Part IThe Story
Two Reids at One Table
In April 2024, Reid Hoffman sat down across from himself.
The other Reid was a deepfake, a video likeness with a synthesized voice. It ran on OpenAI's GPT-4 and had been trained, by Hoffman's account, on "over 20 years' worth of material": the speeches, the keynotes, the bestsellers, the podcasts, two decades of a man thinking out loud in public. Hoffman was among the first public figures to stage this kind of encounter. The two of them talked about AI regulation and about what the machines could and could not yet do. Then, according to the account of the exchange, they talked about ways Hoffman could improve his LinkedIn profile.
That detail deserves a second look. Hoffman co-founded LinkedIn in December 2002, launched it with a handful of invitations on May 5, 2003, and watched it grow into the closest thing the global economy has to a shared résumé, a registry of 1.3 billion people. He sold it to Microsoft for $26.2 billion. And there he was, a quarter century into the project, taking notes on his own self-presentation from a statistical model of his own sentences. Was the machine teasing him? Was he teasing himself? It is hard to say, and the ambiguity is the point. Hoffman told Axios that ReidAI existed to show "the positive things we can do with all these technologies." Much of his career has followed that pattern. He builds a mirror, holds it up to himself, and calls the reflection an optimism.
The question underneath the stunt is old, and he has been circling it since he was a teenager editing role-playing games in Oakland. What is a profile? Is it a person, or the version of a person that a network can read? Hoffman trained as a philosopher of mind and left philosophy because, as he put it years later, the philosophers "didn't understand thought that much better than anyone else." He then built the world's most consequential machine for professional self-presentation. Now he backs, writes about, and converses with machines that can simulate a self well enough to give it career advice.
The irony did not stop there. In the same years the digital Reid was being trained, the public record of the actual Reid was being rewritten by forces no model could steer: a Senate hearing, a presidential order, a tranche of emails from a dead financier's files. The most connected man in Silicon Valley was learning, again, that connections have two ends.
By the Numbers
The Network and Its Builder
8thHoffman's place among Stanford students to declare a symbolic systems major
$10MHoffman's take from eBay's 2002 acquisition of PayPal
$2.34BEstimated value of his LinkedIn stake at the May 19, 2011 IPO
$26.2BMicrosoft's all-cash price for LinkedIn, its largest acquisition until 2022
1.3BRegistered LinkedIn members as of March 2026
37+AI companies backed by Hoffman and Greylock as of May 2023
$2.7BForbes estimate of Hoffman's net worth, May 2026
The Eighth Name on the List
The parents were both attorneys, and they divorced when their son was a year old. He was born in Palo Alto on August 5, 1967, and raised in Berkeley, which in those years was the kind of place where a precocious child could find adult work without anyone thinking it strange. Hoffman found his at Chaosium, the tabletop game company then based in nearby Oakland, where he worked as an editor. He was a teenager refining the rules of imagined worlds: how a character advances, how alliances form, what a player can and cannot do. The vocabulary of that work—levels, connections, the rules governing who can reach whom—would turn out to be a very specific apprenticeship.
He was also restless, and he negotiated about it. Convinced he learned better in small classes, he talked his parents into a private high school close to home. That was not far enough. He then persuaded them to send him across the country to the Putney School in Vermont, a progressive boarding school that took breadth seriously enough to put its students to work on the farm. It is a pleasant thing to picture: a future billionaire of the social graph doing barn chores in the Green Mountains, deep in country where there was nothing to network with but cows.
Stanford came next, and with it the decision that shaped everything after. Hoffman's interests sprawled, and a sympathetic adviser asked him to write out every course he hoped to take. The list pointed toward a new interdisciplinary program called symbolic systems, which combined philosophy, linguistics, psychology, mathematical logic, and computer science. Hoffman has said he was the eighth person to declare it. The major was, roughly, the study of how minds and machines represent the world, and the question that hooked him was how ideas move between people. He read Marshall McLuhan, whose line "the medium is the message" was already a cliché, and Neil Postman, whose Amusing Ourselves to Death argued that television was quietly destroying the public's capacity for sustained argument. A young man who took both men seriously would come to understand that the shape of a communication channel decides what can be said through it. Later he would build channels for a living.
He graduated in 1990 with a bachelor's degree in symbolic systems and cognitive science, won a Marshall Scholarship, and went to Wolfson College, Oxford, where he took a Master of Studies in philosophy in 1993. Wolfson is a graduate college for people who intend to stay in the academy, and Hoffman meant to. He wanted to be a professor, a writer, a public intellectual. At Oxford he worked through both the analytic schools that dominate English-speaking philosophy and the continental traditions with their attention to history and context. He came away believing he had learned how to think and doubting that the academy was where thinking changed anything.
"I concluded that we were a long ways off," he said of artificial intelligence, recalling the period decades later at a fireside chat with the investor Elad Gil. "Got a little bit like maybe should study philosophy for thinking, studied philosophers, didn't understand thought that much better than anyone else. And so then went into entrepreneurship."
The sentence has the quality of a shrug, which hides how large a decision it describes. It also has an ancestry. Hoffman's paternal great-great-great-grandfather was Theophilus Adam Wylie, a Presbyterian minister who served as president pro tempore of Indiana University. Wylie was a man of the pulpit and the lectern, a nineteenth-century custodian of institutions built to shape how people think. His uncle, Eric Hoffman, became a writer. The family had produced preachers and authors. Reid, the would-be professor, would end up doing a version of both through software, books, and an unending sequence of podcasts. Business Insider once ran his own verdict under a headline: the philosophy master's had helped him more than an MBA would have. In 2017, when Britain made him an honorary Commander of the Order of the British Empire, the citation read, in part, "for services to promoting UK business and social networking and the Marshall Scholarship scheme." The Oxford detour he had abandoned came back to him as a medal.
A Dating Service for Golfers
He started in technology as a temp.
Back in Northern California in 1993, Hoffman canvassed his Stanford friends for jobs and landed a temporary position in user experience at Apple Computer. It became a longer assignment as a junior product manager on eWorld, Apple's early attempt at an online service, a cartoon town square of a place where users could gather, send mail, and read news. It did not last. eWorld ended up at AOL in 1996, one of many early experiments that understood people wanted to gather online and misjudged the architecture of the gathering. Hoffman watched it fail from close range. He moved on to Fujitsu, and then, in 1997, he founded his first company.
