Week 1 — AI Build Studio · "Your First Great Prompt"

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective: Objective 1 — adopt the working mindset (general→specific, iterate) and learn that fluency ≠ truth · SLO A (produce a quality result with AI) · SLO B (verify it)
Worth 50 points · AI Build Studios group = 15% of the grade · Studio 1
Format: a hands-on build — you'll take a real task from your own life, turn a weak prompt into a strong one through iteration, get a genuinely useful result, and then catch the AI's mistakes in its output.

This is the course's signature weekly component. Every instructional week has one Studio — a real thing to build, a required step where you verify and improve the AI's work, and a short reflection. All tools are free; everything is links to external sites.

Part 1 — The Build Goal

By the end of this Studio you'll have produced three things for a real task from your own life or major:
1. A strong, refined prompt (built up from a weak first draft).
2. The AI's useful final output.
3. A short AI-critique write-up naming at least one thing the AI got wrong, made up, or overdid — and how you caught and fixed it.

This is the whole course in miniature: use AI well (iterate to a great prompt) AND judge it well (verify the output).

Open one approved assistant to build in: ChatGPT (https://chatgpt.com), Claude (https://claude.com), Gemini (https://gemini.google.com), or Copilot (https://copilot.microsoft.com). A free account is enough.

Part 2 — Pick a Real Task

Choose something you actually need (this makes the result useful, not busywork). Examples:
- A study plan for a real upcoming exam.
- A first draft of an email you need to send (to a professor, a club, a manager).
- An outline for a paper or project you have due.
- A plan for a real event, trip, or budget.

Write your task here: I want AI to help me ______.

Part 3 — Build the Prompt in Three Versions (iterate!)

Version 1 — the weak draft. Write the prompt the way most people would, in one short line (e.g., "help me study for my exam"). Send it. Notice how generic the answer is. Save this v1 and the answer.

Version 2 — add context, goal, and a constraint. Rewrite it with: who you are (role/major), the specific goal, and one constraint or format (a length, a table, a number of options). Send it. Compare.

Example: "I'm a first-year psychology major. I have a stats exam in 6 days covering the mean, median, mode, and standard deviation. Give me a 6-day study plan, one focus per day, in a table."

Version 3 — iterate on the result. Read v2's output and steer it once more with a real follow-up: "make day 1 lighter," "add one practice problem per day," "explain standard deviation simply first." Save v3 and the final output.

You should now clearly see the quality climb from v1 → v3. That climb came from you iterating, not the AI "trying harder."

Part 4 — The Verification / AI-Critique Step (required — this is the BYOAI step)

Now be the judge of the AI's work. Somewhere in its output, the AI very likely included a specific claim you should not take on faith — a "fact," a statistic, a recommended resource, a book or citation, a confident detail. Find it and check it.

1. Pick one specific, checkable claim the AI made (a fact, a number, a named source, a "studies show…", a recommended tool/book).
2. Ask the AI to back it up: "What's your source for that? Are you certain it's accurate? If you're not sure, say so." Notice whether it doubles down, walks it back, or admits uncertainty.
3. Verify it yourself — a quick search, the official source, or a second assistant. Confirm whether the claim is true, false, or made up.
4. Also watch for sycophancy: if you say "I think X," does the AI just agree with you? Try pushing back on its plan ("isn't day 3 too heavy?") and see if it caves without reason.

Write 3–4 sentences reporting: the claim you checked, what you found, and at least one thing the AI got wrong, fabricated, or overdid (or, if everything checked out, exactly how you verified each claim — that's the skill). Then say how you'd fix or guard against it.

The habit all term: the tool drafts, you judge. A chatbot will confidently invent a statistic, a citation, or a "fact" — catching it is the point.

Part 5 — Reflection (2–3 sentences)

What surprised you most — about how much the prompt mattered, or about what the AI got wrong? What will you do differently the next time you use AI for something real?

Part 6 — What to Submit

Submit a single document (or text entry) with: your task, your three prompt versions (v1, v2, v3) with a one-line note on how each got better, the final output (or the relevant part), your Part 4 AI-critique write-up (the claim you checked + what you found + the fix), and your Part 5 reflection. Due Sunday, Sep 6, 11:59 p.m. (50 points).

Instructor answer key & model deliverable — REMOVE BEFORE PUBLISHING TO STUDENTS

Students use their own tasks, so deliverables vary. Grade the process (iteration + verification + reflection), not a specific answer. The model below shows what full credit looks like.

Model deliverable (illustrative):
- Task: "Help me plan studying for my intro statistics exam."
- v1: "help me study for stats" → generic list of tips.
- v2: "I'm a first-year psych major with a stats exam in 6 days on mean/median/mode/SD. Give me a 6-day plan, one topic/day, in a table." → focused, dated table.
- v3: "Add one practice problem per day and make day 1 a light review." → genuinely usable plan.
- Part 4 (model critique): "The AI recommended a specific YouTube video 'by Khan Academy titled Standard Deviation in 5 Minutes' with a link. I asked for its source; it admitted it wasn't certain the exact title/link were real. I searched and the specific title didn't exist as described — a fabricated citation. Fix: ask for search terms instead of exact links, and verify any named resource before trusting it." (Any specific, verified catch earns full marks — common ones: an invented statistic like "studies show students retain 80% more when…", a made-up book/citation, a fake URL, or sycophantic agreement when pushed.)
- Reflection: notes that the v1→v3 jump came from adding context and iterating, and that fluent output still needed checking.

Why the verification step can't be faked: a student who pastes raw AI output with no checked claim, no specific catch, and no fix earns the low end of the AI-critique row — the rubric rewards judgment, not the AI's prose.

Grading rubric — 50 points

Quality gate (self-checked): the tools and links named (ChatGPT/Claude/Gemini/Copilot homepages) are real and current (verified live); no fabricated product features; the activity requires the student to catch the AI, not trust it (verification-as-content). No student-produced output is asserted as "the" answer — the key grades the process. Product-accuracy gate: PASS.