$1.75 Billion of False Validation: Quibi, Juicero, and the Humane Pin

There’s a question I’ve learned to ask whenever someone tells me an idea is validated: validated by what, exactly? Not as a gotcha — as an inventory. Because the answers sort cleanly into two piles. One pile is evidence: a specific person did a specific thing, recently, that they wouldn’t have done if the problem weren’t real. The other pile is enthusiasm: someone said yes to a survey, an investor wired money, an audience applauded a demo. The second pile feels like the first pile. It generates the same confidence, fills the same slide, and supports the same roadmap. It just doesn’t predict anything.

Three companies from three different eras make the cleanest case study I know for the difference, precisely because they had nothing else in common. Quibi was a Hollywood streaming service. Juicero was a kitchen appliance. Humane was an AI wearable. Different categories, different decades of founder pedigree, different failure theaters. Same autopsy: each one had a mountain of the second pile and almost none of the first, and each one’s funding was large enough to keep the question from being asked until the market asked it for them.

Quibi: what $1.75 billion can’t tell you

Quibi is the largest version of the failure, so it goes first. Jeffrey Katzenberg and Meg Whitman — two of the most credentialed operators in American business — raised $1.75 billion to build “quick bites”: premium, ten-minute-ish shows shot for your phone, for watching in the in-between moments of a commute or a queue. The logic was top-down and it was airtight in the way top-down logic always is. People watch video on phones: true. People have gaps in their day: true. Hollywood production quality beats user-generated content: arguable but defensible. Therefore people will pay a monthly subscription for premium short-form mobile video: completely untested, and load-bearing.

The projections said 7.4 million subscribers in year one. The service launched on April 6, 2020, and shut down on October 21, 2020 — roughly six and a half months later — with about 500,000 subscribers. Along the way, Sensor Tower estimated that only around 72,000 people were actually paying as of July 2020; Quibi disputed that figure, and I’d carry the dispute, but even the company’s own end-state number was about seven percent of the year-one projection. Around $350 million went back to investors. Everything else went into content, marketing, and a distribution deal for an assumption nobody had ever watched a real person act out.

Here’s the part I find genuinely instructive rather than merely expensive. Quibi did research. There were surveys, focus groups, market-size analyses — the full apparatus. What none of that apparatus contained was a single observation of the kind a Mom Test interview is built to produce: not “would you watch premium short video on your phone?” (of course people say yes; saying no is rude and costs nothing) but “what did you actually watch, on your phone, in a spare ten minutes, yesterday?” The honest answer in 2019 was YouTube, TikTok, Instagram, Netflix — all free or already paid for, all watchable on a TV later, all shareable. Every one of those properties contradicted a Quibi design decision, including the launch-day inability to cast to a television or screenshot a scene to share. The behavioral answer was available for the price of watching commuters for a week. The hypothetical answer cost $1.75 billion.

Juicero: the thirty-second test nobody ran

If Quibi is the failure at maximum scale, Juicero is the failure at maximum purity, which is why it became the shorthand. The company raised about $120 million — from serious, name-brand investors — for a Wi-Fi-connected machine that pressed proprietary packs of chopped produce into juice. The machine sold for $400. The entire value proposition rested on one physical claim: you need the machine. The press had to do something your hands could not.

On April 19, 2017, Bloomberg reporters squeezed the packs by hand. It took about ninety seconds to get nearly the same amount of juice; some renditions of the test came in faster. The video went exactly as viral as you’d expect, and the company was dead by September 1, 2017 — a little over four months from falsification to shutdown.

The detail that keeps this story in my head is not that the test was devastating. It’s that the test was available the whole time — to the founders, to the engineers, to every partner who diligenced the deal — and cost thirty seconds and one juice pack. This is the sharpest possible illustration of a distinction I keep returning to from the discovery frameworks post: a value proposition isn’t validated when smart people believe it; it’s validated when it survives the cheapest test that could kill it. Juicero’s cheapest lethal test was a hand squeeze. Nobody inside the money ran it, because $120 million of committed capital is a powerful reason not to go looking for the experiment that might embarrass the thesis. The reporters ran it because they had no position to protect. That asymmetry — outsiders can afford falsification, insiders feel they can’t — is the whole disease.

Humane: applause at the demo, returns at the warehouse

Then the failure got an AI era. Humane’s Pin launched in November 2023 at $699 plus a $24-a-month subscription, founded by ex-Apple leaders, demoed at TED to genuine acclaim, and positioned explicitly as the beginning of the end of the smartphone. That positioning choice was the discovery failure in one sentence: it staked the product on the single claim easiest to falsify with behavioral evidence — that people would leave the most validated product in consumer history in a drawer.

