

The tools below are assistants. This appendix gets them working. The judgment stays with you.

Coach’s Note — Everything in Sections A.1 through A.4 is free, needs no install, and needs no admin rights — it runs on a school Chromebook, a borrowed laptop, or your phone. Only the optional private-computer path in A.5 installs software, and even that has a free cloud fallback. Nobody in this room is priced out. Start at the top and stop when you have what your next lab needs.

A.1 — What You Actually Need

The backbone of this entire course is three free chatbots in a browser. That’s it. Most weeks use one of them. A few weeks add a specialized free tool, and exactly one week offers an optional download for people who want to run AI privately on their own machine.

Here is the whole landscape, honestly:

You will never be required to pay for a subscription in this course. Free tiers are rate-limited — they slow you down or cap you after a burst of use — but they are more than enough to learn on. As of mid-2026 the free tiers of the three big assistants are genuinely capable; when you hit a limit, wait a bit or switch to another free tool. I’ll show you how to tell what each lab needs in A.4.

Coach’s Note — A word before you type anything into any of these: never paste a client’s private data, a patient record, a student’s information, a password, or anything confidential into a free consumer chatbot. Free tiers may use your input to improve the model. This matters enough that a whole appendix is devoted to it — read Appendix C — Responsible & Ethical AI Use before Week 6. For now: practice on public or made-up information.

A.2 — Create Your Three Free Accounts

You’ll set up all three market leaders. Why all three? Because Week 2’s lab (Three Models, One Prompt) sends the same prompt to all three and compares the answers — and because “which tool do I reach for?” is a skill you can only build by having more than one on hand. This is the spine rule in practice: you can’t choose the tool if you only own one.

All three are free to start, need no credit card, and work the same on web and phone.

1. ChatGPT (OpenAI)

• Web: go to chatgpt.com and sign up with an email, Google, Microsoft, or Apple account.
• Phone: install the free ChatGPT app from the App Store or Google Play; sign in with the same account.

2. Claude (Anthropic)

• Web: go to claude.ai and sign up.
• Phone: install the free Claude app; sign in with the same account.

3. Gemini (Google)

• Web: go to gemini.google.com and sign in with any Google account (a free Gmail account is enough).
• Phone: use the Gemini app, or Gemini inside the Google app; sign in with the same Google account.

Verify each one works — with one prompt

Don’t just create the accounts. Prove they answer. Open each of the three, one at a time, and paste in the exact same sentence:

Explain what a large language model is to a busy professional, in three sentences, with no jargon.

You’re checking two things. First, that each tool responds — if you get three answers, your toolkit is live. Second, and more interesting: notice they don’t answer identically. One is warmer, one is tighter, one adds an example. Keep those three answers. That side-by-side is exactly the muscle Week 2 trains, and you just did your first rep.

Coach’s Note — If you want to confirm which tool you’re actually talking to, ask it: “In one sentence — what are you, and who made you?” ChatGPT names OpenAI, Claude names Anthropic, Gemini names Google. Useful when you have three tabs open and lose track of which is which.

There are other strong assistants — Perplexity for cited web answers, Grok, DeepSeek, Microsoft Copilot — and you’ll meet them in Appendix B — The AI Tool Directory. For now, the three above are your core kit. Don’t over-collect. Three you use beats ten you signed up for.

A.3 — Set Up Your Prompt Library (a plain notes doc)

The single habit that separates professionals who get value from AI from those who don’t: they save the prompts that worked. A good prompt is reusable capital. You write it once, refine it, and reach for it for the rest of your career. Week 12’s whole project is building this out properly — but you start the container now, empty, on day one.

You don’t need software. Pick whichever of these you already have:

• A Google Doc (free, syncs everywhere, easy to share with a teammate later)
• Apple Notes, Microsoft OneNote, or your phone’s built-in notes app
• A plain text file (prompt-library.txt) on your desktop

Create it, title it “My Prompt Library,” and give each entry the same simple shape:

=== Task: <what you're trying to do> ===
Tool I used: <ChatGPT / Claude / Gemini / other>
The prompt:
<paste the exact wording that worked>
What worked / what I'd change:
<one line of notes>

Add your first entry right now: the “explain a large language model” prompt from A.2, and a note about which tool’s answer you liked best. That’s the library started. Every week, when a prompt earns its keep, it goes here. By Week 12 you’ll organize it by role and task — the wise, Scripture says, lay up knowledge (Proverbs 10:14, ESV). This little doc is where you lay up yours.