SocialNet.com focused, in a description that sounds almost comic now, "on online dating and matching up people with similar interests, like golfers who were looking for partners in their neighborhood." There are worse first ideas. Its premise, that the internet could introduce strangers whose interests overlapped, was sound enough to describe half the consumer internet of the next twenty years. The flaw lay in the detail. A neighborhood golf partner is a weak reason to hand over your identity to a website, and SocialNet never found the reason that would make people do it. The company was hard to explain, hard to grow, and, by Hoffman's later accounting, a long education in what happens when the theory of what people want is slightly wrong.
The failure shows up later in his own words, sometimes obliquely. In 2025, WIRED's Katie Drummond ran him through a set of rapid-fire questions and asked him to name the hardest lesson he had ever had to learn. "Oh gosh, there's a whole lot," he said. "Probably when to give up."
He also rejected the most fashionable lesson available to someone in his position. When founders are told to "fail fast," Hoffman wrote on Greylock's site, "I think, No, no, no! The goal isn't to fail fast—it's to learn fast by tackling your most dangerous potential points of failure." The distinction matters. Failing fast treats the failure as the product. Learning fast treats it as the cost of the product, a tuition payment, and tuition is something you try to pay only once. The most often quoted image from his advice to entrepreneurs comes from these years: "Starting a company is like jumping off a cliff and assembling a plane on the way down." SocialNet was the plane that never flew. He kept the parts, and he kept two of the people. One of them was Allen Blue, who would co-found LinkedIn with him and remain there for decades.
The lesson went beyond a list of mistakes to avoid. The golfers had been the wrong anchor. People would not expose themselves online for a hobby. They might for something more consequential, something with money and status at stake. They might do it for work.
No Place for a COO
Before he could test that idea, he had to survive a different sort of company.
Hoffman had been a member of PayPal's board of directors from the beginning, recruited into the circle around Peter Thiel and Max Levchin, the Stanford-and-Kyiv pairing whose payments startup merged with Elon Musk's X.com. In January 2000 Hoffman left SocialNet and joined PayPal full-time as chief operating officer. It was an operating title at a company whose operations were, by every account, close to chaos: fraud losses mounting, a merger culture still unsettled, executives fighting about everything from software platforms to brand names.
Then the title vanished. Jimmy Soni, in his history of PayPal, The Founders, reconstructed the morning. "On June 1, 2000—nineteen days into his tenure—Musk introduced a reworked executive structure. He would now count seven direct reports." The list put David Sacks over product and Levchin as CTO "with no direct reports up to him." It made Reid Hoffman "senior VP of business development and international." Soni then added the sentence that matters for this story: "Notably, the reorganization left no place for a COO or president."
Hoffman did not quit. He took the narrower job and widened it. Greylock's later biography describes his PayPal role as executive vice president "responsible for all of the company's external relationships": the banks, the card networks, the regulators, the eBay merchants whose livelihoods ran through the service, the international partners. PayPal's internal life was a sequence of coups and counter-coups. Musk was himself removed as CEO later that year. Through all of it Hoffman worked at the boundary between the company and everyone it depended on. He was learning the craft of being a node, the person through whom other people's interests have to pass.
The micro-biographies of that era rhyme with his. Keith Rabois was a PayPal executive who followed Hoffman to LinkedIn and later joined Slide; he would tell the Los Angeles Times that "Reid sees the next move on the Internet better than anyone in Silicon Valley." Thiel was a Stanford philosophy student turned contrarian investor and became one of LinkedIn's backers. Musk was the man who had eliminated Hoffman's title. Two decades later, when WIRED asked Hoffman for the biggest difference between himself and Musk, he answered with one word: "Sanity." The two men would cross again, at a dinner, on a board, and in the wreckage of American politics. The PayPal alumni later took the gangster's nickname "mafia" with evident pleasure. Like a family, they spent the following decades funding, quoting, and occasionally despising one another.
eBay bought PayPal in 2002. Hoffman's share came to about $10 million, a fortune by any normal standard and a small stake by the standards of what came next. The internet was in ruins, and venture firms were writing off entire portfolios. He had a theory about work, a few former colleagues willing to follow him, and enough money to be his own first investor.
The Invitation
The incorporation papers for what would become LinkedIn Corporation were filed in Delaware on March 6, 2003, under a slightly stiffer name: LinkedIn, Ltd. The company had already existed for several months as a team in Hoffman's orbit: Allen Blue and colleagues from SocialNet, a former colleague from Fujitsu, engineers and operators from PayPal. Eric Ly, Jean-Luc Vaillant, Lee Hower, and Konstantin Guericke are among the names in the founding roster. They worked out of Mountain View, at an address on Stierlin Court that would later earn the label "former headquarters."
On May 5, 2003, the co-founders sent invitations. That was the launch: no press event and no advertising budget, just a few people asking a few hundred others whether they wanted to be connected. The company's own description of its design calls it a "gated-access approach," meaning that contact with any professional "requires either an existing relationship, or the intervention of a contact of theirs." It was meant to build trust, and it was an inversion of SocialNet. SocialNet had introduced strangers who shared an interest. LinkedIn would introduce strangers only through people they already knew. Hoffman's insight was that professional reputation already worked this way offline. The introduction, the reference, and the friend of a friend were the actual circuitry of careers. LinkedIn made that circuitry visible and searchable, and then, eventually, saleable.
Growth came slowly and then compounded. The site reached a million users in August 2004. In July 2005 it launched a premium service for "power users like recruiters, analysts and researchers," and in August added business accounts with more powerful search. In March 2006 the company recorded its first profitable month, an early milestone for the social networks of that era, which were mostly famous for burning money. Sequoia Capital had led the first institutional round in late 2003, a $4.7 million investment the company announced under the punning headline "Sequoia Capital 'Links In.'" Bessemer and the European Founders Fund put in $12.8 million in January 2007. In June 2008, a group including Bain Capital Ventures, Sequoia, and Greylock bought 5 percent of the company for $53 million, a post-money valuation of roughly $1 billion.
