Project P1 — First Contact

“what is man that you are mindful of him, and the son of man that you care for him?” — Psalm 8:4 (ESV)

Chapter: 1 — The AI Revolution Due: End of Week 1 Submit: A link to a shared folder (Google Drive / OneDrive / Dropbox) or a single combined PDF containing your journal, your worksheet, your tool-classification table, and your screenshots. Non-technical — no GitHub, no code. See Appendix A for account setup and submission mechanics. Allowed tools: The three major consumer assistants on their free tiers — ChatGPT, Claude, and Gemini — reachable from any browser or phone. That’s it. (Setup help: Appendix A. Tool reference: Appendix B.) AI policy (Phase 1): This whole project is using AI — that is the point. But you are using it to compare and evaluate, not to obey. You will run the same task on three tools and judge the results with your own eyes. Every deliverable ends with an honest AI usage note. And you will follow one hard rule from day one: never paste anything confidential, personal, or client-identifying into a consumer AI tool — read Appendix C first, and use the safe fictional scenario below.

The Setup

Meet Renee Alcott. She runs the front office of Cedar & Main Landscaping, a twelve-person company in a mid-sized town. Renee is not “a tech person.” She schedules crews, chases invoices, answers the phone, keeps the owner sane, and writes the emails nobody else wants to write. For a year now she has heard the same sentence at every chamber-of-commerce lunch — “you should just use AI” — and she has never once opened one of these tools. She isn’t afraid of it. She’s busy, and nobody has ever shown her why or which one.

This week, Renee has a genuinely awkward email to write. A good, long-time client — Belmont Property Group — is now sixty days late on a $4,200 invoice. Renee needs a message that is firm enough to get paid and warm enough to keep the relationship. She has rewritten the first sentence four times and given up twice. It is exactly the kind of small, human, judgment-loaded task where AI might genuinely help — or might hand her something so generic and cold that it damages a relationship she has spent five years building.

You are going to be Renee’s first guide. You will open all three major assistants, hand each of them the same task, and find out — with your own eyes — how they differ, which one you’d actually trust with this, and what you’d have to fix before Renee could ever hit send. Along the way you’ll take stock of the AI already running quietly in a workday like hers. This is first contact: forming your own honest first impressions, and starting the habits — compare, verify, decide — that the whole course is built on.

Setup (the starter)

This chapter ships two starter files in code/. You will use both:

• code/ai-in-my-work.txt — a worksheet to inventory the AI already in a real workday, name each by type, and find where the judgment lives. You’ll fill this in for your own week (or Renee’s).
• code/ai-types-cheatsheet.txt — a one-page reference to the six types of AI. Keep it open while you classify tools.

Copy both into your project folder and make them yours. Do not submit them blank.

Learning Targets

You will demonstrate that you can:

• Get started — create working free-tier accounts on the three major assistants and reach a chat window on each.
• Compare like a professional — run one identical, realistic task across three tools and describe the differences in tone, accuracy, and usefulness concretely, not “they were all pretty good.”
• Verify, not just receive — read each AI output with a cold eye, catch what’s wrong or too generic, and state exactly what you’d change before it went out under your name.
• Classify by type — sort real AI tools you already use into the six types and explain why, using the cheat sheet.
• Own the verdict — form and defend a judgment the AI cannot make for you: which assistant you would adopt, and why.

Normal Tier

Goal: Make first contact with all three assistants, run one real task on each, journal the honest differences, and take stock of the AI already in your work.

Required features

Produce a single document (or folder) named p1-first-contact containing:

1. Three accounts, three screenshots. Create a free account on ChatGPT, Claude, and Gemini (see Appendix A). Include one screenshot per tool showing you reached a working chat window. (No personal data in the shots — a blank or generic chat is fine.)
2. One task, three runs — outputs.txt (or a section in your doc). Give all three assistants the same prompt for Renee’s overdue-invoice email (or your own equivalent awkward-but-safe work email — invented details only, no real client data). Paste each assistant’s reply unedited, each under a header naming the tool and, if the app shows it, the model (e.g., “ChatGPT — GPT-5.x, mid-2026”; “Claude — claude-sonnet-5”; “Gemini — Gemini 3.x”).
3. journal.docx / .pdf — the comparison journal (the heart of the project). For each of the three outputs, write a short honest paragraph: what it did well, what it got wrong or too generic, and what its “personality” felt like. Then a short table scoring each on tone, accuracy of the facts you gave it, and usefulness as a starting draft (1–5 each), with one sentence of justification per score.
4. The verification pass. For each output, name at least one specific thing you would change before sending it — a wrong assumption it invented, a number it got wrong, a phrase too cold or too soft, a promise Renee can’t keep. This is the “own the verdict” step; it is not optional.
5. ai-in-my-work.txt — completed. Fill in Parts A, B, and C for your own workday (or Renee’s, if you prefer to stay in character).
6. tool-types.xlsx / a table — five tools classified. List five AI tools or features you already use, and classify each into one of the six types with a one-line justification, using code/ai-types-cheatsheet.txt.
7. The reflection + AI usage note. Half a page: which assistant felt best for this task and why, one thing that surprised you, and one honest AI usage line (what you asked, what any tool got wrong, what you verified yourself).

