Google Fellowship
In-Person Institute
In-Person Institute — Impact Challenge Showcase 🔘
Thu, Sept 17 (9am–5pm) & Fri, Sept 18 (9am–1pm), 2026 · 1195 Borregas Ave, Sunnyvale, CA 94089
"The definitive outcome of your fellowship is a high-impact project, a major institutional challenge either solved or proposed using Google AI tools, which you will present at the in-person Institute showcase."
🎤 The slot doubled. Selected as a Fellowship Voices presenter, 10 minutes. The general cohort guidance is a strict 6-minute Day Two slot covering presentation and Q&A. I am building the 10-minute version and holding a 6-minute cut ready. Ten minutes is room for one real case study, which six minutes could not carry. Slide-by-slide running order is on the My Presentation page.
⏱ Hard deadline: Friday, August 21. The semester starts the following week and this work stops. September is rehearsal only, in 30-minute blocks, nothing that requires thinking. The one open item outside my control is Google's format spec for the 10-minute slot, and I am not waiting for it. The deck is modular, so if the spec differs, adjusting is a 20-minute edit in September rather than a rebuild.
The lead case study is now Think Gemini
By Thalia Merino, Vashtie Persaud, Elias Cherrier-Vickers, and Hailey Ho-Sang. A Young Ones One Show brief for Gemini, twelve weeks, on responsible AI use in the classroom. It answers a Gemini brief in front of Google, its subject is the fellowship's own subject, and the team used the same empathy-interview methodology Google taught us. Bunny Quest moves to reserve for hallway conversations. Google named four accepted formats, and CPA is two of them at once: a case study plus a governance framework. Both words go on the leave-behind, because they are Google's own categories.
The three prep deliverables Google assigned
Problem Breakdown
A Socratic audit of the institutional challenge
Empathy Interviews
Insights from key campus stakeholders
Vulnerability Map
Exactly where the current prototype's blocks exist
Optional workbook for capturing this work: Resource Hub
1. Problem Breakdown — the Socratic audit
| Lens | Google's question | My answer, in short |
|---|---|---|
| Knowledge | What are the verified facts? | FIT has no AI detection tools and no plan to build any. AI produces a polished UI in seconds. Students on this workflow won 21 Young Ones awards, a Gold Cube for AI Process, and 6 hires before graduation. |
| Assumptions | What are we taking for granted? | That the artifact still proves authorship. That process evidence can be captured without adding unpaid labor. That students document honestly when documentation is graded. |
| Evidence | Where is the data? | External validation from award juries and employers. Missing: an in-classroom control comparison and inter-rater reliability across the 8 dimensions. |
| Perspectives | What might I be missing? | Students may read process capture as surveillance, though the one student I have asked directly (Thalia Merino, Aug 11) called it support, with one condition: the record has to be verbal, because a written trail can itself be AI-generated. Adjuncts carry the grading load. FIT legal has not weighed in on student IP. Employers may not yet ask for a process portfolio, though recorded case-study walkthroughs are already standard in design job interviews. |
| Implications | Second-order effects? | Process grading could reward good documentation over good design. It could also create the first defensible student IP record in AI-assisted design. |
| Reflection | How do my biases shape this? | I teach the workflow I am proposing to assess, so my pilot evidence is my own students. I need outside raters to test whether the rubric holds in someone else's hands. |
2. Empathy Interviews Complete
- James Pearce, Technical Manager for Faculty Research Space & Emerging Technology, June 17. The institutional constraint. Read the report
- Thalia Merino, Motion Design & Art Direction on Think Gemini, August 11. The student view, plus a documented self-correction. Read the report
Two interviews is the answer, and it is a decision rather than a shortfall. A third and fourth were on the plan and I cut them on Aug 11. Pearce gives the institutional constraint, Merino gives the student view. August availability is unreliable, and two solid write-ups beat four thin ones. A chair or dean on adoption cost, and a faculty peer outside CT&D on whether the rubric travels, are both worth doing in the fall, not before Sept 17.