A.4 — Which Lab Needs Which Tool

So you never over-prepare or panic, here is every week mapped to the minimum free tool that does the job. When a week points you to a specialized tool, Appendix B tells you which one and where to find its free tier.

Read that table and breathe. Fifteen of sixteen weeks run entirely on free browser accounts you already made in A.2. Only Week 11 even offers a download, and it’s optional. Nobody needs a $2,000 laptop for this course.

One more free tool worth naming: several labs fill in a small table or matrix (Weeks 3, 7, 12, 16). Any free spreadsheet works — Google Sheets (free with the Google account you already have), Excel or Numbers if you own them, or even a table in a plain document.

A.5 — Optional: Run AI Privately on Your Own Computer

This one section is different from everything above, and I want to be clear about that up front. A.5 is optional, it installs software, and it is the only part of this course that asks anything of your hardware. It exists for one lab — Week 11’s Frontier vs Local Showdown — where you compare a big cloud model against a smaller one running entirely on your own machine, with your data never leaving the room. That privacy is the point: some information is too sensitive to send to anyone’s cloud, and guarding the deposit entrusted to you (1 Timothy 6:20, ESV) is a professional duty, not a preference.

Before you try it, two honest cautions:

1. It needs your own computer, with permission to install apps. A locked-down work or school laptop usually won’t allow it — that’s fine, use the fallback below.
2. It needs a reasonably modern machine. A rough rule as of mid-2026: a small (7–8 billion parameter) model at the common 4-bit size needs about 4–5 GB of free memory to run, so 8 GB of RAM is a bare floor and 16 GB is comfortable. Apple Silicon Macs (M-series) do this especially well. Older or low-memory machines will be slow or won’t run it — and that’s okay.

The gentlest way in: LM Studio

For non-technical users I recommend LM Studio — it’s a normal desktop app with a chat window and a built-in model browser, no typing commands required:

1. Go to lmstudio.ai and download the app for your operating system; install it like any other program.
2. Inside the app, search its model catalog and download a small model (look for a 7–8B model in a “Q4” size — the app labels which ones fit your machine).
3. When it finishes, open a chat with that model and ask it anything. If it answers, an AI is now running entirely on your computer, with no internet — you can even unplug your Wi-Fi and it keeps working.

Prefer a command line, or want the tool most guides assume? Ollama (ollama.com) is the popular alternative — install it, then run ollama run llama3.1:8b in a terminal to pull and chat with a small model. Same idea, more technical feel. Either one satisfies the Week 11 lab.

If your machine can’t do it — the cloud-only fallback

You can complete Week 11 without installing anything. The project gives a cloud-only path: you compare two cloud models (say, a flagship free assistant versus a lighter/faster free one) on the same tasks, and you reason through the privacy tradeoff from the reading and the in-class demo rather than by hosting a model yourself. You lose none of the credit and none of the lesson — the lesson is the privacy-versus-capability tradeoff, and you can learn that with your hands or with your head. Choose the path your hardware allows and don’t fight it.

Coach’s Note — Don’t buy a computer for this course, and don’t feel behind if A.5 isn’t for you. Running a model locally is a genuinely cool thing to have done once — it demystifies the whole business — but it is a bonus rep, not the main lift. The professional skill this course builds is judgment, and judgment doesn’t require a graphics card.

A.6 — When Something Won’t Work

A few snags catch almost everyone. None of them are your fault, and all are quick fixes:

• “It won’t let me sign up.” Some accounts have a minimum age or a region restriction, and school/work Google accounts sometimes block Gemini. Use a personal email or a free Gmail instead, and try from a home network if your workplace filters the site.
• “I hit a limit / it told me to slow down.” That’s the free tier working as designed. Wait a bit, or switch to one of the other two free assistants for that lab — this is exactly why you set up three.
• “It won’t take my file.” Free tiers cap upload size and sometimes the file type. For Week 7, paste the text directly, use a smaller excerpt, or try Gemini or NotebookLM, which handle documents well.
• “The app looks different from these steps.” These tools redesign their screens constantly. The button names drift; the task doesn’t. Look for the obvious “new chat,” “sign in,” or “upload” control — it’s there under some label.
• (A.5) “The local model is painfully slow or won’t load.” Your machine is under-powered for that model — pick a smaller one, or just use the cloud-only fallback. No penalty.

