Notes on Startups, or How to Lose the Future
2026-07-15
“What important truth do very few people agree with you on?”
— Peter Thiel, Zero to One (2014)
“It is well known that this is a very, very difficult environment for business-to-consumer Internet companies.”
— Julie Wainwright, announcing the liquidation of Pets.com, November 7, 2000
“Play long-term games with long-term people.”
— Naval Ravikant, “How to Get Rich (Without Getting Lucky)” (2018)
Form D, filed with the Securities and Exchange Commission, July 2014: the founders of Secret, a company then ten months old, had taken six million dollars off the table. The company was shut the following spring.
“The graveyards are full of indispensable men.”
— attributed, like most things the Valley believes, to somebody who never quite said it
The story is usually told this way. During the Second World War, officers studying the damage on returning bombers proposed more armor where the holes clustered. Abraham Wald pointed to the planes absent from the survey: damage common among survivors marked places an aircraft could be hit and still return; scarce damage around vulnerable areas, including the engines, reflected aircraft that had not come back. The scene is a later simplification. Wald’s actual wartime work addressed the harder statistical problem beneath it: how to estimate an aircraft’s vulnerability from damage recorded only on survivors [1].
Silicon Valley has built a lavish advice industry from the planes that came back. The commencement speaker, the podcast sage, the author of the airport hardcover with the one-word title: each can describe the pivot, the hire, or the refusal to sell that preceded success. None can establish from survival alone that the decision caused it. Similar decisions were made in less celebrated dorm rooms by founders whose companies are now known chiefly to the Wayback Machine and the bankruptcy courts of Delaware. The explanation followed the outcome and often borrowed its authority from it.
In 2012 a Stanford law student named Blake Masters circulated notes from Peter Thiel’s class; two years later, revised with Thiel, they became Zero to One [2]. The book instructs the young that competition is for losers, that the future belongs to the founder with a secret, and that going from nothing to something is the only movement worth making. It is fluent, confident, and occasionally brilliant. It is also a survivor’s account of an industry in which one investment can equal or outperform the rest of a successful fund combined. That arithmetic explains the attraction of venture capital. It says much less about the costs inside the losing companies, or about the ease with which preventable failures of governance can be recorded as acceptable portfolio variance.
This is a book of counting. It goes from one to zero, deliberately, the wrong way down the great man’s number line, and proposes to do for the startup what a coroner does for a preferred family story: establish what happened.
I should declare my credential, which is proximity rather than prophecy. I watched Clinkle at close range: a Stanford payments startup that announced a $25 million seed round, then the largest in Silicon Valley, and had raised about $30.5 million by late 2013 [3]. Its celebrated payment technology was abandoned; experienced executives departed; the company eventually launched a much smaller prepaid-card product called Treats [4]. Some noteholders later asked for repayment, and no final public accounting has been located. Employees lost jobs and years. No commencement speech resulted, which is why the company belongs in this book.
The companies differ, but the institutional roles recur. Founders learn to present biography as evidence. Venture capitalists advertise contrarian judgment while treating another prominent investor’s check as validation. Some business coverage republishes the ascent; other reporters later perform scrutiny that investors and boards omitted. Employees appear in survivor literature as headcount. Here they are the people who paid for the education that everyone else claims to have received.
The book is braided rather than episodic. Separate case studies flatter each disaster with the illusion of uniqueness. Told stage by stage — the promising start, early financing, peak confidence, first warnings, fall, and aftermath — the companies expose recurring incentives and recurring failures of scrutiny. The braid does not make the protagonists interchangeable: fraud is not a mistaken hypothesis, and an orderly wind-down is not a deception. It does let us ask when failure was a necessary experiment and when money, status, or weak governance allowed avoidable harm to continue.
Failure is a price of innovation, but the phrase settles neither the size of the bill nor who paid it. Employees bore lost wages, worthless options, and interrupted careers; limited partners supplied the capital; general partners collected management fees whether an individual investment won or lost, while carry depended on the fund producing profits. The subject of this book is not failure alone. It is the transfer of downside, and the stories used to make that transfer look natural.
The missing planes never reached the lecture hall. This book asks what their absence conceals.
or, How to Lose a Billion (Without Even Trying)
Naval Ravikant’s 2018 thread “How to Get Rich (Without Getting Lucky)” compresses wealth into numbered principles spoken after success [1]. The form is ideal for survivor bias: the result stands beside the maxim, and proximity impersonates proof.
Here is the counter-catechism. It does not claim that the familiar advice always loses. It asks what the aphorisms omit when the same advice precedes opposite outcomes.
1. Seek wealth, not revenue. A valuation arrives by negotiation; revenue arrives from customers. Clinkle’s announced $25 million seed round preceded a public product, Quibi’s $1.75 billion preceded launch, and Fast’s valuation, reported near half a billion dollars, accompanied roughly $600,000 in 2021 revenue [2]. The capital was real. What it proved about demand was not.
2. Rent judgment briefly. Barry McCarthy, the former Netflix chief financial officer, joined Clinkle as chief operating officer in October 2013 and left less than five months later, saying the company was “not nearly as close to scaling” as he had believed [3]. His appointment reassured outsiders; his departure supplied contrary evidence.
3. Play short-term games with people holding long-term promises. Secret’s founders sold $3 million of personal stock apiece in its July 2014 Series B. The company closed nine months later and returned roughly $10 million of the $35 million it had raised; no report indicated that the founders returned their secondary proceeds [4]. Employees held the less liquid instrument: faith in a company that had already paid its principals.
4. Prefer charisma when expertise is inconvenient. Elizabeth Holmes’s early meeting with MedVenture Associates ended when she could not answer basic technical questions. Later investors and directors brought political, military, and commercial stature, but little laboratory expertise [5]. Prestige widened the circle without answering the original questions.
5. Take risk through a stack of other people’s claims. Limited partners supply the capital. General partners receive management fees while a fund operates and earn carry only if it produces sufficient profits. Employees receive illiquid options, and founders may sell liquid shares in later rounds.
6. Borrow distribution and call it leverage. BranchOut reached 25 million registered users by riding Facebook’s viral channels. When Facebook restricted those channels, BranchOut’s active audience contracted sharply [6]. A borrowed channel can be useful; it remains an asset whose terms another company can change.
7. Master fundraising before the customer. A large round can buy time to find a market. It can also weaken feedback when founders, employees, and the press mistake investor competition for customer demand. Clinkle’s financing demonstrated an appetite for Clinkle shares, not for its unlaunched payment product.
8. Scale only the assumptions you have tested. Blitzscaling, in its careful formulation, accepts inefficiency when speed matters and product-market fit exists [7]. Webvan committed to a roughly $1 billion, twenty-six-warehouse expansion before its first facility had proved the economics; MoviePass sold unlimited admission for less than the price of one ticket in many cities. They scaled assumptions faster than evidence.
9. When a metric embarrasses you, invent another. WeWork’s 2018 bond documents presented “community-adjusted EBITDA,” excluding marketing, design, and administrative expenses to turn a $933 million net loss into a positive $233 million measure [8]. The new category changed the presentation, not the cash.
10. Set the culture early; people copy rewarded behavior. A company’s culture is visible in who may challenge a claim, how leaders announce a layoff, and what happens to the person who reports bad news. Slogans matter less than the conduct that receives money, status, and protection.
11. Protect the secret until scrutiny becomes impolite. Secrecy can protect intellectual property. It can also prevent customers, partners, employees, and directors from testing what they have been told. Walgreens entered its Theranos partnership without independently verifying the technology; Clinkle’s secrecy concealed the distance between its demonstration and the product it could ship [9]. In both cases secrecy increased the information gap. It did not, by itself, prove fraud.
12. Let the power law turn each loss into fund-level variance. The power law is real: in a successful venture fund, the best investment may equal or exceed the rest of the portfolio combined [10]. It does not tell us whether any particular failure was unavoidable, whether governance prolonged it, or who absorbed the cost. That accounting begins where the aphorism ends.
Seek leverage, move fast, escape competition, believe. Each instruction can precede success or failure because each leaves out the conditions that matter: what was tested, who could challenge the founder, how much time the capital purchased, and who paid when the thesis failed. The aphorism supplies no way to tell which result is coming.
Promising starts
In the autumn of 2011, a nineteen-year-old Stanford undergraduate named Lucas Duplan returned from a term in London frustrated by the work of exchanging currencies. He came home convinced that the wallet should disappear and that he could make it happen. Within two years, a demonstration of a proposed system for moving money between telephones by high-frequency sound had helped attract Accel, Andreessen Horowitz, Peter Thiel, Intel, the founders of VMware, and Sir Richard Branson. Clinkle announced a $25 million seed round, then the largest in Silicon Valley; securities filings later put its total financing at about $30.5 million [1]. No public product had launched. Its Aerolink technology worked in controlled demonstrations but failed amid the noise of an ordinary coffee shop [2]. The financing arrived before the field test.
Young companies must describe a future before they have much operating evidence. A founding narrative is useful compression; it becomes dangerous when treated as validation, or when the prestige of the people financing it substitutes for a test. Vivid stories did not doom these companies. Narrative and abundant capital did weaken feedback and postpone scrutiny, as contemporaneous documents show: what investors knew, what they could have tested, and which warnings existed before the spectacular rounds.
· · ·
Adam Neumann was raised in part on Kibbutz Nir Am, served five years as an officer in the Israeli navy, and moved to New York with his sister to attend Baruch College, which he did not finish. His first venture, Krawlers, sold padded clothing for infants under the slogan “Just because they don’t tell you, doesn’t mean they don’t hurt” [3]. In 2008 he and the architect Miguel McKelvey, who had grown up in a commune-like collective in Eugene, Oregon, worked with the landlord Joshua Guttman to subdivide vacant DUMBO space into eco-branded offices called Green Desk. They sold their interest and opened WeWork in 2010; its first location at 154 Grand Street was reportedly profitable within a month [4]. The insight was real: post-crash Manhattan had empty rooms and independent workers who did not want to sit alone. The later distortion came when a co-working business was presented as “the world’s first physical social network,” and expansion itself began to stand in for proof of the underlying economics.
The origin story already contained a discrepancy. WeWork lore described a $15 million commitment from the Brooklyn real-estate investor Joel Schreiber, who told the founders, “Whatever it takes, I want to be a partner with you.” In a later deposition, Schreiber testified that he had delivered between $1 million and $2 million [5]. Subsequent profiles repeated the commitment as though it were cash received.