Normal-tier rubric (out of 100)

Medium Tier (+up to 25% extra credit)

Push past first contact. Pick one or more:

1. A second, different kind of task. Run one non-writing task on all three — e.g., “summarize this (invented, non-confidential) three-paragraph vendor policy into three bullets,” or “brainstorm ten low-cost ways a small landscaping company could thank loyal clients.” Note how the ranking of the three tools changes by task type. (This previews the whole point of choosing the right tool.)
2. Try a second modality. On any one assistant that supports it, try a multimodal interaction — upload a photo (something safe, like a plant or a handwritten note) and ask about it, or use voice mode. Describe what worked and what it got wrong.
3. Free vs. the wall. Push a free tier until you hit a limit (a usage cap, a slower model, a “try Pro” prompt). Document exactly what the free tier stopped you from doing — this is real intelligence for deciding whether a paid plan is ever worth it later.

Document each clearly and keep every raw output. Extra credit rewards honest observation, not flattery.

Hard Tier (+up to 25% additional extra credit)

This is the judgment the machine cannot make for you, and it is graded as such.

Write recommendation.docx — a one-page memo addressed to Renee (or to your own manager) that makes and defends a single recommendation: which one of the three assistants should she adopt as her daily default, starting Monday, and why? You must:

• Name the winner and commit. “They’re all good, use whichever” fails. Pick one. Professionals make the call.
• Defend it from your own evidence — cite specific moments from your journal (this tool’s draft needed the least fixing; that one invented a fact; the other one nailed the tone). Not vibes — evidence you gathered.
• Name what she gives up. Every choice has a cost. What is your pick worse at, and when should Renee reach for a different tool anyway?
• Tie it to the spine rule. State plainly what Renee must still do herself every time, no matter how good the tool is — you choose the tool; you own the verdict.

A memo that says “use all three” or “AI can’t be trusted, don’t bother” fails. The grade is in the discrimination: showing you can weigh real evidence, commit to a recommendation, and defend the line — the one thing no assistant can do for Renee.

Submission

Put everything in one shared folder (Google Drive / OneDrive / Dropbox) or export it as one combined PDF, and submit the link per Appendix A. A tidy folder looks like:

p1-first-contact/
  journal.docx            (or .pdf)     — the comparison journal + verification pass
  outputs.txt                            — the three raw, unedited AI replies
  ai-in-my-work.txt                      — your completed worksheet
  tool-types.xlsx         (or a table)   — five tools classified by type
  screenshots/                           — one working-chat screenshot per assistant
  recommendation.docx     (Hard tier)    — the adoption memo

Do not submit screenshots or documents containing real client names, real invoice numbers, personal data, or anything confidential. Use the fictional Cedar & Main / Belmont scenario or invented details of your own. When in doubt, read Appendix C again — this is the habit that protects you for the next fifteen weeks.

Hints (Read Before You Begin)

• Give all three the exact same prompt. If you tweak the wording between tools, you are no longer comparing the tools — you are comparing your prompts. Paste an identical block into each.
• Write your own version first, or at least know what “good” looks like. You cannot judge three drafts of Renee’s email if you have no opinion about what the email should say. Decide the tone you’re after before you read a single AI reply.
• Capture the raw output before you fix it. The unedited reply is your evidence. Editing it destroys the comparison.
• “It sounds nice” is not “it’s correct.” Check every fact each tool used — the amount, the days overdue, the client name. A fluent email with a wrong number is worse than no email.
• Keep it fictional and safe. No real overdue clients. The scenario exists precisely so you never have to paste something you’d regret.
• Don’t rush the classification. Some of your five tools will honestly be two types at once. Say so, and pick the primary — that nuance is the point.

What Mastery Looks Like (Beyond the Rubric)

A mastered submission is one where I can read your journal and hear you thinking. You didn’t write “all three were helpful” — you noticed that one invented a payment-plan offer Renee never authorized, that another was so cautious it buried the actual ask, that the third nailed the warmth but got the amount wrong. Your five-tool table shows you can look at the invisible AI in an ordinary workday and name what it’s doing. And your Hard-tier memo commits — it names a winner, defends it from your own evidence, and states clearly what Renee must still do herself. The accounts and screenshots are table stakes. The judgment is the project.

Coach’s Note — The temptation this week is to be impressed. These tools are genuinely impressive on first contact, and you should enjoy that. But your job here is not to be a fan — it is to be the first person in Renee’s life who used them with a cold eye and told her the truth: where they help, where they’d embarrass her, and which one to trust with what. Impressed and skeptical is the exact posture we’re building. Start it now.

When You’re Done (a short checklist)

• Three free accounts created; one working-chat screenshot each.
• The same task run on all three; raw outputs saved unedited, tool/model named.
• Comparison journal describes concrete differences, not “all good.”
• You named at least one specific fix per output before “send.”
• Five tools classified by type, each justified in a line.
• ai-in-my-work.txt completed (Parts A, B, C).
• Reflection includes an honest AI usage note.
• (Hard) recommendation.docx commits to one assistant and defends it.
• No real client data, PII, or confidential material anywhere in the submission.

A theological footnote. You spent this week handing the same small, human task to three made intelligences and judging what they gave back. Notice what you were doing the whole time: deciding. The machines drafted; you weighed, corrected, and chose. That is not a limitation you’ll grow out of as the models improve — it is the shape of being a person. Psalm 8 marvels that God is mindful of small, ordinary man and crowns him with dominion over the works of God’s hands. You just exercised a sliver of that dominion — building tools, then ruling over them rather than bowing to them. Renee’s client, her relationship, her name on the email: those were entrusted to her, a person God is mindful of, not to a model that cannot be mindful of anything. To use the draft is wisdom. To sign it without reading it is to hand your God-given judgment to the work of your own hands. Keep the verdict. That is what it means to be what man is — mindful, because you are Someone’s beloved, and answerable, because you are His.

See you next week.