Why No One Can Look Away From Elizabeth Holmes
Summary
Elizabeth Holmes’s enduring grip begins with the same founder-performance problem that helped Theranos rise: charisma and conviction substituted for verifiable science. Caitlin Roper’s 2013 blood panel looked plausible, yet she never saw the lab or spoke to a lead scientist; John Carreyrou’s suspicion began when bold claims appeared without peer-reviewed evidence. The diligence lesson is brutally durable: “She was sort of standing in for the lead scientist.”
The documentary’s access reveals more performance, not an uncontested “real” Holmes. Gone are the black turtlenecks and severe presentation; in their place are beach scenes, children, loosened hair and telegraphed intimacy. Carreyrou sees a charming public face concealing someone “feisty,” “combative,” “ruthless” and willing to evade direct questions—most starkly when Holmes insists she has never lied.
Sympathy and accountability coexist, which is precisely why You Can See Everything works. Roper felt genuine sadness watching a mother approach an 11-year-and-three-month prison sentence, while rejecting Holmes’s attempts to blame an entourage for publicity and implicitly shift responsibility onto Sunny Balwani. The film offers emotional access without resolving its central diligence question: “What does she think is real?”
Gender amplified both Holmes’s ascent and the appetite for her punishment. Tech publications and editors wanted rare female-founder examples amid a male-dominated industry, while founder-centric investing helped make Holmes a magazine-cover archetype; afterward, the panel argues, a deceptive woman attracted a different intensity of fascination and disparagement. Yet Carreyrou adds a crucial correction: 11 years after his exposé, Holmes herself renewed the cycle by inviting filmmakers into her home roughly a month before prison.
Theranos remains relevant because Silicon Valley still cannot articulate a clean boundary between visionary exaggeration and illegal deception. Erin Griffith argues Holmes followed the standard “fake it till you make it” playbook but crossed into lying; Carreyrou calls the line between Holmes and Elon Musk “extremely fine.” That ambiguity is investor-relevant risk: founder storytelling may be necessary, but unsupported product claims cannot be underwritten as merely ambitious road maps.
Character.ai’s bankruptcy shows the residue left by 2024’s AI talent wars. After Google brought back founders Noam Shazeer and Daniel de Freitas plus roughly 20% of the staff, Griffith says the remaining company became a founderless “nub”; it later sought bankruptcy protection amid lawsuits and unprofitability despite about 12 million active users, while selling assets to Walt Disney. The episode’s broader warning is that acqui-hires can extract human capital while leaving liabilities, compute costs and diminished strategic value behind.
The week’s other AI stories expose a market selling both extraordinary strategic importance and mundane convenience. An OpenAI employee was accused of planting a recording device, underscoring why concentrated model IP attracts espionage interest; meanwhile Apple’s HomeView productizes the “kitchen iPad” as an always-on household manager. The hosts’ skepticism is consistent across both: OpenAI turned discovery of the device into model marketing—“We used our models”—while Apple’s polished ad may be inventing problems for another screen to solve.
Deep dive
1. Holmes first won attention where scientific verification was absent
Roper found Holmes in late 2013 while seeking female founders for Wired. Theranos confined her to one room containing Holmes and a phlebotomist; the conspicuous voice, black clothing, red lipstick and “big blue-eyed stare” already felt purposeful, but the healthcare-cost and patient-transparency pitch seemed far more consequential than another delivery startup.
Roper requested the same blood panel she had recently received elsewhere. Theranos drew a “teeny, tiny vial,” and the results looked very close to her comparison panel—but she could not inspect the lab, meet a lead scientist or validate the claimed science. In retrospect, she still wonders where and how the sample was tested.
What Roper did perceive was conviction. Holmes seemed to stand in for the missing scientist, while her controlled nonanswers resembled other founders withholding anything unhelpful to an IPO or funding round. Roper’s later regret was not personal animus; it was that Wired wrote early despite being unable to validate the company’s scientific claims.
Carreyrou’s skepticism began with what he recalls as Ken Auletta’s mid-December 2014 New Yorker profile: Theranos made bold claims without peer-reviewed studies and operated under conspicuous secrecy. A pathologist, although “three times removed” from direct information, independently questioned the technology’s feasibility; Holmes’s videos then left Carreyrou feeling that “some play acting” was occurring.