Elizabeth Holmes left Stanford at nineteen with her tuition money and a proposal for a wearable diagnostic patch. Phyllis Gardner, a professor at Stanford’s medical school, told her the science would not work [6]. Holmes built a public identity around the black turtleneck, an unusually deep voice, Steve Jobs, and Thomas Edison. But an early meeting with MedVenture Associates, whose partners understood diagnostics, ended when she could not answer basic technical questions [7]. Later directors brought distinction in government, defense, and diplomacy rather than laboratory medicine. The contrast is not between dull experts and visionary outsiders. It is between questions that tested the device and credentials that reassured people without testing it.
Quibi began with the opposite credential: two long records of conventional success. In August 2018 Jeffrey Katzenberg and Meg Whitman announced that NewTV, later renamed Quibi, had raised about $1 billion before it had a public product or a market test. A second round brought the pre-launch total to $1.75 billion [8]. The investors included Disney, NBCUniversal, Sony, WarnerMedia, Viacom, Lionsgate, MGM, Alibaba, Goldman Sachs, and JPMorgan. Strategic investors may have been buying a hedge against disruption as much as a forecast of demand, but their names still looked like validation. At the January 2020 CES keynote, Whitman compared Quibi’s short episodes to the chapters of The Da Vinci Code. The premise — that viewers wanted Hollywood drama cut into chapters of less than ten minutes and available only on a phone — was tested at full industrial scale rather than in a cheap pilot.
· · ·
The same substitution was visible in 1999.
Greg McLemore and Eva Woodsmall had launched the first Pets.com through the Pasadena incubator WebMagic. In early 1999 Hummer Winblad acquired the domain and recruited Julie Wainwright, fresh from Reel.com, as chief executive; Amazon invested in March. Its press release survives. “We invest only in companies that share our passion for customers,” Jeff Bezos said. “Pets.com has a leading market position, and its proven management team is dedicated to a great customer experience, whether it’s making a product like a ferret hammock easy to find, or help in locating a pet-friendly hotel” [9]. Wainwright called it “a marriage made in heaven.” The company’s prospectus supplied a less celestial measure: in its first fiscal period, Pets.com spent $11.8 million on advertising to produce $619,000 in revenue and sold merchandise for roughly one-third of what it had paid for it [10].
Webvan had a founder with a prior success. Louis Borders had helped transform bookselling, and now proposed to rebuild grocery distribution around automated warehouses and thirty-minute delivery windows. Benchmark and Sequoia each invested $3.5 million in the Series A; SoftBank, Goldman Sachs, and Yahoo followed. Randall Stross, embedded with Benchmark’s partners for eBoys, recorded Borders saying, “I don’t see any reason why an Internet company should take five to ten years to be profitable” [11]. The firm also heard its own objections. “You’re all about marquee players,” Bob Kagle told the deal’s champion; “salesmen are more likely to be sold,” Andy Rachleff added [12]. Benchmark proceeded, and Webvan later committed to a roughly $1 billion contract with Bechtel for twenty-six automated distribution centers before the first had proved profitable.
Boo.com had three founders: Ernst Malmsten, Kajsa Leander, and Patrik Hedelin. They had already built the Swedish online bookseller bokus.com, giving the new venture more than glamour [13]. Boo proposed a fashion store launching in eighteen countries and seven languages at once, with rotating three-dimensional products, an animated assistant called Miss Boo, free shipping, and free returns. J.P. Morgan advised a raise of about $135 million from investors including Bernard Arnault, Goldman Sachs, the Benettons, and the Hariri family. Larry Lenihan of Pequot Capital declined. “You are making extremely aggressive assumptions about visitor numbers to your site,” he told them, “and you are going to blow way too much money getting there” [14]. The warning preceded launch; it did not prevent the syndicate from financing the plan.
The founding of govWorks was recorded on film. Childhood friends Kaleil Isaza Tuzman and Tom Herman started the company to let citizens pay parking tickets and other municipal bills online, while Tuzman’s former Harvard roommate brought a camera. Startup.com preserved the apartment brainstorm, the term-sheet negotiations, and the friendship offered to investors as part of the case [15]. Within a year the company had roughly two hundred employees and $60 million from Hearst, KKR, Sapient, and the New York City Investment Fund. Within two, Herman had been escorted from the building and the locks had been changed.
· · ·
By the 2010s, platform growth and founder reputation could themselves be presented as the product.
BranchOut’s founding pitch was a syllogism delivered by Rick Marini, a founder with a prior exit: LinkedIn served the minority of professionals in white-collar roles; Facebook held everyone else; therefore the professional network for the rest would be built inside Facebook [16]. The company raised $6 million in 2010 and $18 million more in May 2011. At the latter round, Adweek reported that the product had roughly six thousand daily active users [17]. The discrepancy was not hidden; it appeared in the headline. Building inside Facebook made rapid distribution possible and left the terms of that distribution in Facebook’s hands.
Color raised $41 million from Sequoia, Bain, and Silicon Valley Bank immediately before the public launch of its proximity photo-sharing app [18]. The team recounted that Sequoia had said, “Not since Google have we seen this” [19]. TechCrunch, citing multiple sources, later reported that Bill Nguyen had rejected a pre-launch acquisition offer of $200 million from Google; neither company publicly confirmed the approach [20]. The financing valued Color at a reported $167 million. The app existed, but demand and retention had not been observed. When it launched, many users opened a product that depended on nearby strangers using it too and found an empty screen.
Domm Holland arrived in San Francisco with a record as an Australian entrepreneur and a simple demonstration: one-click checkout for the open web. NPR later reported that his earlier towing venture had collapsed amid payment disputes, leaving small operators with unpaid claims; one driver said he was owed about A$27,000, and a creditor’s lawyer estimated another claim at more than US$400,000 [21]. In 2019 Holland founded Fast. Stripe led its Series A in March 2020 and its $102 million Series B in January 2021 [22]. The prior dispute was part of the public record. Whether Fast’s backers missed it or discounted it, it supplied questions that ordinary reference checks could have pursued.
MoviePass offers a revealing before-and-after comparison. Stacy Spikes and Hamet Watt founded it in 2011 as a niche subscription service priced at roughly $30 to $50 a month, depending on the market. In August 2017 Helios and Matheson, led by Ted Farnsworth, acquired control, and MoviePass chief executive Mitch Lowe announced unlimited moviegoing for $9.95 a month — less than one ticket in many cities. Subscribers rose from about twenty thousand to more than three million; losses eventually exceeded $40 million a month [23]. The ownership, price, and market had all changed, so this was no controlled experiment. It did show how quickly a compelling subscriber story could outrun the cost of serving each subscriber. Lowe pleaded guilty to securities-fraud conspiracy in 2024; Farnsworth pleaded guilty to securities fraud and conspiracy in 2025 [23].
· · ·
Early-stage investors decide with incomplete information. They can test technical feasibility, speak to former colleagues, model unit economics, and examine behavior; they cannot eliminate uncertainty. Narrative becomes dangerous when it replaces those tests.
Clinkle’s demonstration made the distinction unusually plain. A former employee told Business Insider that the celebrated demo “is not even what’s going to go out… It’s not actually moving any money” [2]. A Stanford classmate recalled Duplan giving a course presentation by “channeling Steve Jobs to a T,” complete with a “one more thing” reveal [24]. The staging worked. Aerolink still failed in a coffee shop.
Prominent checks then created their own confirmation. Amazon’s investment helped Pets.com approach the public markets; Sequoia’s investment made Color more credible to the press; Stripe’s name reassured later observers of Fast. In venture capital this is called social proof. It can convey real information about prior diligence, but it can also produce an information cascade in which each participant relies on the judgment implied by the previous check.
The lesson is not that plain stories succeed and magnificent stories fail. A founding story’s aesthetic quality has no evidentiary value. The relevant questions are more prosaic: what a bag of dog food costs to ship, whether an audio signal survives a blender, whether users return after the invitation channel closes, and what happened at the founder’s previous company.
Before these companies could demonstrate durable demand, their financing became the headline: Clinkle’s record seed, Quibi’s pre-launch billion, Color’s $41 million before public use. The next stage began when money intended to fund an experiment was reported as the experiment’s success.
Spectacular early successes, and the metrics that weren’t
On the morning of November 13, 1998, a company called theGlobe.com sold shares to the public at nine dollars. Two Cornell undergraduates had assembled the message-board community for about fifteen thousand dollars of their friends’ money. The first trade printed at eighty-seven. By the close the shares stood at $63.50, a gain of 606 percent, then the largest first-day rise in the history of American initial public offerings, and the two twenty-four-year-old founders were worth roughly a hundred million dollars apiece [1]. The company’s revenue model was the banner advertisement; its product was a web page on which strangers talked to one another; it had never earned a profit. That morning supplied a more exciting product: the pop itself, the vertical line on the chart and the crowd gathered to watch it rise.
The survivor literature calls this phase “traction” and recounts it with the false modesty of a man describing his first million as a lucky break. In companies that endured, early success is remembered as evidence that the market had spoken. In the cases here, the same signals — an oversubscribed round, an exploding user chart, marquee names on the cap table — were accepted as validation before anyone established what the chart measured or what the names had verified. A price can contain information about a business; it can also contain information about the other bidders.
George Akerlof’s market for lemons shows how unverifiable quality can depress prices. A venture boom could produce the opposite bidding dynamic. A company with revenue had comparables; a company with only a story preserved the imagined upside of becoming the next Facebook. When several funds feared missing that outcome, opacity could delay falsification while competition lifted the price.
· · ·
On June 27, 2013, Clinkle announced that it had raised twenty-five million dollars in seed financing, “the largest seed round in Silicon Valley history,” a superlative supplied by the company’s publicists and repeated by the press [2]. The announcement contained neither a product nor a launch date; Clinkle would not ship a beta for another fifteen months, and the patent claims surrounding it had not produced a single issued patent [3]. It did contain a list of names: Accel’s Jim Breyer, Andreessen Horowitz, Peter Thiel, Intel, Intuit, Marc Benioff, the founders of VMware, Ross Perot Jr., assorted Stanford professors, and eventually Sir Richard Branson, eighteen-plus investors in all. At the center stood a twenty-one-year-old Stanford graduate named Lucas Duplan, who told TechCrunch, “This is not a small social app. What we’re trying to do here is fundamentally change how people trade” [2]. He added, with the solemnity of a man who had rehearsed the line: “The margin for error is zero.”