3. Vulnerability Map
Updated Aug 12 from the Merino interview. Naming the blocks out loud reads as rigor, not weakness.
| Block | Why it's real | Mitigation |
|---|---|---|
| Faculty grading time | Eight dimensions across a full studio section is more grading, not less | Rubric templates, and the verification layer is critique, which already happens in every class. CPA scores a conversation faculty are already having rather than adding an event to the calendar. It replaces some written feedback rather than adding to it |
| Evidence capture burden | Students must log prompts and decisions while designing | Build capture into existing deliverables and into critique, which is already on the syllabus. Gemini notebooks as the container, not a separate system |
| No detection infrastructure | Nothing verifies that a decision log is authentic | Live presentations are the verification. A student who can't explain a decision out loud didn't make it. Thalia Merino named this failure mode unprompted and proposed the same fix: "I wouldn't recommend 100 percent written, simply because it's easy for people to just AI it" |
| Surveillance, as students experience it | Evidence capture can feel like monitoring rather than support, and a student mid-semester has no reason yet to see the professional value in it | Keep the record verbal and make it the same artifact the industry already asks for, a recorded walkthrough of your own decisions. Thalia only recognized the value after graduating, through portfolio reviews and job interviews, which means the framing has to be taught, not assumed. Measure trust between faculty and students as a first-year metric, because it is the signal that tells me whether the risk is real |
| IP and legal | Student IP protection is asserted but not yet reviewed by FIT counsel | Get a legal read before scaling. Frame the timestamped record as evidence, not as a guarantee |
| Rater reliability | Only I have scored against this rubric | Recruit two faculty to score the same three projects and compare |
| Tool churn | Stitch, Nano Banana, and Gemini all change under me | The rubric grades judgment, not tools. Tool names sit in an appendix, not in the dimensions |
The sprint (Aug 11 → Aug 21)
Five blocks. Nothing here is optional, and nothing outside here happens before September.
| By | Block | Status |
|---|---|---|
| Aug 11–13 | The deck. 11 slides built from the script, Thalia's assets placed, both videos cut, PDF backup exported | Drafted |
| Aug 13 | Decide where outside pilot data goes: second edition, an online companion Field Guide, or the AI Pedagogy Toolkit | Open |
| Aug 14 | The leave-behind, one page both sides. Rubric on the front, the three error types and the critique loop on the back, licensed CC BY 4.0. Content written Aug 20, layout still to do. The rubric page is the online version | Layout |
| Aug 17 | Problem Breakdown written across all six Socratic lenses, naming the single strongest assumption CPA rests on and what would disprove it | Now |
| Aug 18 | Vulnerability Map finalized, Merino interview written up, deck sent to Thalia | Done |
| Aug 19–20 | Three timed run-throughs at 10 minutes targeting 9:00, one timed run of the 6-minute cut, tech check | Next |
| Aug 21 | Buffer. If everything above is done, do nothing. That is what the day is for | Buffer |
September, light touch only
- Week of Sept 8: one timed run-through, 20 minutes
- Sept 15: print the leave-behind
- Sept 16: final tech check
- If Google's spec changed anything, adjust timing only. Do not reopen the content
Cut from the plan, on purpose
Named so the decisions stay visible instead of quietly reappearing in September.
- Bunny Quest and CT302 mapped against the 8 dimensions. Think Gemini carries the case study alone. Bunny Quest stays as a verbal reserve for hallway conversations
- The AI Pedagogy Toolkit specification. This is the ask at the Institute, not a deliverable for it
- An adoption pathway document. Replaced by the book
- A third and fourth empathy interview. Two solid write-ups beat four thin ones
- Futures Lab research and the Pomelli Phase 2 teaser. Both are conversations to have at the Institute, not prep to do before it
- Additional draft-to-final comparison captures. Thalia's assets already do this job
Slide-by-slide running order, measured timings, and the cut order are on the My Presentation page.
Logistics
See the Know Before You Go doc for hotel recommendations. Google will update that same document with the formal agenda closer to the event.