The features from these years read as a series of wagers about vanity and need. In May 2007 LinkedIn launched "who's been viewing your profile," which turned every visitor into a small mystery and every member into a returning user. In April 2008 came "People You May Know," which turned the network's map into a list of nudges. Neither feature was neutral, and Hoffman would later argue that no technology is. Each one rested on a theory of human nature, and the theory was not complicated. People want to know who is looking at them, and they want to know whom they are missing.
He was also learning what kind of executive he was not. In February 2007 Hoffman moved from CEO to chairman and president, and Dan Nye became chief executive. Jeff Weiner, a former Yahoo executive, arrived soon after and would run LinkedIn for eleven years, turning Hoffman's network into a disciplined business. Weiner later set out to build what he called an "economic graph," a digital map of every job, every skill, every company, and every professional in the world economy. In 2011 Hoffman and Weiner shared the Ernst & Young U.S. Entrepreneur of the Year Award. The division of labor held: Hoffman, the theorist of the network, and Weiner, the man who ran it.
The invitation that launched the company eventually came back as a lawsuit. LinkedIn's sign-up flow asked new members for their email password and then offered to invite their contacts, and the address book opened with every address already selected. Members who clicked through sent invitations to everyone they had ever emailed. Contacts who ignored the first invitation received two more. In September 2013, a class action titled Perkins v. LinkedIn Corp. argued that this was not an invitation so much as a hijacking. LinkedIn argued that its messages were protected speech. In November 2014 a federal judge disagreed, ruling that the invitations were advertisements. The court accepted that members had consented to the first message, but not to the reminders. LinkedIn settled in 2015 for $13 million. Case No. 13-CV-04303-LHK sits in the record as a footnote to the founding gesture: a network built on invitations had learned how hard it is to stop sending them.
There were other footnotes. In June 2012, the cryptographic hashes of roughly 6.4 million LinkedIn passwords were stolen and posted online, and security researchers pointed out that the company had not salted its password file. In May 2016, 117 million credentials, believed to come from the same breach, were offered for sale for about $2,200. A network is only as trustworthy as its weakest gate.
His Reidness
In July 2008, a month after the round that valued LinkedIn at $1 billion, the Los Angeles Times reporter Jessica Guynn published a portrait of Hoffman. It is the most vivid document from his middle years, a catalog of excess in everything except the things money usually buys.
"Reid Hoffman is a big man in Silicon Valley," the piece began. "And, try as he might to remain in the background, his stature just keeps growing—literally and figuratively." Guynn counted. Hoffman's email account held ten gigabytes. His LinkedIn network ran to 1,684 connections "and counting." He sat on seven corporate boards and had invested in more than sixty companies. He worked at three computers at once, with a few hundred windows open on the desktop. He was 40, and by his own estimate gaining about ten pounds a year, a consequence of what he called his "active but sedentary start-up lifestyle." He did not step on scales; he simply said he weighed "too much." His friends had hired him a personal trainer, and he was reluctant to give the sessions even a few hours a week.
Then came a detail that rewrites the rest. Until a few weeks before the article ran, Hoffman and his wife, Michelle Yee, whom he had married in 2004, had been living in an 876-square-foot, two-bedroom apartment in Mountain View. The man who had invested early in Facebook, Digg, Flickr, and Six Apart was living in a space smaller than many of his portfolio companies' lobbies. Startups he had backed early had by then sold for a collective $1.4 billion.
— David Siminoff, venture capitalist, to the Los Angeles Times (2008)Reid's big. His ideas are big. His vision is big. His heart and brain are big. He's almost one of those mythic characters like Babe Ruth who, when he ate, he ate nine hamburgers; when he drank, he couldn't see straight; and when he hit the baseball, it would keep going until Monday.
One friend, Guynn reported, referred to him as "His Reidness." The only thing that wasn't big, she wrote, was his ego. He received "scant attention compared with the entrepreneurs he has bankrolled," and that was how the arrangement worked. Hoffman's style of investing was the PayPal job carried over. He did not need the spotlight. He needed to sit in the middle of things.
The most famous example was a meeting. According to David Kirkpatrick's book The Facebook Effect, Hoffman arranged the first meeting between a twenty-year-old Mark Zuckerberg and Peter Thiel. Thiel wrote a $500,000 angel check, and Hoffman invested alongside him in Facebook's first financing round. Hoffman was already running his own social network and might have been expected to see Facebook as a competitor. He seems to have seen it as a different branch of the same tree, and he had the instinct, rare in someone that competitive, to route the opportunity rather than hoard it. Dave Goldberg, the late SurveyMonkey chief executive, put the reputation plainly: Hoffman "is the person you want to talk to when you are starting a company."
In 2009 he joined Greylock Partners. The firm gave him a $20 million seed fund in 2010, making him what TechCrunch called a "Super Angel-Turned-VC." He led Airbnb's Series A, sat on Zynga's board from March 2008 to June 2014, and in May 2012 ranked third on the Forbes Midas List, where the magazine called him "Silicon Valley's uber-investor" who "has had a hand in creating nearly every lucrative social media startup." The same year, Forbes put him on its cover as "Silicon Valley's Best-Connected Billionaire." The description was accurate, and it would later be turned against him.
One person had an unusually close view of those years. Ben Casnocha met Hoffman as a young entrepreneur and writer, became his co-author on The Start-Up of You, and spent, by his own count, roughly ten thousand hours working alongside him. Casnocha's best scene takes place in the air. It was December 2012, and the two of them were flying home to San Jose on a private plane three hours after landing in Las Vegas, where Hoffman had given a talk at a tech event. They debriefed. Hoffman replayed his answer about how Greylock differed from other venture firms, and Casnocha took notes. Then the conversation turned, as it increasingly did, to whether the trip had been worth it.
Casnocha quoted E.B. White: "I wake up in the morning unsure of whether I want to savor the world or save the world. This makes it hard to plan the day." Hoffman, he wrote, liked to savor in his way: intellectual epiphanies, time with friends, Settlers of Catan. But mostly he wanted to save. That night on the plane Hoffman looked exhausted, and Casnocha remembered thinking that his friend should do more things just for himself.