The round was assembled from more than eighteen convertible-note checks, and no investor received a board seat [2] [4]. That choice reduced formal oversight and left no lead director responsible for challenging the founder. It did not erase the investors’ ability to seek information, coordinate, withhold later capital, or negotiate different terms; they had chosen a structure that made those interventions harder. The crowded cap table also served as marketing, each recognizable name reassuring the next. Three years later, when some noteholders asked Duplan to return their money, a Silicon Valley lawyer gave the plain conclusion: “When you have convertible debt investors asking for their money back, things have gone really wrong” [4].
Alyson Shontell of Business Insider later reconstructed the mechanics from former employees. Diane Greene introduced Duplan to Jim Breyer, the investor who had led Facebook’s Series A. Breyer consulted Stanford professors and committed after a demonstration [5]. One former employee described what he saw: “It’s not actually moving any money. It doesn’t actually do anything, but it looks like it does something” [5]. Another described the pattern investors appeared to recognize: “He sells the vision of what every investor wants, which is a 20-year-old, white male, Stanford Computer Science major… He appears to be the next Mark Zuckerberg, and he carries himself that way. Investors invest in people, not products” [5]. By September the waitlist had passed a hundred thousand students and Branson had joined the syndicate, declaring, “I’m excited to be a part of the Clinkle revolution” [6]. In October, Barry McCarthy, the chief financial officer who had taken Netflix public, signed on as chief operating officer and pronounced the software “phenomenal” and the founder “super-talented” [7]. Each endorsement raised the apparent cost of doubting the ones before it.
Little hindsight is needed to identify the central risks. Some were public and others were available to the investors: no shipped product, no issued patent, no lead investor, and no board seat. The financing announcement presented several of them as virtues. Secrecy was “stealth,” the missing product was “pre-launch,” and reduced oversight was “founder-friendly.” Information had not disappeared; the auction had changed its meaning.
· · ·
Clinkle raised on a person; BranchOut raised on a number whose definition received less attention than its size. Rick Marini’s company was “LinkedIn on Facebook,” a professional network for what he called “the other 90%” of workers. In the spring of 2011 it raised an eighteen-million-dollar Series B led by Redpoint on the strength of its position atop Facebook’s viral channels [8]. At the time, the app had roughly six thousand daily active users, a figure Adweek put in its headline [9]. BranchOut’s growth team concentrated on invitations: pre-checked friend requests, quiz-bait wall posts, and other aggressive distribution tactics that Facebook then permitted. By April 2012 the registered-user counter read twenty-five million, a mark LinkedIn had needed five years to reach and BranchOut had added in roughly ninety days [8]. Mayfield led a twenty-five-million-dollar Series C near the top of that curve. “We’re not even thinking about selling,” Marini announced. “We want to go big” [8].
A registered user was someone who had once authorized the app, perhaps while dismissing a quiz about which friend would make a good boss. The count said nothing about return visits, yet it produced a large, inexpensive number for a funding chart. Daily activity measured something more demanding. BranchOut’s rounds were raised while the contrast between those measures was visible in the trade press; the investors chose the larger one.
Viddy, “the Instagram for video,” plugged into Facebook’s Open Graph in February 2012. Every clip a user watched announced itself to that user’s social circle, and the counters responded: top free app in the App Store, three hundred thousand registrations a day, and twenty-seven million registered users by May [10]. Justin Bieber invested; Mark Zuckerberg joined the service. NEA then led thirty million dollars at a valuation of three hundred and seventy million, an eighteenfold markup in three months, with Goldman Sachs and Khosla in the syndicate. NEA’s Pete Sonsini said Viddy was “positioned to be the breakaway leader” [10] [11]. TechCrunch’s Alexia Tsotsis quoted a source saying, “it’s more money than I’ve ever seen. Seriously, ever seen” [10]. Within three weeks of the round’s confirmation, Facebook announced a crackdown on auto-sharing spam. The platform dependence was not obscure; Facebook’s machinery had generated the chart on which the round was priced.
· · ·
The pre-launch round gave this method its cleanest expression. In March 2011, Bill Nguyen’s proximity-photo-sharing app Color raised forty-one million dollars before a member of the public had opened it. Sequoia supplied about twenty-five million, more, the team said, than the firm had initially given Google [12]. Nguyen recounted Sequoia’s pitch in four words: “Not since Google have we seen this” [13]. At that point this meant a demonstration and a domain name that had cost $350,000. Google reportedly offered two hundred million dollars for the company before launch and was refused [14]. Eighteen months later, Color’s assets went to Apple for a reported seven million. The users finally consulted by the launch found an empty app unless another user stood within a hundred and fifty feet, and they rated it accordingly [12].
Velocity itself became a metric. Secret, the anonymous-confession app in which Silicon Valley whispered about itself for one vicious season, reached a hundred-million-dollar valuation in under six months and raised roughly thirty-five million dollars in eight. Index’s Danny Rimer called it “essentially a form of online graffiti; a place where we can truly unburden ourselves” [15]. The round closed about ten weeks after usage had peaked. Investors were therefore pricing not only current engagement but the chance that another fund would pay more. That can be a liquid-market strategy; in a private company, the next buyer is also the exit.
· · ·
WeWork produced the period’s fullest valuation ladder: roughly ninety-seven million dollars in 2012, four hundred and fifty million in 2013, a billion and a half in early 2014, five billion by that December, ten billion in 2015, sixteen in 2016, twenty in 2017, and forty-seven in 2019 [16]. The company subleased desks throughout; one important change at each round was the mark established by the previous one. By 2014 the register included JPMorgan, T. Rowe Price, Wellington, Goldman Sachs, Harvard’s endowment, and Benchmark, “each name,” as later reporting put it, “making the next feel safe” [16] [17]. An October 2014 pitch deck projected a 2018 operating profit of $941.6 million. In 2018 the company instead reported a net loss of $1.9 billion on $1.8 billion of revenue [16]. The widening discrepancy did not prevent the next private mark.
Masayoshi Son added the decisive rung. On December 6, 2016, the SoftBank chief arrived an hour and forty-five minutes late from a meeting with the president-elect and told Neumann, “I’m so sorry, but I only have 12 minutes.” He toured WeWork’s headquarters, pulled Neumann into his car, ignored the prepared deck, sketched terms for a $4.4 billion investment on an iPad, signed in red ink, and handed Neumann the pen [17]. Neumann later put the decision at twenty-eight minutes including the walk to the car. Investors have sometimes decided quickly; what distinguished this episode was that both men retold it as evidence of conviction. Forbes published the boast in 2017, extending SoftBank’s endorsement to other institutions considering the leases beneath it.
Theranos assembled social proof from statesmen. By 2014 its board included George Shultz, Henry Kissinger, William Perry, Sam Nunn, Bill Frist, and General James Mattis: two former secretaries of state, two former secretaries of defense, and two former senators [18]. Frist was a physician, but the board had no working specialist in laboratory diagnostics. Its prestige nevertheless reassured investors including the Waltons ($150 million), Rupert Murdoch ($125 million), the DeVos family ($100 million), and the Cox family ($100 million) [18]. Walgreens added a second endorsement when it signed a retail partnership without independently verifying that the technology worked [18]. The names did not prove the blood tests; they reduced the pressure on the next party to ask for proof.
By the decade’s end the method no longer required a user chart. Quibi raised $1.75 billion before launch, the first billion while the company was still called NewTV, from Disney, NBCUniversal, Sony, WarnerMedia, Alibaba, Goldman Sachs, and JPMorgan [19]. The strategic investors may have been buying protection against the future that Jeffrey Katzenberg described before there was an app to test. Fast, the one-click-checkout company, drew similar credibility from Stripe, which led its Series A and its $102 million Series B [20]. Stripe’s participation did not validate Fast’s revenue or economics, but its name made both easier for later investors to assume.
· · ·
Public markets added disclosure requirements and short sellers, but during both booms they also served as a final, larger round. Webvan went public in November 1999, five months after launching its grocery service. Its final prospectus reported approximately $4.2 million in net sales through September 30, not the $395,000 from its first month of commercial operation that later accounts often repeat [21]. Goldman Sachs led the offering, and Webvan closed its first day valued in the billions. Pets.com followed in February 2000 with Merrill Lynch as underwriter. Its shares traded as high as fourteen dollars before beginning their descent to nineteen cents [22]. The offering documents contained warnings that the first-day price did not absorb.
Across these cases, spectacular early success consisted chiefly of prices and growth measures set inside a competitive financing process. Registered users could rise without durable use; private marks could rely on the previous mark; a famous investor could substitute for work the next investor had not done. None of those signals was meaningless. The mistake was to treat them as independent confirmation when they had often been produced by the same auction.
Survivors also exhibit famous backers and vertical charts, which is why those signs carried weight. Clinkle and Square, BranchOut and LinkedIn, differed in the product and usage evidence available at the time. The financing market treated them as members of the same promising class and paid before those differences had been tested. Later memoirs call that moment recognition. At the time, it was still a wager.
Peak hubris, and how capital rewarded it
On the twenty-third of September, 2013, in the London offices of the Virgin Group, a twenty-two-year-old named Lucas Duplan stood beside Sir Richard Branson and set fire to forty thousand dollars. The bills were fake: four banded stacks of prop hundreds, torched for the cameras to symbolize the death of cash at the hands of Clinkle, a payments company that had raised twenty-five million dollars but would never launch the ultrasonic system on which the money had been raised [1]. The photographs were judged too vivid even for Clinkle’s publicity department, which had lately produced a television spot in which people’s heads disappeared. They stayed in a drawer until 2016, when Forbes obtained them [2]. Before it shipped a product, a payments startup had posed beside a billionaire while burning money.
Memoirs tend to file such scenes under “excess,” as though a sound enterprise had suffered regrettable catering. In these cases, the extravagance also communicated certainty. Investors repeatedly rewarded founders who promised escape from ordinary margins, and the founders learned to perform that promise with jets, festivals, grand pronouncements, and public contempt for restraint. The clearest instruction came from one of the investors himself.