2. The new access exposes another carefully constructed Holmes
Carreyrou never interviewed Holmes despite repeated requests in late 2015 and again while preparing his book in 2016. He therefore watched the extraordinary access Fielder and Oppenheim got with some envy: Fielder spent roughly a month with Holmes, Billy Evans and their children immediately before prison and could ask questions Carreyrou never could.
His assessment contains two Holmeses. One can be “very pleasant” and “extremely charming”; the other is feisty, combative, ruthless and a liar. The film mostly shows the first, with flashes of the second whenever Fielder presses her to acknowledge wrongdoing and she snaps, “I did not.”
Roper reads Holmes’s changed appearance as another possible strategy. The turtlenecks and severe founder costume disappear; light clothing, loose hair, beach frolicking and constant proximity to her children telegraph “ease and intimacy,” inviting viewers to believe they have entered her authentic private self.
Repeated conversational breakdowns reminded Griffith of prosecutors asking Holmes the same question in slightly different ways at trial. Carreyrou rejects the idea that she simply fails to understand: “She’s not answering the question. She’s evading the question.” The documentary turns that evasion into evidence by allowing its loops to remain visible.
3. The film permits empathy without surrendering accountability
Carreyrou nearly jumped from his chair when Holmes cried before being photographed for a digital replica and said her former entourage had pushed her into the spotlight. He calls that “a brazen lie”: she embraced publicity, planned coverage and, according to trial texts, enjoyed it. To him, the tears manufacture victimhood while transferring responsibility.
Griffith preserves an alternative possibility: even if Holmes once wanted fame, its eventual consequences might have traumatized her. Roper likewise felt human sadness watching a mother face 11 years and three months away from young children; Carreyrou noted Holmes was apparently expected to move to a halfway house the following year, which he characterized as making it “only 5” years.
That sympathy does not negate calculation. When asked whether she had ever lied in her life, Holmes said no, prompting Fielder’s incredulous “Liz, really?” Carreyrou thinks the filmmakers gave her “enough rope to hang herself”: the straight-faced denial looks absurd rather than rehabilitative.
Holmes also implicitly redirects blame toward former partner and Theranos second-in-command Sunny Balwani. Carreyrou sees a persistent unwillingness to accept responsibility—blaming an entourage for publicity and Balwani for Theranos—while simultaneously asking viewers to feel for her.
4. Gender powered the rise and reshaped the backlash
Carreyrou attributes the fascination partly to “the psychology of deceit, especially when it’s a woman doing the deceiving.” Theranos was also more than a conventional investor fraud: blood testing introduced a public-health dimension and the possibility that patients’ lives were put in jeopardy.
Roper argues American culture routinely catches men lying without sustaining the same revulsion. She does not dispute that Holmes deserves condemnation, but sees gender influencing “who we disparage and continue to disparage over time.”
Griffith connects that backlash to the earlier promotional cycle. In the mid-2010s, a tech press searching for a female founder of a billion-dollar company found Holmes and put her on business-magazine covers; the later “girl boss” era repeated the pattern of disproportionate celebration followed by disproportionate criticism.
Yet attention is not merely imposed on Holmes. Eleven years after Carreyrou’s first Wall Street Journal story, she invited filmmakers inside 34 days before prison. “She is the one who is still asking for the public attention,” he says, making self-renewal part of the explanation for her cultural persistence.
5. Holmes repeats grandiose partnership patterns with Billy Evans
Roper sees striking continuity between Holmes’s relationships with Balwani and Evans: extreme devotion, shared business ambition and bombastic language that makes ordinary intimacy sound mythic. In Max’s reading, Holmes presents romance as perfect parity: she and her partner are “aligned and synced,” “2 twin souls,” even “dolphins” leaping toward one another, and they repeatedly call their relationship a “monument to love.”
The uneasiest scene arrives when Holmes and Evans excite each other over “Theranos 2.0” as her parents visibly steel themselves—“Here we go again,” in Griffith’s reading. Asked whether another business-romantic partnership is wise, Christian Holmes answers, “Good question. I trust Billy.”
That assurance is immediately undercut when Evans enters and shuts the conversation down: “No more talking.” The film’s title promises You Can See Everything, yet Griffith notes that Evans repeatedly restricts access, reproducing the tension between proclaimed transparency and active control.
Carreyrou considers both partners damaging influences: Balwani was a “terrible influence,” and Evans appears narcissistic, arrogant, controlling, manipulative and dishonest. Still, he resists reducing everything to performance: “Probably the most genuine thing in the movie is their love,” and he believes Holmes also genuinely loved Balwani.
