About the unverified marks on this pageFigures tagged unverified are PACE-local estimates that could not be confirmed against any published source, so they are labelled rather than quietly presented as fact. Every other factual claim on this page — statutes, standards codes, gene biology, primary-source citations — was independently checked; the full claim audit, including what came back wrong, lists every claim with its source.
Try it Tuesday · 90 minutes
Run this next week
90 minutes, no fabrication. Minutes 0–15: each team writes one value proposition and, beneath it, the single riskiest assumption it depends on, in one sentence. Minutes 15–45: teams use the landing-page prompt to generate a one-file site plus two headlines that make genuinely different promises, then paste all three headlines into a shared doc. Minutes 45–75: a hallway test — every student walks to another team, shows the three headlines cold, and records which one the stranger understood first and what they thought the product was. Minutes 75–90: teams report the gap between the headline they loved and the one that actually communicated. The deliverable is the hallway-test notes, which grade whether the team can tell traction from self-flattery.
The full assignment
Make it 3D print · laser cut
A physical MVP good enough to put in a customer’s hands. Not a finished product: a form study, a fit test, a packaging mockup. The point is that customer discovery interviews change completely when there is an object on the table.
Build it AI chat · no code
A working landing page for the venture, built in one prompt, with an email capture and three headline variants. Students run traffic from their own networks and report the actual conversion rate, not their opinion of it.
4 weeks, realistically
The 4-week version
- Wk 1Teams commit to one value proposition and write the riskiest assumption underneath it in a single sentence.
- Wk 2Build the landing page live in class and submit the print or cut file at the start of the week so fabrication runs in parallel with the digital test.
- Wk 3Run the test: fifteen real interviews and whatever traffic they can generate. No pivoting yet, just collection.
- Wk 4Prototype in hand, data in hand, teams present what they were wrong about. Grade the quality of the disconfirmation, not the size of the number.
Starter prompt & what the model actually did
The prompt, run for real
Build me a single-page landing site for a new product, as one self-contained HTML file I can host anywhere. The product is: [describe it in two sentences]. The customer is: [describe them]. Include a headline, three short benefit statements, an email signup field, and nothing else. Then give me two alternate headlines that make a meaningfully different promise, not just reworded versions of the first. Keep the copy plain and specific; no marketing filler.
Here's the file. Save it as `index.html` and host anywhere.
```html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Split Bowl</title>
<style>
:root { color-scheme: light dark; }
* { box-sizing: border-box; }
body {
margin: 0; font-family: system-ui, sans-serif;
display: grid; place-items: center; min-height: 100vh;
padding: 24px; line-height: 1.5;
}
main { max-width: 440px; width: 100%; }
h1 { font-size: 1.9rem; margin: 0 0 24px; }
ul { list-style: none; padding: 0; margin: 0 0 28px; }
li { padding: 10px 0; border-top: 1px solid #8884; }
form { display: flex; gap: 8px; }
input {
flex: 1; padding: 12px; font-size: 1rem;
border: 1px solid #8888; border-radius: 6px; background: transparent; color: inherit;
}
button {
padding: 12px 18px; font-size: 1rem; border: 0; border-radius: 6px;
background: #1a1a1a; color: #fff; cursor: pointer;
}
@media (prefers-color-scheme: dark) { button { background: #fff; color: #111; } }
</style>
</head>
<body>
<main>
<h1>Grab plate lunch. The rice stays crisp, not soggy.</h1>
<ul>
<li>Two sealed compartments keep rice and toppings apart until you eat.</li>
<li>Collapses flat to fit in a backpack between classes.</li>
<li>Reusable silicone — skip the styrofoam takeout box every time.</li>
</ul>
<form onsubmit="this.innerHTML='<p>Thanks — we\'ll email you.</p>';return false;">
<input type="email" required placeholder="you@hawaii.edu" aria-label="Email">
<button type="submit">Notify me</button>
</form>
</main>
</body>
</html>
```
The form just shows a confirmation on submit — wire the input up to Mailchimp, Buttondown, or a Google Form action when you're ready to actually collect addresses.