· · ·
As SoftBank completed its $4.4 billion investment in WeWork, Masayoshi Son put a question to Adam Neumann. In a fight, he asked, who wins: the smart guy or the crazy guy? Neumann answered: the crazy guy. Correct, said Son, but “you and Miguel are not crazy enough.” He instructed Neumann to make WeWork “ten times bigger than your original plan” and said that, on those assumptions, the valuation was cheap [3]. A major technology investor had examined a company that rented desks and treated its founder’s appetite for risk as an asset. Neumann told friends that Son had instructed him to be crazier [4]. The Neumanns called Son “Yoda.”
Neumann answered that mandate by building a court around the company. WeWork acquired a Gulfstream G650 for some sixty million dollars. On one crossing, according to later reporting, marijuana smoke drove the crew to use oxygen masks; on a flight to Israel, a cereal box containing the drug was left aboard for the return leg, and the jet’s operator recalled the aircraft [4]. WeGrow, Rebekah Neumann’s “conscious entrepreneurial school,” charged up to $42,000 a child. The company invested in a wave-pool business, a superfood concern, and a spa. Neumann’s personal holding vehicle owned the trademark “We,” and the company paid him $5.9 million for it when it rebranded [5]. Former Twitter chief executive Dick Costolo called the S-1’s self-dealing “so egregious” [6]. At an executive retreat Neumann raised a glass “to nepotism” [4]. At an all-company gathering he said WeWork would give the world’s 150 million orphans “a new family: the WeWork family” [7].
The valuation, Neumann explained in 2017, was “much more based on our energy and spirituality than it is on a multiple of revenue” [4]. It is one of the clearest admissions in the book, and the professional custodians of other people’s pensions received it as vision.
The smaller scene is more revealing. In June 2016, WeWork laid off roughly seven percent of its staff. Some weeks later, at an evening all-hands, Neumann addressed the remaining employees somberly on the necessity of the cuts. Then, according to the Journal’s reconstruction, employees entered carrying trays of plastic shot glasses filled with tequila. Darryl McDaniels of Run-DMC followed, embraced Neumann, and performed “It’s Tricky” while the trays circulated. Workers still thinking about their fired colleagues described themselves as stunned [4]. The company had turned a layoff into entertainment for the people who remained.
· · ·
In April 2018 WeWork sold bonds while presenting a metric called “community-adjusted EBITDA.” Its reconciliation began with a $933 million net loss for 2017 and arrived at positive community-adjusted EBITDA of $233 million after excluding interest, taxes, depreciation, stock compensation, and major operating expenses including marketing, design, and administration [8]. “I’ve never seen the phrase ‘community adjusted Ebitda’ in my life,” said Adam Cohen of Covenant Review; the Financial Times later called it “perhaps the most infamous financial metric of a generation” [8]. The arithmetic was disclosed. The invention lay in presenting necessary costs as though they belonged outside the business.
· · ·
At Clinkle the same conduct had a younger cast. The jobs page advertised laser tag and “runners to help with chores” [9]. When the app was breached in January 2014, the exposed database yielded Duplan’s profile photograph: the founder grinning over roughly thirty thousand dollars in banded cash. The company called the image “playful” and “self-deprecating” [10]. At an all-hands Duplan described the culture differently: “We are not Google. At Google, people ride around on bikes, smiling. We’re more like the Marines… You can’t be upset if your best friend is fired, because it just means they weren’t the best” [9]. In September, Clinkle rented its growth team a party bus to celebrate one hundred thousand waitlist signups, then fired about a dozen members by telephone that Sunday [9].
One Stanford classmate recalled Duplan delivering a presentation “channeling Steve Jobs to a T — even capping off with a one more thing reveal” [11]. Elizabeth Holmes adopted the same costume with graver consequences: black turtleneck, unblinking stare, and a voice that colleagues said she had lowered below its natural register [12]. By 2014 she was on the cover of Fortune and on the Forbes list as the youngest self-made female billionaire. She told investors that Theranos devices had been deployed on medevac helicopters over Afghanistan and projected revenue above one hundred million dollars. The Defense Department had deployed nothing; 2014 revenue, according to the Securities and Exchange Commission, was slightly over one hundred thousand dollars [12]. The presentation differed from reality by a factor of roughly a thousand.
FTX gave the performance a different costume. Sam Bankman-Fried raised a Series B extension of exactly $420.69 million from exactly 69 investors [13]. Sequoia published a profile that treated his playing League of Legends during its pitch meeting as evidence of unusual ability; a partner’s note from the meeting read, “I LOVE THIS FOUNDER” [13]. Bankman-Fried mused publicly about buying Goldman Sachs. Celebrated status coexisted with a missing chief financial officer and little independent board oversight, yet the round closed. Sequoia removed the profile in November 2022 [13].
· · ·
Age and experience offered no protection. Jeffrey Katzenberg and Meg Whitman, veterans of Disney, DreamWorks, eBay, and Hewlett-Packard, raised $1.75 billion for Quibi before launch, pre-sold $150 million of advertising, projected 7.4 million first-year subscribers, and bought Super Bowl airtime [14]. At launch, its phone-video app blocked screenshots and could not cast to a television, limiting the sharing through which mobile video found an audience. Katzenberg resisted calling the programs “shows”; they were “movies in chapters,” and Whitman compared them at CES to chapters of The Da Vinci Code [14]. The experienced team committed a ten-figure budget before learning whether viewers wanted the format.
Experienced executives had done this before. Two decades earlier George Shaheen left Andersen Consulting to run Webvan after only months of commercial operation and negotiated a supplemental pension of $375,000 a year for life [15]. Webvan also signed a roughly billion-dollar agreement with Bechtel for twenty-six automated warehouses before the first had reached break-even. Shaheen told Forbes that Webvan would “set the rules for the largest consumer sector in the economy” [15]. The SEC delayed the IPO after executives continued giving interviews during the quiet period [15]. Kozmo.com, whose couriers delivered single pints of ice cream across Manhattan free of charge, agreed to pay Starbucks $150 million over five years for co-marketing when Kozmo’s annual revenue was $3.5 million [16]. Its founder said of Blockbuster, then roughly a thousand times larger, “I’m going to put them out of business” [16]. Aboard Concorde, meanwhile, the founders of Boo.com argued about cost control while the company burned more than seven million dollars a month and supplied $100,000 apartment allowances. Kajsa Leander demanded that co-founder Patrik Hedelin “get a better grip on the financials” at roughly ten thousand dollars a seat [17].
The conduct did not require consumer glamour. Michael Marks, the industrialist who had built Flextronics, took more than two billion dollars of SoftBank’s money into Katerra to create “the Foxconn of buildings.” The company bought architecture and construction firms, erected factories ahead of demonstrated demand, and dismissed much of the construction trades’ accumulated expertise. A 2019 investigation found roughly a dozen unfinished projects; when asked, Katerra could identify one it had delivered on time [18]. Zume raised $375 million from SoftBank at a $2.25 billion valuation to bake pizzas in robotic trucks, then discovered that turning vehicles sent cheese to one side [19]. It pivoted to compostable packaging. Fast, a checkout-button company, paid the Chainsmokers roughly one million dollars for a conference performance in a year when its revenue was about six hundred thousand dollars [20]. Each expense was defensible as promotion or expansion in isolation; together they show how readily capital arrived ahead of operating evidence.
The founders also kept personal ledgers. When Secret raised twenty-five million dollars at a valuation above one hundred million, ten weeks after usage had peaked, its two founders sold three million dollars of stock each without telling employees. Some employees learned from anonymous posts on Secret [21]. David Byttow bought a red Ferrari. The company closed within ten months of the round [21]. Color founder Bill Nguyen supplied a more expansive claim after declining a reported two-hundred-million-dollar offer from Google: “IBM didn’t survive the PC, none of the PC guys survived the web, and I don’t think any of the web guys will survive the post-PC world” [22]. IBM plainly survived, as did leading firms from the PC era. The false history was part of the pitch: each technological transition was described as total extinction, with Color exempted from the rule.
· · ·
The founders remained responsible for the companies they ran; investor appetite does not excuse self-dealing, deception, or contempt for employees. It does help explain why the conduct persisted. Son instructed Neumann to be crazier and increased the plan; Sequoia recorded “I LOVE THIS FOUNDER” while FTX lacked a chief financial officer and effective independent oversight; Quibi’s syndicate underwrote an executive record before testing the format. Jets, parties, burned banknotes, Marine speeches, lifetime pensions, and secondary sales all signaled that ordinary limits did not apply. Capital did not require every extravagance, but it repeatedly rewarded the claim of exemption that the extravagance advertised.
That claim survived until an employee, filing, customer, or reporter supplied a fact the performance could not absorb.
First cracks, and the institutions that failed to test them
On the nineteenth of April, 2017, two Bloomberg News reporters, Ellen Huet and Olivia Zaleski, performed an unusually economical act of due diligence. Juicero, a San Francisco company, had raised about a hundred and twenty million dollars from Kleiner Perkins, Google’s venture arm, Campbell Soup, and others to build a Wi-Fi-connected machine that pressed proprietary packets of chopped produce into juice. At the time, the machine cost $399, down from its $699 launch price. It exerted four tons of pressing force (“enough to lift two Teslas,” its founder liked to say), contained four hundred custom parts, and refused to press a packet until it had consulted the internet. The reporters picked up a packet and squeezed it with their hands. In their test, the juice emerged as fast as, and sometimes faster than, the machine could manage it [1].
The experiment required no laboratory, subpoena, or forensic accountant. Two Juicero investors told Bloomberg they learned that the packets could be hand-squeezed only after investing. Hardware investor Ben Einstein later tore down the machine and ranked it among the most complex devices he had disassembled: machined aerospace-grade parts, a camera to read packet codes, and engineering far beyond the task [2]. Abundant capital had financed technical complexity before anyone established that the press itself was necessary.
The warnings took different forms: a resignation, a prospectus, an analytics chart, a regulator’s file, the squeeze of a hand. Some were public; others reached boards or executives first. Responses differed too. Theranos threatened employees and a newspaper, while Facebook changed a parameter and exposed companies dependent on it. Across these cases, people with authority often delayed independent inquiry, leaving outsiders with less financial attachment to test claims that insiders had accepted.
· · ·
Some warnings came from inside the building.