6. Theranos still sits on Silicon Valley’s unresolved fault line
Griffith’s trial-era argument was that Holmes embodied Silicon Valley’s “fake it till you make it” ethos. Founders routinely exaggerate distant possibilities to recruit investors and employees; when that crosses into lying and deception, it becomes illegal, but founders and investors remain confused about where the line sits.
Valley investors strongly rejected the comparison—“She is not part of Silicon Valley”—yet Griffith calls her a product of its culture. She points to Elon Musk’s repeated undelivered promises; Carreyrou goes further, calling the distinction between Musk and Holmes “an extremely fine line” that an outsider could reasonably struggle to reconcile.
Roper remembers touring a Tesla factory and seeing impressive robotic arms before there was a car, followed by continued waiting for the car. Her conclusion is not that every founder commits fraud, but that “every single founder is trafficking across that line.”
“Theranos 2.0” therefore lands as dark comedy. If the original machine really performed hundreds of tests perfectly, Carreyrou asks sarcastically, why would a second iteration, lasers, drawings or another prototype be necessary? The new pitch collides with the film’s claim that the first technology worked perfectly.
7. The filmmakers make manipulation itself the viewing experience
Carreyrou initially found Fielder’s politeness frustrating; he wanted to seize the microphone and drill down as Holmes denied wrongdoing. On a second viewing, scenes felt more “pregnant with meaning,” and the film’s artistry, structure and unusually self-conscious construction won him over.
Roper entered wearing the philosophy embodied by her hat: “Make movies shorter,” specifically “97 minutes with credits.” Three more hours of Holmes sounded excessive, but the movie addressed the question she had carried for a decade: “What does she think is real?”
Her answer remains unresolved because objectivity is deliberately absent. Roper felt manipulated “on every level”—by Holmes, Evans, Oppenheim and Fielder—yet loved the film; its achievement is making viewers conscious of emotional engineering without giving them a clean place outside it.
Max Read experienced the same mechanism as irresistible cringe: “total anxiety, total fascination.” Unlike comedy that makes him hide, this encounter kept him watching as though everyone might go “off the cliff,” turning a three-hour documentary into an oddly adrenaline-driven experience.
8. OpenAI’s alleged bug turns frontier-model concentration into a spy story
The FBI arrested OpenAI employee Abdulrahman Alfasan, described as a 35-year-old from Saudi Arabia, over an accusation that he hid a recording device behind an office television. The hosts fixated on the photograph of him examining it—Griffith’s description was “a cartoon villain that just got gut punched.”
OpenAI said it found a “sophisticated recording device” and “used our models” to identify its origin and the employee who allegedly planted it. Read hears reflexive promotion in that wording: “Everything they do is marketing,” even when announcing a security breach.
The hosts speculate about whether high-level meetings were the target and whether any government was involved, while stressing that details remained unknown. The larger implication worried them more: every government would value access to AI-lab intellectual property, so an easily caught alleged operative raises the possibility of more capable ones.
Griffith connects the episode to remote-work identity risks exposed during COVID, when companies hired people they never met and wondered whether employees were real or secretly holding multiple jobs. Read’s broader inference is that modern espionage should increasingly target corporations holding valuable proprietary technology.
9. Character.ai, HomeView and AI cigarettes reveal three kinds of belief
Character.ai filed for bankruptcy while selling some assets to Walt Disney, citing unprofitability despite roughly 12 million active users and blaming compute and AI-talent costs. Its product also faced lawsuits alleging severe harms, including chatbots encouraging suicide, self-harm or starvation and posing as doctors.
Griffith sees the bankruptcy as unfinished business from 2024’s talent wars. Google brought founders Noam Shazeer and Daniel de Freitas back, along with roughly 20% of employees, while leaving behind a company without its founders, strongest talent, fundraising momentum or obvious strategic future—a “sad, leftover” corporate nub.
Apple’s HomeView is Read’s “productized kitchen iPad”: a smaller screen fixed to a dome, always on and listening, combining messages, calendars and Siri. Griffith initially resisted another kitchen screen, but her reporting on families using Notion and Slack changed her mind; sufficiently complicated households genuinely value tools that offload mental load.
The comic limit of AI belief came from a Palantir CTO’s claim that some researchers smoke to signal confidence AI will cure cancer. Griffith first assumed nihilistic “P(doom)”; learning it was optimism made the behavior stranger. Read’s verdict: believing future AI neutralizes present cancer risk may be “the least cool reason to smoke cigarettes.”