Week 8 · Practice, Present, Reflect

This is our last regular class. Eight weeks ago, "write an Abstract" was a wall. Now your whole paper sits on Overleaf and is pushed to GitHub — done. There's no new writing today, and honestly only two things matter: first, you'll pair up and present your project to each other — out loud, face to face; then we'll look back together at how far you came in eight weeks. Everything else here — the questions people ask, what to do when you're stumped, the sprint plan — is prep for the real defense on 7/24-26.

One thing to settle your nerves first. A defense is not a test of whether you know everything about your topic. It asks you one thing: why did you do it this way? Why this question, why this method, why this conclusion. You don't have to be the world's expert on your subject. You just have to be the person who knows your own project best — and that's already you.

Project names on this page are just tags (e.g. #short-video-&-teen-attention) — no one is named. Keep this page open. Stuck? Mia is here.

01 · The questions a panel keeps asking

Here's the good news. Whether it's your partner today or the panel on 7/24, people ask the same handful of questions, over and over. Get these ready and almost nothing can catch you off guard. So right now, write a short answer to each one — a couple of lines you can say out loud. That little stack of cards is your whole prep, and you'll reach for it the second you present in the next section.

Write all of them now, in your own words, on cards you can hold. You'll use them in the next section — and again the night before the real thing.

02 · Present it to a partner

This is the heart of today. Reading about presenting does nothing. Standing up and saying your project out loud, to one real person — that's the only thing that makes it click. So we do it for real, in pairs. The point isn't to trip each other up — it's to get the words out of your mouth and land them on one listener. Here's the setup:

Each pass runs about 5 minutes per person, in two beats:

One pass — about 5 minutes per person: 1 · PRESENT (3-4 min) stand up, out loud, to your partner: what you looked at -> how you did it -> what you found -> what it means | 2 · TWO LINES + Qs (1 min) they tell you ONE "this part was clearest" and ONE "this part I didn't get", then ask 1-2 questions then swap -> now they present, you listen two passes total, so both of you present once
Three things to actually do while you present — don't just nod at them. (1) Don't read a script. The second you start reading, your partner can hear it — and it stops sounding like you. The words live in your mouth, not on a page. (2) Say the jargon in plain words. Not "statistically significant" — "this gap almost certainly isn't just luck." If a term makes your partner's eyes glaze, that's the term to fix. (3) You're not being tested — you're being understood. The whole win is one person walking away able to say back what you studied. That's the bar.

03 · When they stump you

You will get a question you can't answer. That's normal — it happens to everyone, and it's not the disaster it feels like. The only real mistakes are panicking and making something up. Three moves. You'll act each one out.

Stumped? Three moves — no panic, no bluffing: 1 · PAUSE wait a few seconds before you speak. Feels like forever to you; to them, you're just thinking. 2 · ASK "You're asking whether XX, right?" — asking is fine, and it buys you a few more seconds. 3 · THINK turn "I don't know" into your thinking, out loud: "I'm not sure, but my guess is ... because ..., and to check it I'd ..."

Move 3 is the one that turns a dead end into a win. You're showing the panel how you think — and that's exactly what they're grading:

A stumped moment — ✗ vs ✓:
✗ (freeze, then bluff) "Uh… well, considering all the overlapping factors and the variance across the data…" — a wall of words over a blank.

✓ (pause, then think) "I'm not sure about that one. My guess is [X], because [reason]; to be sure I'd have to [collect / check / compare Y]." — honest, and you can see the brain working.
Saying what's wrong with your own study is a plus, not a minus. When the panel asks "what's the weakness here?", beat them to it — "I only asked 30 students at one school, so this can't speak for all teenagers." That's what a clear-eyed researcher sounds like. Pretending your study is perfect is what actually gets you chased around the room. And never bury a simple question under a pile of jargon — a panel sees through it in one second.

04 · Get the talk out of your mouth

Your slides are done — you built those yesterday. Now the part that actually carries the room: the words coming out of your mouth. The talk lives in you, not on the screen. Here's how to get it ready.