At Theranos, lab associate Erika Cheung watched Edison machines fail quality controls while patient samples were run regardless. Tyler Shultz emailed Elizabeth Holmes about failed quality control and manipulated proficiency testing. The company’s president rebuffed him, and lawyers from Boies Schiller confronted him at his grandfather George Shultz’s house with a temporary restraining order and accusations of trade-secret theft. His family spent more than four hundred thousand dollars on legal defense [3]. Cheung filed a complaint with federal regulators that helped trigger a surprise inspection of the Newark laboratory; after leaving Theranos, she was followed and served papers by an unidentified man [4]. When John Carreyrou of the Wall Street Journal assembled their testimony in 2015, Theranos sent David Boies, whose firm had accepted part of its fees in Theranos shares, to press the newspaper for five hours [5]. On October 15, 2015, Holmes answered the resulting story on television: “This is what happens when you work to change things.” The patients whose samples had been run on failing machines had no part in that exchange.
The board did not commission an independent validation of the technology. Major investors, including Rupert Murdoch and the Walton and DeVos families, invested without audited financial statements [6]. Simple technical checks were available. Google Ventures sent a staffer to Walgreens, where the advertised finger-stick required a venous draw; a medical fund ended a 2004 meeting after Holmes could not answer basic questions [7]. Several investors with relevant experience passed, while some of the most celebrated investors did not require the same verification.
At Clinkle, the warnings came from senior employees the company had hired to reassure investors. Barry McCarthy, the chief financial officer who had taken Netflix public, arrived in October 2013 and resigned in under five months. “They’re not nearly as close to scaling the business as I thought they were when I came in the door,” he said [8]. Chi-Chao Chang, a Yahoo veteran announced as vice president of engineering, saw the product plans on his first day and resigned after twenty-four hours [9]. By November 2013, five months after the record seed announcement, Business Insider’s Alyson Shontell had counted thirty-one departures [10]. Aerolink, the ultrasonic payment system on which Clinkle had raised its money, was shelved that fall after ambient noise, including a coffee-shop blender, interfered with it in field tests [11].
Clinkle’s more than eighteen investors, including Accel, Andreessen Horowitz, Peter Thiel, and Richard Branson, held convertible notes without a board seat [12]. That structure limited formal control but did not make action impossible: investors could seek information, coordinate, withhold further capital, or press for governance changes. The public record instead shows little collective response as executives departed. When Forbes asked Duplan about reports that staff were leaving, he replied, “a fact check would be worthwhile.”
· · ·
Other warnings arrived in documents that institutions were required to read.
The Pets.com prospectus of February 2000 disclosed that, in the period through September 1999, the company had spent $11.8 million on advertising to generate about $0.6 million in sales, with those sales covering roughly a third of merchandise cost [13]. Those were early-period figures rather than its complete record at the IPO, but the direction of the economics was clear: every additional bag of dog food sold below cost and shipped free enlarged the loss. Merrill Lynch underwrote the offering, and its analysts maintained their enthusiasm until the stock had fallen about ninety percent [14]. Five weeks after the IPO, Barron’s published Jack Willoughby’s “Burning Up,” which calculated that dozens of internet companies were only quarters from exhausting their cash. “That unpleasant popping sound,” Willoughby wrote, “is likely to be heard before the end of this year” [15]. A succession of failures made the warning well founded.
Webvan’s warning was arithmetic of the same nakedness. Its Oakland warehouse, the showpiece journalists toured like a lunar facility, needed roughly three thousand orders a day to break even and was averaging about 2,160, some twenty-seven percent of capacity. The company was meanwhile committed through a billion-dollar agreement with Bechtel to building twenty-five more warehouses [16]. In 2000 it booked $178.5 million in sales against $525.4 million in expenses. Randall Stross, embedded in Benchmark’s partner meetings, recorded partner Bob Kagle warning David Beirne, the colleague who had staked the firm on marquee hires, that salesmen are more likely to be sold [17]. The firm heard Kagle’s warning and invested.
When WeWork filed its S-1 on August 14, 2019, one document put the company’s risks before a public audience. It reported a 2018 loss of $1.9 billion on $1.8 billion of revenue, $47.2 billion in lease obligations, and members able to cancel on much shorter terms. It also disclosed founder transactions, including the $5.9 million trademark payment [18]. Much of this was already known to the board, banks, and SoftBank. Bond investors had seen “community-adjusted EBITDA” in April 2018 [19]. Masayoshi Son had abandoned a proposed $16 billion buyout on Christmas Eve 2018, then invested another $2 billion at a $47 billion mark whose derivation people close to the deal could not clearly explain [20]. After the filing, Scott Galloway wrote that anyone endorsing the company above ten billion dollars was “lying, stupid, or both” [21]. Thirty-three days later, WeWork postponed the offering. Public scrutiny had changed the price assigned to facts insiders had possessed for months or years.
· · ·
A platform change could expose dependence as clearly as a resignation or filing.
Between mid-May and the end of June 2012, Facebook changed its news feed to suppress pre-checked invites, quiz-bait wall posts, and auto-shared video views. BranchOut, which had shown 13.9 million monthly active users around its April round, lost 7.4 million of them in thirty-seven days [22]. By August founder Rick Marini acknowledged that he had built on “shifting sands” [23]. Viddy confirmed its round on May 11; Facebook announced its auto-sharing restrictions on May 31; daily active users fell from roughly five million to one million within a month [24]. Twenty days separated the funding confirmation from the platform change. Investors had priced audiences that the companies did not control.
A markdown carried the same warning in financial form. Fidelity, which had led Zenefits’ $500 million round at $4.5 billion, cut its carrying value by nearly half in late 2015 [25]. BlackRock marked its Jawbone debt down sixty-nine percent within eight months of extending it; Sequoia and Khosla declined to join the next round, which a sovereign wealth fund completed [26]. Those decisions expressed doubt without a public announcement. Employees, who might have organized their lives differently with that information, often learned months later from journalists — or abruptly from payroll.
At Zenefits, BuzzFeed News reporter William Alden established in autumn 2015 that unlicensed employees were selling insurance; in Washington State, they handled eighty-three percent of the sampled deals. Records later showed that chief executive Parker Conrad had written a browser macro that allowed brokers to evade required training and shared it as a productivity tool [27]. An internal memo from June 2015 had already asked employees to stop leaving cigarettes, beer cups, and used condoms in the stairwells [28]. The licensing violations were visible in company records and state databases before an outside reporter made them public.
Customers could see MoviePass’s warning. On July 26, 2018, the $9.95 unlimited-cinema service ran out of money, went dark nationwide during the opening weekend of a Tom Cruise film, and resumed after a five-million-dollar emergency loan [29]. Management then covertly throttled heavy users by invalidating passwords and demanding photographs of ticket stubs, conduct the Federal Trade Commission later alleged in its complaint [30]. At Quibi, the warning preceded launch. Meg Whitman privately threatened to resign in 2018 over Jeffrey Katzenberg’s treatment of her. Their truce left product decisions to negotiation between a chief executive and a founder-chairman with blurred authority [31]. The governance problem was known inside the company nearly two years before launch.
· · ·
These warnings were not identical. Theranos retaliated against employees; Clinkle’s fragmented noteholders lacked a board representative; Pets.com and WeWork disclosed damaging figures in filings; Facebook altered the distribution on which BranchOut and Viddy depended. Some facts were hidden from outsiders, but the relevant board members, executives, or investors often had enough information to investigate sooner. Describing the later collapse as “sudden” erases the people who raised concerns and the institutions that postponed a response.
The incentives did not require a conspiracy. Venture firms collected management fees on committed capital; carried interest depended on eventual gains, while new fundraising also depended on attractive interim marks and access to sought-after deals. Founders needed the next round, and board members risked being called “not founder-friendly” when they pressed for controls. None of those incentives made scrutiny impossible, and each actor retained a duty to exercise judgment. They did make an outside test unusually valuable. Reporters including Carreyrou, Shontell, Alden, Willoughby, Huet, and Zaleski had less reason to preserve the private valuation, while employees, regulators, counterparties, and skeptical investors provided other checks when their evidence reached daylight.
A warning did not make collapse inevitable. Refusing to investigate it made the eventual failure harder and more expensive to arrest.
Downfall
On the morning of November 7, 2000, while the rest of the United States was busy failing to elect a president, a company in San Francisco announced that it would stop taking orders for dog food at eleven o’clock Pacific time two days hence. The stock, which had been offered to the public at eleven dollars in February, closed at nineteen cents. The wind-down would eliminate 255 of roughly 320 jobs. The chief executive, Julie Wainwright, issued the statement that the age required of her — “It is well known that this is a very, very difficult environment for business-to-consumer Internet companies” — and declined further comment to Forbes, which had already filed her company under its recurring feature, “Disaster of the Day” [1]. From the February 11 initial public offering to the shutdown announcement, 270 days had elapsed [2].
Pets.com sets the pace: 270 days, IPO to wind-down, with no fraud, raid, or palace coup. The prospectus reported $5.8 million in net sales through December 1999 against $13.4 million in cost of sales; the company recovered about forty-three cents for each dollar the merchandise cost it [1]. Merrill Lynch’s bankers could read the same arithmetic while collecting their fees. In many of the collapses that follow, the decisive facts were likewise visible before the end. What changed was that a public market, a reporter, a counterparty, a court, or a federal agency stopped extending the credit of disbelief.
Once money or permission is withdrawn, the announcements acquire a dialect of their own. A company does not run out of money; it encounters “a very, very difficult environment.” It does not fire thirty of its fifty people; it makes “some strategic changes,” in the words of Yik Yak’s Tyler Droll [3]. Zirtual told four hundred employees by overnight email that it had “paused all operations,” a pause that included their pay and health coverage [4]. The same language that had enlarged an ordinary product now reduced an irreversible firing. At the end, the description was the last thing still shipping.
· · ·
The pivot is where denial most often acquires a product roadmap.
Clinkle offers the clearest example. It had raised thirty-odd million dollars on an ultrasonic payment technology that ambient coffee-shop noise could defeat and that was shelved within months of the round [5]. In September 2014, three years and one Apple Pay announcement into its existence, it launched a prepaid Visa debit card for college students. A lottery gimmick called Treats let users send a friend every seventh card swipe; the recipient could then “spin” for a refund of up to twenty-five dollars [6]. TechCrunch used the headline “Mobile Wallet Laughingstock Clinkle Finally Launches”; Re/code filed “Six Months Later, Clinkle Is What We Thought It Was” [6]. The revolution that was to change how people trade was paying beta users ten dollars to swipe five times and had installed at Berkeley a vending machine that dispensed twenty-dollar bills [7].