**Two alternate headlines** (different promise, not reworded):
1. **"Eat between classes without buying another takeout box."**
Leads on cost and waste reduction — the recurring-savings angle — rather than food quality.
2. **"Packs flat. Holds a full plate lunch. Fits your backpack."**
Leads on portability and capacity — the "it actually fits in your bag" objection — rather than freshness or waste.
The original headline promises food quality (crisp rice); these two promise saving money and fitting your gear. Pick based on which objection you think stops students from buying.
- Unsupported benefit claim in the headline: 'The rice stays crisp, not soggy'Asserted as a product fact before any prototype exists — the model manufactures a benefit the founder has not tested.
- Genuinely different alternate headlines (cost/waste vs. portability)The two alternates make distinct promises rather than reworded versions, which is what the prompt demanded and what makes them worth A/B testing.
- Honest scaffolding note: 'wire the input up to Mailchimp... when you're ready to actually collect addresses'The model flagged that the email capture is non-functional — a place it correctly declined to fake capability.
- Fabricated audience psychology: 'the objection you think stops students from buying'Presents a hypothesis about customers as if known; a student should treat it as the thing to test, not a finding.
Assessment
Rubric
| Criterion | Weight | What it assesses |
|---|---|---|
| Riskiest-assumption clarity | 20% | The stated riskiest assumption is genuinely the thing most likely to sink the venture, specific enough to be tested. |
| Quality of disconfirmation | 35% | Grades whether the team sought evidence that would prove them wrong and reported it honestly, rather than counting vanity signals. |
| Signal vs. noise interpretation | 25% | Distinguishes a real preference signal from politeness, curiosity, or the team's own hope. |
| Iteration decision | 20% | The pivot-or-persevere call is justified by the evidence collected, not by attachment to the original idea. |
What this replaces
Swapping it into a real course
Replaces
The written business plan, or the polished pitch deck, submitted at the end of the venture course.
What is lost
The comprehensive financial and operational planning a full business plan forces.
What is gained
Students get a real signal from a stranger in one session instead of a graded document nobody outside the class will read; the assignment rewards disconfirmation over presentation polish.
Where AI is bad at this
The failure your students should catch
Ask a model to build a landing page and it produces confident, conversion-optimized copy for a market it has never observed — asserting benefits, inventing a target-customer objection, and implying demand no interview has established. Its fluency at marketing language makes an untested idea look validated, which is the precise opposite of what customer discovery is for.
Budget & logistics
What it costs to run
- Instructor prep1 hours
- Class time90 minutes
- Per-student cost$0 for the Tuesday version; roughly $8–15 per team in print or cut stock unverified for a physical MVP in the four-week version.
- Fabrication file dueFirst day of Week 2 — submit the print or cut file at the start of the week the landing page goes live. PACE quotes 7–10 business days (up to 14 calendar days), so it clears before the week-4 presentation.
- Calendar dependencyThe digital test needs no lead time. Submitting the MVP file at the start of week 2 clears the up-to-14-calendar-day turnaround before the week-4 presentation.
Three sizes
Scale it to the time you have
One session
The 90-minute value-proposition and hallway-test session above; no fabrication.
4 weeks
The seeded four-week plan: riskiest assumption, live landing page plus fabrication file, real interviews and traffic, then a what-we-were-wrong-about presentation.
One semester
Run a real venture through a full build-measure-learn loop with multiple pivots, a fabricated MVP iterated twice, and a final report graded on what the team was demonstrably wrong about.
Reserved · live build
This space is intentionally empty. During the Day 8 session it will be filled in live — fill the liveBuild field in content/entrepreneurship.json and rebuild.
Tell us what happened
Run it, then say how it went
If you try this — the 90-minute version, the prompt, any part of it — send back what you tried and what happened, especially anywhere the model was confidently wrong. That is the material the next session is built from.
Email feedback on Entrepreneurship →
Opens a draft in your mail client, already addressed and titled. No form, no account, nothing to sign up for.