— Ben Casnocha, '10,000 Hours with Reid Hoffman: What I Learned'He is among the most selfless and externally-generous people I've met in my life.
The scene has a quiet poignancy. Hoffman had arrived at the condition everyone in Silicon Valley says they want: almost no constraints on time, access, or money. He found it oppressive in a specific way. Every choice now carried an enormous opportunity cost, because the list of things not done was nearly infinite. He fell back, as he often did, on a game. Hoffman has said that board games are good training for business, and the Financial Times once ran an interview with him under the headline "Board games inspired my business strategy." Asked by WIRED whether he preferred cooperative or competitive games, he chose cooperative, which is a revealing answer from a man who spent his life winning competitive markets.
Sixty Dollars a Head
LinkedIn filed to go public in January 2011 and traded its first shares on the New York Stock Exchange on May 19 under the ticker LNKD, priced at $45. The stock rose as much as 171 percent during the day and closed at $94.25. Hoffman's stake was worth an estimated $2.34 billion, not counting whatever he stood to gain through Greylock. The company had gone from about 500 full-time employees in 2010 to about 2,100. That year it earned $154.6 million in advertising revenue alone, more than Twitter's $139.5 million, an awkward fact for anyone who still thought of LinkedIn as the dull cousin of social media. In 2011, members joined at a rate of about two every second.
A public company needs more surfaces to sell, and the next few years built them. In September 2012 LinkedIn let members "endorse" one another's skills. The feature allowed only positive endorsements, and because the algorithm suggested skills members might have, people sometimes appeared to be requesting endorsements for abilities they did not possess. In October 2012 came the Influencers program, an invitation-only roster of "thought leaders" that by May 2016 numbered more than 750, with Richard Branson, Narendra Modi, Bill Gates, and Martha Stewart among them. Acquisitions followed in quick succession: SlideShare for $119 million, Pulse for $90 million, Bizo for $175 million, and in April 2015 Lynda.com for $1.5 billion, the company's largest. LinkedIn engineers built and, in 2011, open-sourced Apache Kafka, a piece of data infrastructure that ended up running far outside LinkedIn's walls. In February 2014 LinkedIn launched a Simplified Chinese version under the name 领英, Lǐngyīng, which translates roughly as "leading elite." Weiner acknowledged in a blog post that the company would have to censor some content to comply with Chinese rules.
The intellectual case for all of this came in book form. On February 14, 2012, Hoffman and Casnocha published The Start-Up of You, which urged readers to treat themselves as businesses-of-one, each the "CEO of their own career." It became a New York Times and Wall Street Journal bestseller. The Economist found "a number of astute observations about shifts in the world of work." It also served as an instruction manual for using LinkedIn well. Two years later came The Alliance, written with Casnocha and Chris Yeh, which argued that neither lifetime employment nor free agency fit a networked economy and proposed treating employer and employee as "allies" for defined "tours of duty." The Financial Times was lukewarm. Readers who knew the territory, it suggested, would find little that surprised them. Hoffman was building the platform and writing its philosophy at the same time.
Not everyone took that philosophy at face value. Academic researchers working with LinkedIn data found effects its founders had not advertised. In 2017 the sociologist Ofer Sharone interviewed unemployed workers and described a "filtration effect" that "has little to do with evaluations of merit." It turned on how profile photographs looked to hiring managers, on whether job seekers could build narratives that fit the screening, and on how a single résumé penalized people who worked across fields. In 2019, sociologists interviewing recruiters found that the high-skilled market had moved to LinkedIn, where recruiters courted the already employed, while active job seekers were pushed onto crowded job boards. In 2020, economists found that high-status people self-select into professional networks, not the dissatisfied workers who might need them most. The network rewarded those who were already well connected. Anyone who knew Hoffman's biography would not have been surprised.
The market's patience gave out on a single day. In February 2016, after an earnings report, LinkedIn shares fell 43.6 percent to $108.38, erasing about $10 billion in market value in one session. Four months later, on June 13, 2016, Microsoft announced it would buy the company for $196 a share in cash, $26.2 billion in total, financed with debt. That worked out to roughly $60 per member. It was the largest acquisition in Microsoft's history until the Activision Blizzard deal in 2022. Microsoft promised that LinkedIn would "retain its distinct brand, culture and independence," and that Weiner would stay on and report to Satya Nadella. The deal closed on December 8, 2016. On March 14, 2017, Hoffman joined Microsoft's board.
What was Microsoft buying? Analysts said integration with Office. Possibly it was also buying a habit. By November 2023 LinkedIn had a billion members, and by March 2026 it had 1.3 billion, with about 1.4 billion monthly visits. Only about one percent of monthly users share content in a given week. The rest look and are looked at, which is to say they use the feature Hoffman's team launched in May 2007. In May 2024 LinkedIn added daily puzzle games—Pinpoint, Queens, Crossclimb—with completion times compared across one's network, streaks, two "freezes" per week, and notifications twice a day. Hoffman had once edited game rules in Oakland. Now the network he built had turned into a game. In July 2026, after a period in which it had invited users to "expand" their posts with AI, LinkedIn introduced a button that let members flag content that "seems like AI slop."
Homo Techne
The symbolic-systems major came back to its first subject.
Hoffman was among the earliest donors to OpenAI, the nonprofit research lab founded at the end of 2015 by Sam Altman, Elon Musk, and others. It was the second time he and Musk had shared an organization. When Musk left OpenAI's board in February 2018, citing conflicts with Tesla's work on self-driving cars, the lab was short of the money it had expected. Hoffman helped engineer the solution, a "capped-profit" structure with a for-profit subsidiary that could raise outside capital. He led the investment into that subsidiary through his family foundation and joined the board, where he served from 2019 to 2023. On March 3, 2023, he resigned, citing potential conflicts among his OpenAI seat, Greylock's AI investments, and a company he had co-founded. When OpenAI's board briefly ousted Altman later that year, The Information reported that Hoffman had been the one calming nerves at Microsoft. He was again the person at the boundary.