The end came over a suspected conversation with a banker. In May 2015, Lucas Duplan fired his chief financial officer and vice president of engineering after suspecting they had spoken with an intermediary about an acquisition. He had not told the staff that Apple had sent a dozen people for a ninety-minute inspection, talks he would neither “confirm or deny.” Seven employees, most of what remained of a staff that had once numbered seventy, resigned the same day [8].
By December 2015, Duplan was pitching a “Treats SDK” in emails that identified him only as “a fellow founder / ceo (graduated Stanford undergrad 2 years ago, raised $30m),” omitting the company’s name [9]. In January 2016, Forbes reported that convertible-note holders had asked for repayment. “When you have convertible debt investors asking for their money back,” lawyer Craig Jacoby said, “things have gone really wrong” [5]. By May the website had ceased to exist. No announcement was made, and no final public accounting of the funds has been located.
Clinkle’s pivot was denial in its comic form; failed social apps repeated the pattern in bulk. BranchOut, having lost 7.4 million monthly users in thirty-seven days when Facebook adjusted the machinery BranchOut had mistaken for its own product, relaunched as a standalone site, asking users who had never formed a habit to form it now at a new address; when that failed it pivoted again, into workplace chat, directly into the ascending path of Slack, and was gone within the year [10]. Secret, bleeding users, rebuilt itself in December 2014 as what TechCrunch called “a shameless clone of Yik Yak,” a redesign so estranged from the founding product that the co-founder who owned its identity quit within six weeks [11]. Yik Yak itself, under the gravity of a Sequoia valuation that demanded a mainstream future its geofenced campus product could not deliver, abolished its own anonymity in August 2016. In the words of a former employee, “people lost their ever-loving minds” and left en masse; the November mea culpa, “We messed up,” restored the anonymity but not the users [3]. These pivots differed, but each arrived after the original feedback had become financially unacceptable. The product was asked to repair the financing story.
· · ·
The mechanics of the fall are clearest where the documentation is richest: the six weeks in which WeWork’s offering came apart.
On August 14, 2019, the company filed an S-1 disclosing, among its energies and spiritualities, a $1.9 billion annual loss, $47.2 billion in lease obligations, and 169 appearances of the founder’s first name [12]. Thirty-three days later the offering was postponed as the marketed valuation fell from forty-seven billion dollars toward ten. On September 18, the Wall Street Journal’s Eliot Brown published the profile — the Gulfstream, the marijuana in the cereal box, the ambitions to live forever and become president of the world — that gave the board what it had lacked for nine years: a reason visible from outside [13].
Then came the weekend. JPMorgan’s Mary Erdoes told Adam Neumann that investors found an IPO with him at the helm “untenable.” Jamie Dimon — investor, underwriter, lender, and the man Neumann called “my personal banker” — “added to the pressure on Sunday,” advising him the offering was dead if he stayed. At dinner that night, three of his own directors urged him to go [14]. The board that had accepted the supervoting shares and self-dealt leases acted only after the bankers put a price on inaction. On Tuesday, September 24, Neumann resigned as chief executive, citing the “significant distraction” of scrutiny into facts disclosed in his own prospectus. On September 30 the S-1 was withdrawn. The company, burning more than three billion dollars a year, was weeks from insolvency; on October 22 SoftBank administered a rescue at an eight-billion-dollar valuation, a markdown of thirty-nine billion dollars in nine months, attached to an exit package for the founder reported at $1.7 billion. On November 21, 2,400 employees were laid off with ordinary severance [12]. The sequence is unusually complete: filing, reporting, banker intervention, founder removal, rescue, layoffs.
WeWork took nearly six weeks from filing to the founder’s fall; Quibi announced its shutdown on its 199th day of operation; FTX took nine days from a CoinDesk article to Chapter 11. A company dependent on continued confidence can lose access to cash far faster than its operations deteriorate. Financing can disappear between one meeting and the next.
Quibi shows that executive experience can improve the staging without saving the company. Having raised $1.75 billion before launch and attracted, by mid-2020, a paying audience that would embarrass a regional cable channel, Jeffrey Katzenberg and Meg Whitman shopped the company and found little to sell: Quibi had licensed rather than owned its content [15]. They shut it down expecting roughly $350 million to remain for shareholders after wind-down costs, rather than spending the balance defending the thesis [15]. On October 21, 2020, announcing to employees that their company would cease to exist, Katzenberg offered a source of consolation: the song “Get Back Up Again,” from the animated film Trolls [16]. Several hundred adults were losing their jobs during a pandemic; their billionaire chairman prescribed a cartoon anthem. The open letter published that day offered an unresolved either/or: the idea “wasn’t strong enough,” or the timing was wrong [15].
· · ·
The character of a shutdown is often clearest in how the employees find out.
Pets.com executed an orderly, board-approved wind-down and gave its workers an announcement. Webvan’s two thousand learned on Monday, July 9, 2001, that operations had ceased that day. Its former chief executive, George Shaheen, had resigned three months earlier with the explanation that “a different kind of executive is needed to lead the company at this time,” taking a negotiated retirement benefit of $375,000 a year for life [17]. Bankruptcy converted the pension into an unsecured claim. Kozmo’s eleven hundred discovered the end of their company on the morning of April 11, 2001, in the most literal way available: they arrived at work, mid-shift for some, and found the warehouse doors locked [18]. Zirtual’s four hundred received their 2 a.m. email announcing the “pause” [4]. Fast’s employees, in April 2022, got roughly a week’s notice and no severance from a company that three months earlier had paid the Chainsmokers a million dollars to play a conference party, a sum exceeding its entire annual revenue [19]. The band was paid; the staff was not.
As cash dwindles, disputes move from paper valuations to things that can still be divided: the last cash, the last titles, the last version of the story. Duplan fired his CFO over a suspected conversation with a banker. Neumann sued SoftBank when the tender offer shrank; his board’s special committee sued too. The settlement ultimately included a $578.4 million stock purchase, a $105.6 million direct payment, and extension of a $430 million loan [12]. At Katerra, the board discovered near the end that financials it had been shown were, in the phrase later reporting attached to the bankruptcy record, “intentionally misstated.” When Greensill, the financing carousel propping up its working capital, collapsed in March 2021, Katerra followed within ninety days, leaving subcontractors and unfinished projects [20]. Jawbone spent its final solvent year suing Fitbit on three fronts, lost, and was cut off by its customer-service contractor, which announced that “Jawbone is not able to pay us for past services” [21]. At govWorks, a documentary crew filmed the fight consuming a friendship begun in childhood: Kaleil Isaza Tuzman called for the removal of his co-founder Tom Herman, escorted his oldest friend from the building, and changed the locks [22]. The friendship had been pitched to investors as the company’s greatest asset. It went with the company.
· · ·
In other collapses, the decisive intervention came from the state, after boards and investors had failed to establish what public investigators later found.
The FBI arrived at uBiome’s San Francisco headquarters on April 26, 2019. A federal indictment filed two years later alleged that the company had submitted more than $300 million in reimbursement claims, collected more than $35 million, used a captive network of health-care providers to approve tests, and misled insurers about retesting and clinical validity [23]. The founders went on leave within days; clinical operations were suspended within two weeks; Chapter 11 followed within five months and Chapter 7 weeks later. Theranos was the larger case. Its board of statesmen did not stop the blood-testing. Inspectors from the Centers for Medicare and Medicaid Services, acting after a complaint from Erika Cheung, a lab associate in her twenties, found “immediate jeopardy to patient health and safety” and in July 2016 banned Elizabeth Holmes from owning or operating a laboratory for at least two years. The company later voided or corrected tens of thousands of patient test results, any of which might already have informed a medical decision, and dissolved in September 2018 after the indictments [24]. MoviePass, having discovered that its ten-dollar all-you-can-watch product was destroying it, secretly prevented heavy users from using it by invalidating passwords and demanding ticket photographs. The Federal Trade Commission later alleged that these tactics were deceptive. MoviePass chief executive Mitch Lowe pleaded guilty to conspiracy to commit securities fraud; Helios and Matheson chief executive Theodore Farnsworth pleaded guilty to securities fraud and conspiracy [25].
FTX compressed the procedure into nine days and performed it before a global audience. On November 2, 2022, CoinDesk’s Ian Allison published a leaked balance sheet showing that the hedge fund attached to the exchange was capitalized largely with the exchange’s own token [26]. On November 6, a rival announced it would sell; a run began; on November 8 withdrawals halted; and on November 9 Binance, having signed a letter of intent to acquire FTX, walked away after one day of due diligence. Sequoia, Paradigm, Temasek, Tiger, SoftBank, BlackRock, and the Ontario Teachers’ Pension Plan had already invested [26]. Chapter 11 came on November 11. The man appointed to sift the wreckage, John J. Ray III, had liquidated Enron, and his first-day declaration carries the weight of a connoisseur’s judgment: “Never in my career have I seen such a complete failure of corporate controls and such a complete absence of trustworthy financial information as occurred here” [27]. To Congress he offered a shorter account: “This is really old-fashioned embezzlement.” Nine days after the leaked balance sheet, the company was in bankruptcy.
· · ·
By the time the last employee receives the email, a premise has been financed past its tests, a pivot has failed to rescue the story, and bankers, counterparties, or the state have asserted control. The jobs are gone; cash, losses, and reputation remain to be distributed.
The companies here did not fail for one reason. In case after case, however, abundant capital pushed choices in the same direction: spending to the plan, scaling to the deck, and postponing the test of whether the business could become real. The fall was not scripted. The conditions for it were financed.
