Sunday, May 17, 2026 · 3:30–5 PM · VISTA Hamilton Hub
Plate 1
A faculty talk in the VISTA Hamilton Hub: the internet, from the mid-2000s onward.
In May 2026 the club ran AI Literacy Week for the whole school. Two challenges went out by email, one a day. The third was a live game on Sunday afternoon, when we met in the VISTA Hamilton Hub to share what came back.
Two by email · one live
Challenge 1 · By email
Can AI draw a real Pomfret faculty member from words alone? Part 1 used a text-only prompt. Part 2 added a reference photo and ran it again, to compare.
Challenge 2 · By email
Three short reasoning questions, run against the AI model of your choice.
Challenge 3 · Live on Sunday
The closing game at the main event. Three categories, three rounds.
How the first two worked
Each took about 5–10 minutes. Paste the club’s prompt into the AI model of your choice, then submit your answer on a Google Form.
Sunday, May 17 · 3:30–5 PM
(01)
Cayden Auyang ’27
The results from Days 1 and 2, then the three you’ve probably used: ChatGPT (OpenAI), Claude (Anthropic) and Gemini (Google). Same family, different strengths.
(02)
Luke Ryan ’27
NotebookLM, Cursor and Claude Design, then a live demo that built a ten-slide deck from one prompt.
Plate 2 · The prompt
“Make a slide deck for an Oyster Farm business on Block Island that focuses on sustainability. …”
The live Claude Design demo started from one typed prompt.
Plate 3 · The deck it built
Ten slides from that one prompt.
AI-generated demo content: the oyster farm and every detail in the deck are invented placeholders.
(03)
Jay Youm ’26, alumni advisor
Where AI is heading, from Mars rovers to cancer treatment.
(04)
Challenge 3, played live
The closing game, and the week’s third challenge. Three categories, three rounds.
Countries with four letters · Body parts with three letters · Periodic-table elements of five letters or fewer
Round 1
Pure recall
No phones, no AI.
Round 2
AI as coach
The AI may only give hints, never answers.
Round 3
Learning check
Phones away. Fresh recall.
Did the AI actually teach you anything?
(05)
Awards for the most involved participants across the week. Pizza for everyone.
Also
Faculty
On the internet from the mid-2000s onward: creators versus users, then social media. Pictured at the top of this page.
Across nearly every submission, adding a reference photo made the image noticeably more accurate.
One submission rated it less accurate.
One model refused to generate a person.
“What you give the AI matters as much as what model you pick.”
“The car wash is 50 meters away. Should I walk or drive there?”
Every model, across three companies, advised walking.
A car wash is for the car.
“AI doesn’t actually understand context the way you do. It pattern-matches on the words you give it.”
“Sally has 3 brothers. Each of her brothers has 2 sisters. How many sisters does Sally have?”
Every model got it right.
Well-defined logic is where AI is sharp.
Three labeled boxes, and exactly one label is true.
The models disagreed, even with themselves.
One Claude run flagged a likely typo in the puzzle itself.
From the event deck
(1)
Context beats cleverness.
From Exhibit 1
What you give the AI matters as much as what model you pick. Add references. Add examples. Be specific.
(2)
Common sense is still on you.
From Exhibit 2
AI pattern-matches on words. It doesn’t understand purpose. The car wash test proved that across every model.
(3)
Same model, different answers.
From Exhibit 4
Three Claude runs of the same puzzle gave three different answers. Don’t take the first response as truth.
AI Youth Alliance provided some of the materials, and the club wrote its own challenges and shared its findings back with them. They will also help the club reach more local organizations.
Run by
The Civic AI Club. Open to all Pomfret students and faculty.
Speakers
Cayden Auyang ’27, Luke Ryan ’27, Jay Youm ’26 (alumni advisor) and a member of the faculty.
When and where
Sunday, May 17, 2026, 3:30–5 PM, in the VISTA Hamilton Hub.
Next
AI Literacy Week comes back this school year.