The company he had co-founded was Inflection AI, announced in March 2022. His partner was Mustafa Suleyman, a co-founder of DeepMind and by then a Greylock colleague and longtime friend. Suleyman had grown up in London, left Oxford, worked in human-rights policy and conflict resolution, and then built one of the most important AI labs in the world. Like Hoffman, he had moved from the study of human problems to the engineering of machines meant to address them. Inflection aimed, in CNBC's paraphrase, to make it "easier for humans to communicate with computers." It lasted two years as an independent ambition. In March 2024 Microsoft licensed Inflection's technology and hired Suleyman, chief scientist Karén Simonyan, and most of the company's roughly seventy employees, in a deal reported at about $650 million. In September 2024 Britain's Competition and Markets Authority cleared the arrangement. It was the second time a Hoffman company had ended up in Redmond.
In January 2025 he started over again with Manas AI, an AI-driven drug discovery company. His co-founder was Siddhartha Mukherjee, the Pulitzer Prize-winning oncologist whose cancer biography made a disease into a narrative; the chemist Jonathan Baell joined the founding team. The company raised $24.6 million in seed funding from Hoffman, General Catalyst, and Greylock. The co-founder pattern is telling. Hoffman chooses partners who explain things: a writer about cancer, a researcher of human conflict, a coauthor of career advice. He likes to build alongside people who could also be teaching a seminar.
He also kept writing. On March 13, 2023, he published Impromptu: Amplifying Our Humanity Through AI, which he said he had written with GPT-4 as his "author's co-pilot." It was a book about the machine written partly by the machine, and it became a Wall Street Journal bestseller. On January 28, 2025, with the journalist Greg Beato, he published Superagency: What Could Possibly Go Right With Our AI Future, which sorted the public into four tribes: doomers, gloomers, zoomers, and bloomers. Hoffman counted himself a bloomer, a technology optimist committed to "smart risk-taking versus blind risk-taking." Buyers could request AI-generated custom covers with personalized blurbs. The book reached the New York Times bestseller list. He has called proposals to pause advanced AI development "foolish" and "anti-humanist," and he has described himself in a phrase that sounds tossed off but is carefully balanced: "I'm a tech optimist, not a tech utopian."
The fullest version of the argument came in December 2024, at the London School of Economics, where he had first visited as an Oxford master's student in the early 1990s. He opened with Heraclitus: one cannot step into the same river twice, because the river changes and so does the person stepping in. Then he made the claim that runs through his whole career.
— Reid Hoffman, lecture at the London School of Economics, December 2024We are more than Homo sapiens. If we merely lived up to this scientific classification and just sat around thinking all day, we'd be much different creatures than we actually are. We humans are Homo techne: humans as toolmakers and tool users.
"We shape our tools," he told the room. "Then our tools shape us." Fifteen months earlier, on September 8, 2023, he had given the commencement address at the Bologna Business School, at the oldest university in the Western world. There he told graduates that AI should stand not for "artificial intelligence" but for "amplification intelligence," called it "a steam engine of the mind," and said that "any new venture must have a strong theory of human nature underlying it." He then had the speech translated by AI into more than a hundred languages "with no human editing," as a demonstration. It is easy to smile at the gesture. It is also consistent with everything else he has done. He has never trusted an argument he could not ship.
There is a strain of self-awareness in his optimism that the critics tend to overlook. Asked by the Observer in 2025 what kept him up at night, he did not name rogue superintelligence. He named lock-in: AI systems, he said, "are not just tools but adaptive infrastructures," learning systems that "improve through use, which will theoretically make it even harder for challengers to unseat incumbents down the line." He worried this "could lead to a world where we see a decline in the kind of entrepreneurial dynamism" that creates jobs and competition. The man warning about AI lock-in had built one of the most locked-in networks in the history of the internet. He knows how that works from the inside, perhaps better than anyone.
The Donor Table
Every few months, Theodore Schleifer reported in Vox in 2020, Hoffman sent an invitation to some of the billionaires who fund the Democratic Party. In Schleifer's phrasing, "He'd like to add you to his political network." The sentence reads like LinkedIn interface copy, and that is part of its sting. The invitations led to closed-door sessions under the Chatham House Rule, in Washington, Palo Alto, or later on Zoom, where advisers to dozens of the party's biggest donors compared notes and heard pitches from campaign officials. That year Hoffman was spending as much as $100 million of his own money to defeat Donald Trump, and his aides said he had steered hundreds of millions more. The Vox article, based on four dozen interviews, found him "as popular in San Francisco as he is despised in parts of Washington." His team believed the party was broken and needed well-funded disruption, and some of the disruption had gone badly. "Whether they love Hoffman or hate him," Schleifer wrote, "Democrats are scared to cross him—and lose access to his wallet." Hoffman declined to be interviewed.
The record shows a donor who treated politics like a portfolio, with the expected rate of write-offs. He was listed among the founders of FWD.us, the pro-immigration group, in 2013. He gave $150,000 to Lawrence Lessig's Mayday PAC in 2014, and that same year spent $500,000 on an independent committee devoted to attacking a San Francisco Assembly candidate. In 2016 he created a card game modeled on Cards Against Humanity to mock Trump; the game-maker was still playing. In 2018 he supplied half of the $35 million used to start Alloy, a data company meant to repair the Democrats' voter files; it shut down in 2021. In December 2018 The New York Times reported that he had put $100,000 into a project that "adopted Russia-inspired political disinformation tactics on Facebook" during the 2017 Alabama Senate race. Hoffman apologized and said he had not known what the nonprofit he funded, American Engagement Technologies, was doing. He helped fund E. Jean Carroll's lawsuit against Trump. On May 9, 2023, a jury found Trump liable for sexual abuse and defamation and awarded Carroll $5 million. In 2022 he gave at least $500,000 to Mainstream Democrats, a super PAC supporting moderates, including Henry Cuellar, who was facing a progressive challenger. In 2024 he publicly said he hoped a Harris administration would replace FTC chair Lina Khan, which put him at odds with his own party's left.