Turn each term into plain words — a few examples:
"statistically significant" → "this difference almost certainly isn't just chance"
"gene expression" → "a gene is like a light switch — it can be turned on or off"
"sample of n = 30" → "I asked 30 people"
"correlation" → "the two tend to move together — which doesn't mean one causes the other"

One more, and it's optional but worth it. Want it to flow even smoother tomorrow? Give one short TED talk 8 minutes tonight — Chris Anderson, TED's secret to great public speaking (~8 min). He runs TED and has watched thousands of these. Four moves from it you can use tomorrow:

Steal these four — straight into tomorrow's talk: 1 · WHY CARE open with a reason to care, so they lean in 2 · ONE STEP build the idea in their heads, one step at a time 3 · FAMILIAR explain the new with something they already know 4 · SPEAK IT don't read a script — say it in your own words

05 · The sprint to 7/21

Your map from this room to the stage. Fill in your own gaps as you read — for "what's still missing", copy straight from the GitHub issues your teacher opened on 7/11. One landmark to circle: 7/15 OH is the last in-person rescue. Bring the problems you can't crack alone.

How to use it: copy your gaps straight from the 7/11 GitHub issues; rehearse at least twice, no exceptions, with one run for someone outside your field; and treat 7/15 OH as your last in-person rescue — bring any hole you haven't closed that day.

06 · Carry these to the stage

Two things travel with you from the writing desk all the way onto the defense stage. First, the four red lines — the same ones that ran through every writing week. Break one, and no slide, no rehearsal, and no smooth answer can save the project:

① Never invent or change data — not one digit. No made-up response, no edited value, no result you didn't get.
② References must be real — ones you checked yourself. No AI-invented papers, authors, or DOIs.
③ A small sample means say it's small. Never stretch it to "all teenagers", and never turn a correlation into a cause.
④ Anywhere you used AI, say so, APA-style. Which step, what for — the content and the data are yours.

Second, the nerves — because they're normal. They hit hardest right at the opening and settle once you're a sentence or two in. And here's the part people don't believe: the audience basically can't tell you're shaking. They're watching what you say, not your hands. The one cure that always works is rehearsing until it's in your body — at least twice, one run to someone who doesn't get your topic. Practice enough and even a stumble, you catch it and keep going.

07 · Look how far you came

Before you leave, look back — and look at the real thing you picked up. It wasn't any one tool. Week by week, you learned to drive AI to turn your idea into something real, one step at a time — and Cursor / codex was your workbench the whole way. Eight weeks ago the terminal scared you and "write an Abstract" was a wall. Look what you can make AI do now:

Eight weeks — each week, one more thing you learned to make AI do: Week 1 picked the one social question you'd study — your starting line, no tools yet Week 2 opened Cursor / codex for the first time, had it get your dev environment running Week 3 had AI turn one sentence into your first webpage that actually ran Week 4 had AI put the site online in three commands Week 5 had AI plug into another AI (DeepSeek / MCP) — AI directing AI Week 6 had AI iterate like an engineer (issue -> PR), make images, read Supabase to analyze your data Week 7 had AI help you write the research up into a paper Week 8 had AI lay it out as a PDF (Overleaf) and build the slides (Guizang) the tool kept leveling up. so did you — you learned to run it.

Here's the thing to actually keep. From Week 2 — the first time you opened Cursor / codex — to now, running it through a whole research project start to finish, what you learned isn't any single tool. It's this: use AI as your workbench, and you bring the ideas and the calls. You point; AI does the work. Change the topic, change the field — you'll still have it.

So take a minute — a real one. Turn to the person next to you and answer these out loud, or write one down. This is the reflection that closes the course:

That's the course. From here it's 7/15 OH (the last in-person rescue), 7/21 to submit paper + slides + poster, and 7/24–26 the defense. Whatever number your data landed on, you did real research, honestly — be a little proud of that. The hard part is already behind you. See you on stage.

08 · Before you go — class feedback

Last one, and it's for the whole course. Two minutes — tell us what landed across all eight weeks, and what you still want help with before 7/21. It goes straight to the teacher and shapes OH and the run-up to your defense.