The numbers came in the way the interviews would have, had anyone run them against behavior instead of imagination. Roughly 10,000 units sold against an internal target of 100,000. By mid-2024, leaked internal figures reported by The Verge suggested daily returns were exceeding daily sales — a claim I’d flag as reported rather than confirmed, but one the endgame did nothing to contradict. In February 2025, HP bought Humane’s assets for $116 million, against roughly $230 million the company had raised, and on February 28 the devices were bricked. People had paid $699 for hardware that stopped working because the vision behind it met the market and lost.

What did Humane have instead of evidence? The best-looking pile of enthusiasm the industry can produce: elite pedigree, a standing-ovation demo, breathless launch coverage, and hundreds of millions in committed capital. Every element of that pile answered the question “is this impressive?” None of it answered the question a behavioral test asks: when this person needed to text someone, check a map, or kill four minutes yesterday, what did they reach for, and what would this device have to beat? The answer was a phone, the phone was already in their pocket, and the Pin — slower, screenless, projecting onto your palm in daylight — beat it at nothing. Demo applause measures the performance. Return rates measure the product. Humane had validated the performance.

The through-line: capital is anesthesia

Line the three up and the pattern stops looking like three colorful startup stories and starts looking like one mechanism wearing three costumes. Quibi in 2020, Juicero in 2017, Humane in 2024. Streaming, hardware, AI. $1.75 billion, $120 million, roughly $230 million. In every case there existed a cheap, fast, behavioral test that would have falsified the core premise — watch what commuters actually watch; squeeze the pack; ask a hundred people what they did the last time they deliberately left their phone behind. In every case that test went unrun, while expensive, slow, opinion-shaped “validation” accumulated instead.

And here’s the inversion that I think is the actual lesson, the one that generalizes beyond schadenfreude: the money didn’t just fail to buy evidence — it actively substituted for it. A bootstrapped founder meets reality in month three, because payroll forces the question. A founder sitting on nine figures can run on projections, surveys, and applause for years, because nothing forces contact with users until the burn does. Capital extends the runway of a true premise and a false one identically. The more of it you have, the longer a false premise can outrun the collision — which means the size of the raise and the quality of the validation aren’t merely uncorrelated. Past a certain point they can be anti-correlated, because the raise itself becomes the thing everyone points to when asked how they know. Investors wired money: validated. Katzenberg believes: validated. TED stood up: validated.

None of those is a user doing a thing. All of them are people with reputational or financial positions expressing enthusiasm — the exact signal the Mom Test exists to teach you to distrust when it comes from a single customer being polite over coffee. The failure of Quibi, Juicero, and Humane is that entire capital structures behaved like that polite customer, at nine and ten figures, for years.

The fix is not more research budget. All three had research budgets. The fix is insisting that somewhere in the pile marked “validation” there exists at least one item of the first kind: a specific person, a real past behavior, a falsifiable claim tested at its cheapest point. That item costs almost nothing. Its absence, on this evidence, costs about two billion dollars.

Put it to work

1. Write down your idea’s hand-squeeze test. For your current bet, name the single cheapest experiment that could falsify the core premise — not the roadmap, the premise — and state what result would kill it. If you can’t name one, the premise isn’t specific enough to be tested, which is worse. If you can name one and haven’t run it, ask honestly whether that’s an accident or self-protection.
2. Sort your evidence file into the two piles. Take everything currently cited as validation for a live initiative and label each item: behavior (someone did something costly — used, paid, returned, churned, built a workaround) or enthusiasm (someone said something — survey answers, stakeholder excitement, funding, applause). If the behavior pile is empty, you have Quibi’s file with smaller numbers.
3. Treat committed money as a bias to correct for, not a signal to cite. The moment budget is allocated, everyone attached to the initiative — including you — acquires a reason not to run the killing test. Schedule the falsification attempt before the commitment ceremony, and have someone without a position run it. Bloomberg found Juicero’s truth in thirty seconds because Bloomberg had nothing invested in not finding it.

Further reading

• Variety, “Quibi Is Shutting Down” (October 2020) — the shutdown report with the subscriber arc from 7.4 million projected to roughly 500,000 actual, six months after launch.
• Bloomberg, “Silicon Valley’s $400 Juicer May Be Feeling the Squeeze” (April 2017) — the hand-squeeze test itself; still the best single artifact in the falsification-over-enthusiasm canon.
• TechCrunch, “Humane’s AI Pin is dead, as HP buys startup’s assets for $116M” (February 2025) — the endgame numbers: units versus target, raise versus sale price, and the bricking date.
• Rob Fitzpatrick, The Mom Test — the interviewing discipline that separates the two piles at the level of a single conversation; these three stories are the same discipline failing at the level of a capital structure.