Capstone Showcase

Teacher note

Accounts to pre-stage

• Nothing new. Every account this arc needs was created earlier, and the GitHub account from Module 7 is the one that matters, because the project ships to a URL.
• Check before kickoff that everyone's Module 7 URL is still live. A few will have gone stale, and fixing that is a five-minute job at kickoff and a panic at showcase.
• If the mid-point goes async, the shared board needs commenting open to the whole class. Test it with a second account.

Fifteen minutes before

• Re-read the paragraph each student submitted from Module 7 so the five-minute conversations are fast and specific.
• Have the scope-check questions ready: who has this problem, what will exist at the end, what would you cut first.
• Open the link wall and put its address on the board.
• Set a visible timer for the showcase talks and test it. Fixed time is the equaliser and it only works if it is enforced.
• Decide the showcase format (lightning talks or stations) and tell the room which one, well before the day.

Wifi fallback

Kickoff survives on paper: briefs get written and approved with no device at all. The gallery walk works with printouts pinned to a wall. The showcase is the one that needs the internet, and the async gallery option is the insurance policy: projects go on the shared board and comments happen between sessions.

Depends on earlier modules

• Module 7 gave every student a live URL and a GitHub account. The capstone ships to that URL.
• Module 6 is where the ethics notes in the brief come from, and the case-study homework is a good source of project ideas.
• Modules 2 through 5 supply the verification and attribution habits that the rubric grades directly.
• A student who kept the personal agent going since Module 2 already has a capstone. Point that out at kickoff, several of them have not noticed.

Teacher note · the rubric, and why it is shaped this way

Grade the process, not the polish. A simple project with strong evidence of iteration outscores an impressive prototype with no evidence of learning, and saying that out loud at kickoff changes what students build.

Suggested weights

• Artifact, 30. The thing itself, working.
• Iteration and validation evidence, 30. What they tried, what failed, what they changed because of it.
• Verification and attribution, 20. Which claims were checked, which tools were used, said plainly.
• Presentation, 20. Clear, honest, within the time.

Plus the bolt-on rule: points come off for any AI output that was never validated against reality, whether that reality is an audience, a dataset, or a manual check. And the honest-limitation slot in the brief is graded up, not down. "Testers could not finish the task, so we redesigned it" is the best sentence a student can write.

Teacher note · when the arc compresses

Up to two of the three sessions can get reassigned to other things, so plan for that rather than being ambushed by it. The minimum viable capstone is kickoff plus showcase in a single combined evening: brief approval and work time early, lightning talks late, with the gallery walk going async on a shared board between sessions. It works. It is tighter, and the briefs have to be scoped smaller from the start.

Anticipated wrong turns

• The brief is a topic, not a problem. "Something about AI in healthcare" is not a brief. Ask who has the problem until a person appears in the answer.
• An engineer scopes a six-month project. The most common failure at this end of the room. Ask what they could finish by the showcase and cut everything else, then say plainly that ambition is not what the rubric rewards.
• A novice apologises for their idea. The second most common. A working household budget with verified totals is a strong capstone and it hits the rubric harder than a half-built app.
• Two students have the same idea. Fine, and often good. Let them work near each other and compare approaches at the mid-point.
• Somebody has no idea at all. Go back to their Module 7 exit ticket, or to the case study they wrote in Module 6. The idea is usually already written down in their own handwriting.
• The approval queue stalls the room. Rolling conversations while everyone else works, and post the order on the board so nobody sits waiting for their turn.

Discussion, with the answers you are steering toward

"Who has this problem?"

A named person or a specific group. "People" is not an answer.

"What will exist at the end?"

Something clickable. A page, a tool, a document, a game. Not a plan to build something.

"What would you cut first if you ran out of time?"

If they cannot answer, the scope is not understood yet. This one question catches most of the over-scoped projects.

Anticipated wrong turns

• Everyone writes "looks good". Require the question sticky. A question is harder to fake than praise.
• Somebody has nothing to show. Have them show the brief and talk through where they are stuck. Getting three stickies on a stuck point is more valuable than a demo anyway.
• Feedback gets treated as instructions. Say clearly that stickies are data, not orders. Deciding what to ignore, and saying why, is part of owning the project.
• The confident projects get all the visitors. Rotate on a timer and assign the first stop of each group.
• The gap before the showcase is too short. A week or more, minimum. Capstones need time away from the room, and this is the single easiest scheduling improvement available.

Discussion, with the answers you are steering toward

"What is the most useful sticky you got?"

Usually a question that exposed an assumption. Read two or three out loud to the room.

"What are you going to ignore?"

Encourage this. A student who can defend ignoring feedback understands their own project.

Anticipated wrong turns

• The demo does not work live. Have everyone bring a screenshot or a short recording as a backup, said as a rule at the mid-point, not on the night.
• Talks run over. Visible timer, hard stop, said in advance and then actually enforced on the first speaker. If it slips once it slips all night.
• The polished project wins the room's attention while the iterated one gets none. Your framing fixes this: ask the polished one what failed along the way, in public, and reward the answer.
• Somebody skips the limitations slide. Ask for it out loud. It is graded up, and letting it be skipped teaches the wrong thing to the whole room.
• A student freezes. The async option existed for exactly this and should have been offered weeks ago. Take the questions yourself and move on gently.

Discussion, with the answers you are steering toward

"What did the AI do badly?"

The best answers are specific and unflattering. Praise them out loud, it changes what the next speaker says.

"What would you do next?"

Every project should have an answer. Finished is not the same as complete.

"Who is going to keep this running?"

They are, and they can, because they own the account and the files. That is the last word of the course.

Spotlight sweep across the arc

The capstone has three share moments instead of one, and each has a different job. Kickoff names every project aloud. The gallery walk makes unfinished work public and safe. The showcase is the real thing.

Prompts to hand the student showing

• Kickoff: "What are you making, and who is it for?"
• Mid-point: "Where are you stuck, and what would unstick you?"
• Showcase: "What changed, honestly measured?"
• Showcase: "What would you still fix?"

Prompts for the room

• "Who else is solving something similar?"
• "Who wants to try this one at the stations?"
• "Whose project would you use next week?"

Wander block menumost of the kickoff session

The kickoff is a wander block from start to finish, which is why it works. Both instructors circulate, the approval conversations happen rolling, and the menu below is for people who are between things rather than a schedule.

• Build. The default and the best use of the time.
• Fix your Module 7 URL if it went stale. Do this at kickoff, not at showcase.
• Find a tester. The person two seats over counts, and their confusion is real data.
• Raid the earlier modules: the flyer sprint, the budget builder, the custom assistant and the workflow map are all half-built capstones already.
• Look at the games shelf's build paths if the teaching-game tier appeals. The templates are license-checked and one of them has no build step at all.
• Talk to an instructor about scope, again, as many times as needed. Scope is the thing that sinks capstones.

Synthesis, the last ten minutes of the course

Put the link wall up and leave it up. Then go back to the sticky-note board from Module 1, the hopes, the worries, and the things people had heard. Read a few. Some got answered, some got confirmed, and a few are still open, which is honest and worth saying. That board and that wall next to each other is the course in one image.

Exit ticket

Last one of the course, so make it forward-looking: what is the next thing you are going to make, and what would stop you? The second half of that question is the useful half, and the answers are the best input you will get for teaching this again.

Homework handoff

Nothing goes home after the showcase, which is the point. What goes home is the URL, the account, and the files, all of which they keep. If a project is worth continuing, say so directly to that student before they leave the room.

One last thing worth doing: ask permission to keep the best projects on the link wall for the next group. Seeing what the last room built is the most persuasive thing a new student can be shown.