Aftermath, and who got paid
On the fifteenth of August, 2022, Marc Andreessen announced his firm’s investment in Flow, Adam Neumann’s residential real-estate startup. The check was reported at three hundred and fifty million dollars, then the largest in Andreessen Horowitz’s history. Flow had not begun operations, but it was expected to manage more than three thousand apartment units Neumann had acquired in Miami, Fort Lauderdale, Atlanta, and Nashville [1]. Andreessen praised him in the language of the parish newsletter: “We love seeing repeat-founders build on past successes by growing from lessons learned. For Adam, the successes and lessons are plenty.” The date matters. WeWork, the vehicle of those successes and lessons, would not file for Chapter 11 for another fifteen months. By then a company once priced at forty-seven billion dollars would be worth on the public market about forty-four and a half million, roughly one-thousandth of its peak [2]. Andreessen Horowitz had funded the resurrection before the death. Lazarus at least had the decency to be buried first.
The aftermath is described in two vocabularies. Survivor accounts call it “lessons learned”; an accountant would call it a distribution of losses. Destroying capital by bad judgment was rarely a crime. Legal consequence arrived when prosecutors could prove a recognized offense, often fraud against investors or customers. Professional consequence followed no such clear rule, and the losses still reached people who had not chosen the risks.
· · ·
The “$1.7 billion exit package” was the October 2019 headline, but it described a structure rather than a payment: the right to sell up to $970 million into a tender offer, a $185 million consulting fee, and a half-billion-dollar credit line. Litigation changed the terms. A later SEC filing records the settlement terms: in February 2021, SoftBank bought $578.4 million of stock from a Neumann-affiliated investment vehicle at $19.19 a share, and WeWork booked a $428.3 million expense for the premium. SoftBank also paid Neumann $105.6 million directly and extended his $430 million loan for five years [3].
Four weeks after the original package was announced, some 2,400 employees, about a fifth of the company, were laid off with ordinary severance [4]. Neumann’s former chief of staff, Medina Bardhi, had by then filed a complaint with the EEOC alleging pregnancy discrimination and describing marijuana smoke in the sealed cabin of a chartered jet [5]. She received a filing number. He received, in time, Flow, which raised again in April 2025 at a valuation of two and a half billion dollars with Andreessen Horowitz once more at the table [6].
National Public Radio compared the check with data showing that startups with Black founders across the United States had raised $324 million in the second quarter of 2022, less than the single commitment to Flow [7]. The comparison does not prove that those dollars would otherwise have gone to those founders. It does show the scale at which a familiar, well-connected founder could be refinanced after failure.
Masayoshi Son, the man who had ordered Neumann to be “crazier,” performed his contrition on earnings calls: “My judgment in investment was not right in many ways,” then, by May 2020, “I was foolish,” and finally, in 2023, “a stain on my life” [8]. He continued to manage one of the largest pools of capital on earth. Bloomberg’s tally of the damage ran to $11.5 billion in equity and a further $2.2 billion in debt exposure [2]. Surgery and aviation have licenses or commands that can be withdrawn; venture capital has no equivalent credential. Son’s stain remained a metaphor. The employees who lost their jobs did not appear in it.
· · ·
Jeffrey Katzenberg assigned Quibi’s loss to history. In a May 2020 interview with the New York Times, he said: “I attribute everything that has gone wrong to coronavirus. Everything” [9]. The explanation was awkwardly timed: households were locked indoors and streaming video at record volume. By October he called the answer “flippant” [10]. Quibi’s principals nevertheless made one defensible decision in the shutdown: they stopped and expected roughly $350 million to remain for shareholders after wind-down costs rather than spend the balance defending the original wager. Katzenberg returned to his venture firm; Meg Whitman was later appointed United States Ambassador to Kenya; Roku bought distribution rights to most of the library, reportedly for under $100 million [11].
External explanations recur in the shutdown statements: the pandemic, the capital markets, the platform, the “environment.” Pets.com used the last of these; Webvan’s chairman blamed the capital markets; Katerra cited the pandemic and Greensill [12]. Those conditions mattered, but the margin, retention, and utilization problems in these cases were older. The first draft of a startup obituary is usually written by the company, and the world is given a speaking part.
· · ·
Clinkle never announced its own death. The website stopped resolving around May 2016. In January of that year Forbes reported that holders of convertible notes had been asking for repayment since the previous summer. “When you have convertible debt investors asking for their money back,” Cooley partner Craig Jacoby observed, “things have gone really wrong” [13]. Some ten to fifteen million dollars reportedly remained in the accounts in 2015, but no public accounting established what was ultimately returned. Lucas Duplan resurfaced in 2018 attached to a blockchain venture and an aspirant venture fund. In 2023 Forbes, which had placed him on its 30 Under 30 list during Clinkle’s collapse, transferred him to its Hall of Shame [14]. A magazine listicle was the visible consequence.
The contrast with the funding announcements is sharp. Viddy was “positioned to be the breakaway leader”; Yik Yak had “tapped our desire to connect authentically”; Secret offered “a place where we can truly unburden ourselves.” I found no comparable shutdown statement in the public record from NEA, Sequoia, Index, or Goldman for those companies [15]. Google Ventures’ Bill Maris was a conspicuous exception after Secret’s founders had taken three million dollars each from the Series B. Some employees learned of the cash-out from anonymous posts on Secret itself. “It’s like a bank heist,” Maris told the New York Times. “I think they should return all the money. Some went to taxes, some went to a red Ferrari, which is apparently now sold” [16]. Investors eventually recovered about ten million of the thirty-five million they had put in; the public record does not show the founders returning any of their six million [17].
· · ·
Legal consequence did arrive in cases where prosecutors proved specific offenses. Elizabeth Holmes is in a federal prison camp in Texas, sentenced to eleven and a quarter years and jointly liable with Sunny Balwani for $452 million in restitution; she had petitioned for clemency by early 2026 [18]. Sam Bankman-Fried is serving twenty-five years, his conviction affirmed on appeal in June 2026 [19]. MoviePass chief executive Mitchell Lowe pleaded guilty to conspiracy to commit securities fraud in 2024; Helios and Matheson chief executive Theodore Farnsworth pleaded guilty to securities fraud and conspiracy in 2025 [20]. Kaleil Isaza Tuzman of govWorks was convicted in 2017 for securities fraud at his next company and received probation after spending ten months in a Colombian prison [21].
Set beside those prosecutions the companies whose products or strategies simply failed. Juicero raised one hundred and twenty million dollars for a wifi-connected machine whose juice-extraction function Bloomberg reproduced by squeezing the packet by hand [22]. Clinkle, Quibi, Webvan, Pets.com, Kozmo, Color, and BranchOut produced losses without criminal cases arising from the failures themselves. That distinction matters: a system that treated an unsuccessful experiment as a crime would stop experimentation. The harder question is why preventable governance failures so often carried little professional consequence for the people with authority while employees, customers, and creditors absorbed the loss. Holmes’s split verdict — conviction on investor counts, acquittal on patient counts — creates an uncomfortable contrast, though no single jury verdict can serve as a general theory of whose injuries the law recognizes [18].
The investors’ reckonings were mostly private. Sequoia deleted the commissioned profile of Bankman-Fried that had celebrated his video-game pitch meeting, wrote $213.5 million to zero, and issued a letter to limited partners opening, “We are in the business of taking risk” [23]. It later apologized on a call. Temasek wrote off $275 million and cut the compensation of its investment team and senior management after an internal review [24]. Among the FTX investors named here, that was an unusually concrete institutional consequence.
The banks had refined loss-shifting into a fee business a generation earlier. eToys’ creditors spent eleven years suing Goldman Sachs, alleging that the bank had underpriced the IPO so favored clients could enjoy the 283 percent first-day gain and return a share of their profits through other business. The case settled for $7.5 million after New York’s highest court held that an underwriter could owe its client a fiduciary duty concerning the pricing conflict [25]. Employees who had taken eToys equity received nothing from the bankruptcy.
· · ·
Employees experienced the accounting directly. Fast, the one-click checkout company that raised $120 million against six hundred thousand dollars of annual revenue, evaporated in roughly a week in April 2022. Its people received one week’s pay and no severance; its founder moved on to another venture [26]. Zirtual’s four hundred assistants learned by a 2 a.m. email that the company was “pausing all operations,” meaning that their paychecks and health coverage had ceased while they slept [27]. WeWork’s 2,400 layoffs arrived weeks after the founder’s nine-figure arrangements. Zirtual’s founder, Maren Kate Donovan, later published an unusually direct account of the missing CFO, the two-person board, and the payroll arithmetic, blaming neither virus nor market nor platform [27]. Candor did not restore the jobs, but it left a usable record.
Professional rehabilitation was common. Parker Conrad left Zenefits after disclosure of a founder-authored tool for circumventing state insurance-licensing requirements, then founded Rippling, whose valuation eventually exceeded Zenefits’ [28]. Katzenberg returned to his fund; Neumann received financing for Flow twice; Duplan explored raising a fund of his own. The failure post-mortem became a rewarded literary genre, and Tuzman spoke publicly about the lessons of govWorks during years in which prosecutors later established that he was defrauding investors in his next company [21]. Sometimes the lesson did alter practice. Michael Moritz has said that Sequoia’s memory of Webvan helped make Instacart prove one city before opening a second [29]. The learning was real; the people who paid for it were not the ones who retained it.
“Fail fast” can now be given an accounting. A founder may retain the war story and fundable experience; an investor records a portfolio loss; a general partner continues to manage the fund. The employee receives a gap in the résumé and a health-insurance cliff; the small creditor, a haircut; the pension beneficiary, a loss too diffuse to contest. The phrase is usually uttered in the first person and experienced in the third.
· · ·
Julie Wainwright offers a more complicated account of consequence. Pets.com’s FY2000 annual report, filed in 2001, records approximately $2.8 million in severance and other compensation for ten corporate officers, including $1.4 million in retention bonuses and about $1.4 million in severance. Wainwright personally received a $225,000 retention bonus and $235,000 in severance; the filing also records a separate $50,000 performance bonus [30]. These were not Neumann-sized sums, but they rule out a simple martyr story. She also paid personally. Within a day of the shutdown announcement, after her husband woke her at four in the morning, she lost the company and her marriage [30]. “My work is gone, I’m getting a divorce, and I don’t have children.” Recruiters told her no one would hire her again. She later built The RealReal, which went public in 2019 at a valuation north of $1.6 billion; in 2022 its board removed her [30]. Her story does not prove that accountability was distributed fairly. It shows how unevenly professional and personal consequences could land.
Her sock puppet did better. The rights sold in 2002 for $125,000 to an auto-finance company, which gave the puppet a new slogan: “Everyone deserves a second chance” [31]. An original now resides in the collection of the Henry Ford museum. The era’s mascot had been liquidated and put back to work selling debt.