He joined the Pentagon's Defense Innovation Board in October 2022. He also turned up wherever the establishment gathered: a regular at Bilderberg since at least 2011, a Council on Foreign Relations member since 2015, and, according to Forbes, listed in the directory of Peter Thiel's private Dialog gatherings. The PayPal circle still overlapped, even as its members spread across the political spectrum.
When much of Silicon Valley moved toward the second Trump administration, Hoffman did not. "Not kissing [Trump's] ring, like many others have, is probably an economic limiter," he told The Guardian in 2025, "but it's better to be principled." In the same interview, asked why anyone should trust the AI optimism of a man who profits from AI, he gave the most revealing description of his method on record.
— Reid Hoffman, to The Guardian (March 2025)I tend to start with my beliefs and follow with my money. And sometimes that does mean doing things that are against my economic interests.
The costs were not only economic. In December 2024 The Sunday Times ran an interview under the headline "LinkedIn's Reid Hoffman: I've had threats of violence since Musk slanders." The colleague who had eliminated his PayPal title had become, a quarter century later, one of his loudest attackers.
The Most Connected Man in Silicon Valley
In 2023, ahead of Microsoft's annual meeting, a conservative group called the National Legal and Policy Center filed a proxy memorandum urging shareholders to vote against Hoffman's reelection to the board. Its first exhibit was a phrase. Hoffman, it said, was "the most connected man in Silicon Valley," and it cited his own biography at an MIT research initiative as the source. The memo used the compliment as an indictment. The thing he had been praised for, the density of his network and the speed with which he could reach anyone, was presented as evidence of poor judgment. The argument concerned one connection in particular.
What is established is this. Hoffman's contact with Jeffrey Epstein reportedly began at the MIT Media Lab, where he helped solicit donations from Epstein. The lab's director at the time, Joi Ito, was a friend; with Ito, Hoffman had created the MIT Disobedience Award, a $250,000 prize for acts of disobedience with positive social impact. In July 2013, Epstein met with Hoffman and others on the MIT campus. In 2019 it was reported that Hoffman had hosted a dinner attended by Epstein, Elon Musk, and Mark Zuckerberg. In September 2023 The Wall Street Journal reported that Hoffman had visited Epstein's private island for a weekend in 2014. Hoffman apologized in 2019, said his contact had been limited to fundraising for MIT, and said he regretted it.
In the autumn of 2025, the episode became a weapon. On October 7, testifying before the Senate Judiciary Committee and pressed about Epstein's ties to Trump, Attorney General Pam Bondi repeatedly brought up Hoffman and called him "one of Epstein's closest confidants." Hoffman denied it. On November 14, after newly released documents detailed Trump's own ties to Epstein, the president directed the Justice Department to investigate Epstein's relationships with several Democrats, Hoffman among them. Hoffman responded the same day, calling the probe "political persecution and slander," saying he had never been a client of Epstein's or engaged with him except in fundraising, and demanding that Trump "release all of the Epstein files: every person and every document in the files." He then paid for a public-service announcement, produced by World Without Exploitation, in which Epstein's victims called for the files' release. It ran nationally during Monday Night Football. Hoffman, who had spent his life arguing that transparency makes networks more efficient, was now paying to make one of the darkest networks of his era visible, including his own place in it.
In February 2026, more files were released. Among them was an email Hoffman had sent on Christmas Eve 2014, mentioning "ice-cream...for the girls" and another item meant to "strike your funny bone for the island." The fragments circulated with little context, and their meaning was argued over by people who had already decided what they believed.
Nothing in this section changes what LinkedIn is or what Hoffman built. It does change what a reader hears in the founding image: the gated network, the trust that passes through a mutual contact, the introduction that vouches for you. That architecture works in both directions. Every connection lends a little reputation and borrows a little. Hoffman knew this better than almost anyone. He had made it the basis of a $26.2 billion company and then had to watch it work against him.
At the end of 2025, Hoffman made his friends a Christmas present. He told WIRED about it in December, when asked for his "one, personal, killer use case for AI." He had generated a holiday album, AI-made music for a season he wanted to celebrate with "irony as well as affection," including a song about ugly sweaters. Everyone on the list knew where it came from. In Drummond's half-teasing summary, it went from his heart into the AI and on to their Christmas trees.
There was one more detail, and he offered it with evident pleasure, as if it were the punchline he had been saving.
"And it's on records," he said. "We put it on records."
Part IIThe Playbook
The principles below come from the pattern of decisions in Part I, not from Hoffman's own frameworks, though they sometimes overlap with them. He has published his methods more generously than almost any founder of his generation, in books, podcasts, a free Stanford course on blitzscaling, and a thousand LinkedIn posts. The more useful lessons are often in what he did rather than what he prescribed, and in the places where the two diverge.
Twelve Principles
- 1.Begin with a theory of human nature, not a product.
- 2.Treat your first company as tuition you pay only once.
- 3.Lose the title, keep the seat.
- 4.Launch as an invitation, and know when to stop inviting.
- 5.Monetize the minority who need you most.
- 6.Hand the operating keys to an operator.
- 7.Be the router, not the endpoint.
- 8.Sell the map, not the quarter.
- 9.Write the manual for your own market.
- 10.Ship the argument, iterate the ethics.
- 11.Disclose the conflict, then argue anyway, or step back.
- 12.Audit the edges of your network before someone else does.
R
The Decision Record
Inflection points in Hoffman's career where a single choice set the trajectory for the next decade.
1993
Leaves an Oxford philosophy track for technology, starting as a temp at Apple.
1997
Founds SocialNet. Matching strangers by interest fails, but the team survives.
2000
Loses the PayPal COO title in Musk's June 1 reorganization; stays to run external relationships.
2003
Launches LinkedIn on May 5 with invitations and a gated-access trust model.
2006
LinkedIn reaches its first profitable month on premium subscriptions.
2007
Steps aside as CEO for Dan Nye; Jeff Weiner follows and runs the company for 11 years.
2009
Joins Greylock, institutionalizing the angel network built after PayPal.
2016
Sells LinkedIn to Microsoft for $26.2B four months after a 43.6% one-day stock drop.
2019
Helps restructure OpenAI as a capped-profit company and joins its board.
2023
Resigns from OpenAI's board on March 3 to avoid conflicts with Greylock and Inflection.