A.7 — Submitting Your Work

Every project in this book is submitted the same way: one link to a shared folder, or one combined PDF. No GitHub, no code, nothing technical. Here is the whole procedure once, so every project can simply say “submit per Appendix A.”

Option 1 — a shared folder (recommended).

1. Make one folder per project in Google Drive, OneDrive, or Dropbox, named like the project asks (e.g., p1-first-contact).
2. Put every deliverable inside — documents, tables, screenshots — using the file names the project lists.
3. Turn on link sharing: in Google Drive, right-click the folder → Share → change “Restricted” to “Anyone with the link — Viewer.” (OneDrive: Share → “Anyone with the link can view.” Dropbox: Share → Create link.)
4. Test the link before you submit it: open it in a private/incognito browser window. If you can see the files while signed out, so can your instructor. If you get a permission wall, fix step 3 — a link the grader can’t open is the most common submission failure there is.
5. Submit the link wherever your course collects work (Canvas, email, or whatever your instructor names).

Option 2 — one combined PDF. If sharing links is blocked where you work or study, export everything into a single PDF (most word processors will Print → Save as PDF; screenshots can be pasted onto pages) and upload that file instead.

Three rules, every week:

• Names matter. Use the file names from the project’s Submission section — the grader is checking against that list.
• Nothing sensitive. No real client names, patient details, student records, or credentials anywhere in a submission — invented details only (Appendix C is the rule book).
• The link must open. Test it signed out, every time.

A.8 — Verify Your Toolkit, Then Count the Cost

Before Week 1, confirm the pieces work. This takes five minutes:

• Three accounts live: ChatGPT, Claude, and Gemini each answered your A.2 verify prompt.
• (Optional) Mobile: at least one app installed and signed in on your phone.
• Prompt library started: your “My Prompt Library” doc exists and has its first entry.
• You know where the map is: you can find the A.4 table to check what your next lab needs.
• (Optional, Week 11 only) Private path: LM Studio or Ollama answers a question offline — or you’ve noted that you’ll use the cloud-only fallback.

If those are true, your toolkit is ready and you owe no one a dollar.

Tear down what costs — count the cost before you commit

Here is the discipline that keeps a free course free. Almost everything above stays free forever. But a few things in this space can quietly start charging:

• Watch the “upgrade” nudges. Every free tool will offer you a ~$20/month “Plus / Pro” plan. You do not need one for this course. If you ever choose to try a paid tier, set a calendar reminder and cancel before it renews.
• Free trials of specialized tools (some image, video, or voice tools in Weeks 9–10) may ask for a card for a trial. Prefer the ones with a true no-card free tier; if you start a trial, write down the cancel-by date the day you start it.
• The local models in A.5 cost nothing to run — no per-use charge — but they use your electricity and disk space. Delete models you’re done with from inside LM Studio or with ollama rm.

That habit — sit down first, add up what a thing will cost, and only then commit — is not just budgeting. It’s wisdom the Lord Himself put in the mouth of the Coach:

“For which of you, desiring to build a tower, does not first sit down and count the cost, whether he has enough to complete it?” — Luke 14:28 (ESV)

You are building something this term — an AI-augmented way of working that will outlast the class. Count the cost of the tools before you lean on them. Choose the free path where the free path is enough. Keep the money, the data, and the final decision in your own hands. You choose the tool. You own the verdict.

Now go make your three accounts. I’ll see you in Week 1.

Up next: Head into Chapter 1 — The AI Revolution — with your toolkit live. Keep Appendix B — The AI Tool Directory open when a lab names a specialized tool, read Appendix C — Responsible & Ethical AI Use before you paste anything real, and use Appendix D — Glossary whenever a word is new. See you in the gym.