On the eighteenth of May, 2020, Masayoshi Son stood before his shareholders to announce the largest operating loss in the history of SoftBank — 1.36 trillion yen — and illustrated it with a cartoon. The slide showed a herd of unicorns tumbling into a chasm labeled, in his own deck’s words, the “Valley of Coronavirus,” while above them one unicorn, sprouting wings, sailed over the abyss. “I believe some of them will fly over the valley,” he explained [1]. A year of extraordinary capital destruction had been reduced to a child’s drawing of a flying horse. The presentation was covered, with light amusement, as a matter of style.
The flying unicorn was not the cause of SoftBank’s losses. It was a candid picture of the method: accept a field of wreckage because one exceptional company may clear the valley. That wager can produce great companies. It can also make the cost of being wrong nearly irrelevant to the person placing the bet, while remaining decisive for the people inside the company.
· · ·
Read together, the Valley’s most influential texts offer a remarkably consistent account of ambition.
From Thiel, via the Stanford lecture notes that became Zero to One: that competition is for losers — he published the claim under that title in the Wall Street Journal, four days before the book appeared [2] — that “monopoly is the condition of every successful business,” and that the great company is built on a secret the world cannot see. From Naval Ravikant, in the 2018 tweetstorm: seek wealth rather than money or status; use leverage; prefer code and media, which cost almost nothing to replicate [3]. From Paul Graham: “a startup is a company designed to grow fast,” “the only essential thing is growth,” and “you can use growth like a compass to make almost every decision you face” [4]. From Marc Andreessen: software is eating the world, and observers should stop “constantly questioning their valuations” long enough to understand how the companies work [5]. From Reid Hoffman: blitzscaling means “prioritizing speed over efficiency in the face of uncertainty,” with counterintuitive rules that include tolerating bad management, letting fires burn, ignoring customers, and raising too much money [6]. And from Son: a corporate vision “designed with the time span of 300 years,” a fund assembled around the Singularity — “that’s why I’m in a hurry” — and the instruction that WeWork’s founders were “not crazy enough” [7].
None of these maxims is, by itself, an instruction to deceive anyone. Thiel’s monopoly still requires a product customers value; Graham repeatedly tells founders to make something people want; Hoffman presents his rules as temporary hazards to be managed, not permanent exemptions. Yet the phrases can perform a second function in a financing culture hungry for exceptional outcomes. A secret can turn a request for evidence into a failure of imagination. Growth can postpone the profit-and-loss statement. Speed can excuse controls that would otherwise be ordinary. A large early round can delay the next external test of performance. Advice intended for rare circumstances becomes dangerous when prestige turns it into a general permission slip.
Taken as a culture rather than a shelf of books, these ideas can become a philosophy of not being stopped: an elect distinguished by conviction, a secret available to initiates, prophets whose wealth confirms their authority, and a doctrine of grace by which failure is converted from evidence into tuition. The texts do not require that result. The incentives surrounding them make that reading attractive.
· · ·
The arithmetic is the power law. Thiel states it plainly: in a successful venture fund, the best investment may equal or outperform the rest of the fund combined. Shikhar Ghosh’s study of roughly two thousand venture-backed companies found that about three in four failed to return investors’ capital and more than ninety-five percent fell short of their own projections [8]. Correlation Ventures reported that sixty-five percent of more than twenty thousand financings returned less than the original investment [9]. Horsley Bridge data spanning three decades found six percent of deals producing sixty percent of returns [10]. AngelList’s Abraham Othman modeled seed-stage returns and argued that, absent foresight, investing broadly across every credible deal available on the platform could outperform selective portfolios [11]. That last result is not a verdict on every venture market; it is evidence of how severely a few outliers can dominate a sample.
If a small number of outliers produces most of a fund’s return, many losses are expected rather than disqualifying. Management fees continue while the portfolio develops; carry depends on the rare gains. Limited partners, which can include pension funds and university endowments, absorb the financial variance. Employees absorb years, foregone wages, and option values. A company that was a workplace to hundreds of people becomes one line in fund-level performance. The arithmetic explains why investors tolerate failure. It does not decide who should bear its cost.
Founder-written post-mortems often describe the company more gently than they describe its market. In CB Insights’ widely cited 2014 review of 101 such accounts, forty-two percent selected “no market need” as a cause of death [12]. The sample was self-selected and the categories overlapped; it is not a failure rate for startups as a whole. Even so, the admission is striking. An industry whose foundational motto is make something people want repeatedly financed products whose own founders later said nobody needed. Under a power-law portfolio, that contradiction need not threaten the fund.
· · ·
Phil Rosenzweig gave the underlying analytical error its business-school name: the halo effect, by which success literature studies winners and discovers that winners share traits [13]. Blitzscaling risks turning that error into practice. Tim O’Reilly put the objection directly: “a strong case can be made that blitzscaling isn’t really a recipe for success but rather survivorship bias masquerading as a strategy.” For every PayPal that assembled the airplane after leaping from Hoffman’s cliff, he wrote, “there is a dotcom graveyard of hundreds or thousands of companies that never figured it out” [14].
Blitzscaling did not cause every failure in this book; several companies predated the term, and others failed by fraud or ordinary bad judgment. It does describe a recurring allocation pattern: answer uncertainty with more capital, weak retention with a larger growth target, and governance questions with urgency. When the method works, the winner enters the canon. When it fails, the loss enters portfolio variance. The asymmetry is the point, not proof that speed or scale is always mistaken.
Abraham Wald, with whom this book began, would have recognized the sampling error. The doctrine studies the planes that came back. It adds speed, capital, conviction, and tolerance for disorder where the survivors displayed them, while the companies destroyed by those same exposures are less likely to enter the manual. The sampling error passes from the page into practice.
· · ·
The strongest objection is that failure is the price of innovation. It is correct. A failed drug trial, an unsuccessful prototype, or a company closed after an honest market test can produce useful knowledge; a regime that punished such failure would finance less discovery. But experimental failure is different from prolonging a disproved premise, suppressing evidence, disabling governance, or transferring the downside to people who did not choose it. The objection justifies risk. It does not justify every way of taking or distributing risk. Steve Blank, who taught the Valley’s curriculum for decades, told Charles Duhigg that venture capital had become preoccupied with “optimizing their own profits and chasing the herd,” wasting money that might have financed useful innovation [15]. His complaint concerned incentives that can reward the wrong experiments for too long.
The industry nevertheless describes itself as contrarian while relying heavily on social proof. A prestigious investor validates the founder; the next investor treats the first as evidence; the rising valuation validates them both. Sequoia’s commissioned profile of Sam Bankman-Fried captured the mechanism in its most embarrassing form: while he played a video game during the pitch meeting, a partner typed, “I LOVE THIS FOUNDER” [16]. After FTX collapsed, the profile was deleted rather than corrected. Contrarianism had become a consensus about who looked contrarian.
The asymmetry is easiest to see in one fact: Andreessen Horowitz described WeWork as a “paradigm-changing global company” when it financed Adam Neumann’s next real-estate venture, fifteen months before WeWork filed for bankruptcy [17]. The institution converted failure into experience quickly enough to invest in the lesson while employees and creditors were still absorbing the loss.
No one has to design these outcomes for the incentives to make them predictable. Management fees reward the stewardship and deployment of committed capital; carry rewards the rare outlier; a later round can validate an earlier mark on paper; and the power law makes many losses compatible with a successful fund. None of those incentives requires misconduct. Together they can reduce the cost of being wrong for the decision-maker while preserving it for the company. The unicorn falls into the valley, the winged one flies over, and the man at the podium owns a share of the wings but little of the fall.
· · ·
Warnings were also issued in public and in time to matter. Knowledge alone was not the problem.
Bill Gurley of Benchmark published “On the Road to Recap” in April 2016, explaining how dirty term sheets, founder worship, and valuation-as-press-release had made the unicorn market dangerous “for all involved” [18]. The memo anticipated much of the WeWork era and changed less than its accuracy deserved. Two decades earlier Bob Kagle had warned inside Benchmark that his partners were falling for “marquee players” rather than businesses; the firm proceeded into Webvan [19]. Information existed. Acting on it could mean surrendering the deal to the fund across the road.
Theranos’s lawyers threatened John Carreyrou’s reporting; his sources were surveilled and threatened; Tyler Shultz spent four hundred thousand dollars of his family’s money on lawyers rather than recant and watched his grandfather side with the founder [20]. Susan Fowler’s restrained account documented failures Uber’s board had not exposed [21]. Effective correction came from different places in these cases: employees, reporters, counterparties, regulators, and eventually prosecutors. Boards and investors were conspicuous less by their total absence than by how often they acted after someone outside the financing relationship made inaction costly.
Whistleblowers and skeptics were often told that they “didn’t get it.” The phrase turns a diligence question into a test of temperament. Several specialists who asked how the Theranos device worked passed on the investment; prominent generalists invested [20]. Institutional investors put roughly two billion dollars into FTX without independent investor representation on its tiny, founder-dominated board [16]. In both cases, resistance to scrutiny was presented as evidence of unusual vision. Valuation then recorded the enthusiasm of those willing to accept the premise.
· · ·
The familiar bomber story is a simplification, but the sampling problem is real: surviving aircraft cannot by themselves reveal what destroyed the missing planes [22].
This book has tried to return missing companies to the sample. They are not interchangeable. Some principals committed fraud; most did not. Some products were honest experiments; others remained financed after decisive evidence had turned against them. What joins the cases is narrower: narrative and abundant capital repeatedly postponed tests that ordinary businesses must eventually pass, while the people choosing the delay were often protected from its full cost.
The church will survive this book. The seminars are already full again; the aphorisms have new numbering; somewhere on Sand Hill Road a term sheet is being prepared for a young founder who resembles the last one. That founder is owed the base rate as well as the benediction.
In the familiar parable, the armor goes where the holes are not. Silicon Valley’s manuals are written from the planes that came back, their bullet holes annotated by survivors and sold to the next aircrew. The engines are less photogenic: a product someone needs, a price above cost, a colleague who tells the truth, a board that reads the numbers, someone willing to say no while it can still matter. They receive less attention because the planes that lost them are unavailable for comment. Their absence is the evidence.