2025
Launches Manas AI with Siddhartha Mukherjee; publishes Superagency.
Principle 1
Begin with a theory of human nature, not a product
The symbolic-systems degree and the Oxford philosophy were not detours. In his 2023 Bologna address Hoffman said that "any new venture must have a strong theory of human nature underlying it," and his career reads as a series of experiments in that claim. SocialNet assumed people would expose themselves to find a golf partner. LinkedIn assumed they would do it for their careers and would value an introduction through a friend over a stranger's cold contact. "Who's viewed your profile" assumed curiosity about one's own audience would bring people back. Each feature was a small hypothesis about motivation, and the ones that worked were the ones with the more accurate psychology.
What matters here is the order of operations. Founders often start with a capability, something a new technology makes possible, and then search for someone who wants it. Hoffman, trained by McLuhan and Postman to think about how media shape behavior, started from the behavior. The professional reference, the warm introduction, and the fear of being invisible all existed long before the website. LinkedIn gave them a software surface.
Tactic: Before writing a product spec, write one paragraph describing the specific human desire or fear your product exploits, and test that paragraph against how people already behave offline.
Principle 2
Treat your first company as tuition you pay only once
SocialNet failed, and Hoffman has spent decades insisting that the lesson is not to celebrate failure. "The goal isn't to fail fast—it's to learn fast by tackling your most dangerous potential points of failure," he wrote. What he took from SocialNet was specific. The interest-based matching was too weak a reason to share an identity, and the team was strong enough to keep. Allen Blue came with him. The conceptual move from introducing strangers who share an interest to introducing strangers through people they already trust is the move from SocialNet to LinkedIn, and it could only have come from someone who had watched the first version stall.
His "hardest lesson," he told WIRED, was "probably when to give up." The admission is useful because it reverses the usual mythology of grit. Giving up well means stopping the company while keeping the insight and the people, and that is a skill most founders never deliberately practice.
Tactic: When you shut down a project, write down the one assumption that proved wrong and the two people you would rebuild with, and start the next venture only when you can name the replacement assumption.
Principle 3
Lose the title, keep the seat
On June 1, 2000, Elon Musk reorganized PayPal's executive team and, in Jimmy Soni's words, "left no place for a COO or president." Hoffman, the COO, became senior vice president of business development and international. Many executives would have read the demotion as a signal to leave. Hoffman instead took on the company's external relationships and made himself essential where the company met the outside world: banks, card networks, partners, regulators.
The payoff came in three forms. He stayed through the eBay sale and left with $10 million. He kept his place in the PayPal circle, which became the deal-flow network for the next decade. He also learned the job he would do for the rest of his career, which was managing boundaries between parties who need each other and do not trust each other. Titles can be taken away by a reorganization. A relationship map cannot.
Tactic: If you are demoted inside a company you still believe in, volunteer for the role that touches the most outside parties, because that network leaves with you.
Principle 4
Launch as an invitation, and know when to stop inviting
LinkedIn launched on May 5, 2003, with no announcement. The co-founders sent invitations. The "gated-access approach," in which contact required an existing relationship or a mutual introduction, made trust the product's default setting. Growth came along the edges of existing relationships, which is cheaper than paid acquisition and produces a better-quality network.
The same mechanism later produced Perkins v. LinkedIn. Address books opened with every contact preselected, and two automated reminders followed each ignored invitation. A federal court found that members had consented to the first message, but not to the follow-ups, and LinkedIn paid $13 million to settle. The founding growth loop had turned into a liability once the company pushed it past consent. An invitation carries a member's reputation, and every automated reminder spends a bit of it without asking.
!
The Invitation Spectrum
Where LinkedIn's growth mechanics sat on the line between consent and extraction.
| Mechanic | Year | Effect | Read |
|---|---|---|---|
| Co-founders send personal invitations | 2003 | Trust-based seed network | Scale |
| "Who's viewed your profile" | 2007 | Curiosity-driven return visits | Scale |
| "People You May Know" | 2008 | Network densification | Scan |
| Algorithm-suggested skill endorsements | 2012 | Engagement; criticized as meaningless | Try |
| Preselected address-book invites plus two reminders | 2013–15 | $13M class-action settlement | Risk |
Tactic: Build your growth loop on members vouching for you, and set a hard internal rule that no message goes out under a member's name without an affirmative action from that member.
Principle 5
Monetize the minority who need you most
LinkedIn launched premium products in 2005, a service for "power users like recruiters, analysts and researchers" and business accounts with better search. By March 2006 it had its first profitable month, early among the major social networks of its era. The company did not need most members to pay. It needed the small group for whom the network was a working tool to pay well. Recruiters searching for passive candidates were the obvious customers, and research later confirmed that high-skilled hiring had moved onto the platform.
The logic still holds. Today about one percent of LinkedIn's monthly users post weekly. The audience is huge and mostly passive, and the revenue comes from the people who need to reach that audience: recruiters, salespeople, advertisers. This is a two-sided structure in which the many supply the value and the few pay for access.
Tactic: Identify the members for whom your free product is already a professional tool, and design your first paid tier around the specific search or reach capability they would otherwise have to give up.
Principle 6
Hand the operating keys to an operator
In February 2007, four years into LinkedIn, Hoffman moved from CEO to chairman and president. Dan Nye took over, and Jeff Weiner soon followed and ran the company for eleven years, through the IPO, the economic-graph strategy, and the Microsoft sale. In 2011 the two men shared Ernst & Young's U.S. Entrepreneur of the Year award, and the shared award is a fair summary of the arrangement.
The move ran against Hoffman's own experience at PayPal, where he had held the operating title and lost it. Here he gave it up voluntarily, while the company was growing, to someone better suited to running a scaled organization. That freed Hoffman to do what he was best at: investing, writing, theorizing, and routing opportunities. The LA Times portrait of 2008, with its seven boards, sixty companies, and three computers, describes someone who could not have also run a fast-growing company's daily operations, and seemed to know it.
Tactic: Write down which parts of your company would get better if you stopped doing them, and if the list includes the daily operating cadence, start recruiting the operator before the board asks you to.
Principle 7
Be the router, not the endpoint
Hoffman's most consequential investment in Facebook was not his own check. It was the introduction that, according to David Kirkpatrick, brought Mark Zuckerberg to Peter Thiel and produced a $500,000 angel round. Hoffman invested alongside Thiel even though he was building a social network himself. The instinct to route an opportunity to the best-matched party instead of holding onto it is why Dave Goldberg called him "the person you want to talk to when you are starting a company."
He has a name for the underlying practice: network intelligence, reaching out to the smartest people you know to test what has to go right and what might go wrong. Greylock's biography puts the principle in the first person: "Networks and marketplaces are central to all of my investing and thinking." A router's value grows with the number of good connections that pass through it. A node that only collects loses its usefulness, because people stop sending things its way.
Tactic: Once a month, make one introduction that benefits two people and does not directly benefit you, and keep a log, because routing compounds the same way an investment does.
Principle 8
Sell the map, not the quarter
In February 2016 LinkedIn's stock lost 43.6 percent in one day and closed at $108.38. In June, Microsoft agreed to pay $196 a share. Earnings did not change in those four months. The audience did. To public markets, LinkedIn was a growth stock that had missed a forecast. To Microsoft, it was a map of the world's professional workforce, the "economic graph" Jeff Weiner had set out to build in 2012, and a natural complement to Office.
The lesson concerns which buyer can value what. A public market discounts a network by its next quarter. A strategic acquirer values it by what the graph makes possible across its own products. Hoffman also turned the exit into a seat: he joined Microsoft's board in March 2017, which later placed him at the center of Microsoft's AI partnerships and, eventually, its acquisition of Inflection's team.
Tactic: If you are building a network, keep a written thesis on which strategic buyer gains the most from your graph specifically, and update it every year, because that buyer's valuation will not track your stock price.
Principle 9
Write the manual for your own market
The Start-Up of You told readers to act as the "CEO of their own career" and build their professional networks, and the obvious place to do that was LinkedIn. The Alliance gave HR departments a vocabulary for the networked workplace. Blitzscaling, the free Stanford course, the Masters of Scale podcast that launched on May 3, 2017, the Masters of Scale book, and Possible all did the same thing in different settings. Hoffman wrote the instructions for the world his companies depended on.
This was not just marketing. It was category creation done in public, and it gave Hoffman an influence over how founders, recruiters, and policymakers thought that no balance sheet would have bought. The approach carried over into AI: Impromptu and Superagency are, among other things, arguments for adopting the technologies his portfolio builds. The method works because the books are useful, and it requires disclosure because the author benefits from them.
Tactic: Publish the framework you use to make decisions in your market, with a clear statement of your financial interest in it, because a widely adopted framework works like a distribution channel.
Principle 10
Ship the argument, iterate the ethics
Hoffman does not treat ethics as a gate to clear before launch. He treats it as something refined through use. "Innovation isn't just unsafe; it actually leads to safety," he told TechCrunch, pointing to brakes, airbags, and seat belts, which were developed as cars spread. He calls the approach "iterative deployment": put the tool into many hands, watch what happens, and adjust. He backs it with a philosophical claim, made on the HBR IdeaCast, that "one of the illusions that are sometimes promulgated is that technology is essentially value neutral." Values come with every design. The only question is whether you revise them in public.
His AI demonstrations follow the same logic. He wrote a book with GPT-4, translated a commencement speech into more than a hundred languages without human editing, and gave a televised interview to his own deepfake. Each one was a deliberate public test of an argument about what the tools could do.
A
Bloomer vs. Doomer: The Two Postures
How Hoffman's stance on AI differs from the precautionary one he argues against.
| Precautionary posture | Hoffman's posture |
|---|---|
| Pause frontier development until risks are understood. | Pause proposals are "foolish" and "anti-humanist." |
| Regulate first, deploy second. | Support "intelligent regulation," but rely on iterative deployment to steer outcomes. |
| Technology is a neutral tool that humans misuse. | Technology always carries values; design choices encode them. |
| Focus on what could go wrong. | "Everyone... focuses way too much on what could go wrong, and insufficiently on what could go right." |
| Biggest risk: runaway capability. | A risk he names that is less discussed: lock-in that entrenches incumbents and suppresses entrepreneurial dynamism. |
Tactic: For any product with social effects, set a fixed review interval after launch to examine who is being disadvantaged, and publish what you changed as a result.
Principle 11
Disclose the conflict, then argue anyway, or step back
Asked by The Guardian why anyone should listen to an AI optimist who profits from AI, Hoffman did not dodge. "An economic interest doesn't necessarily make what someone is saying wrong," he said, "and I try to be transparent and not hide mine." He then described his method: "I tend to start with my beliefs and follow with my money." When the conflicts became structural, he acted on them. On March 3, 2023, he resigned from OpenAI's board, citing the overlap with Greylock's AI investments and his role at Inflection.
The discipline is to name the conflict before your critics do, and to recognize the point at which naming it is no longer enough. Hoffman has clearly held to the first part. The second required giving up a board seat at the most important AI company in the world, which is the costliest form of credibility available.
Tactic: Keep a written list of every role whose interests could conflict with your public arguments, and set in advance the threshold at which you will resign one of them rather than merely disclose it.
Principle 12
Audit the edges of your network before someone else does
The trust model that made LinkedIn work, in which reputation passes through mutual connections, also applies to its founder. The Forbes cover line "Silicon Valley's Best-Connected Billionaire" became, in a 2023 proxy memo, the first piece of evidence against him. A fundraising relationship that Hoffman says was limited to MIT turned into Senate testimony, a presidential investigation order, and years of reputational damage. The 2017 Alabama disinformation project, which he says he did not know about, produced a public apology. In both cases Hoffman's explanation turned on not fully knowing what someone in his network was doing.
His later response, demanding full release of the Epstein files and paying for victims' calls for transparency, was consistent with his long-held view that visible networks behave better. It also came late. A network is judged by its worst connection, and by how long that connection was allowed to continue.
Tactic: Once a year, list the ten relationships and funded entities with the most reputational exposure to you, and for each one confirm you know what they are actually doing with your name and money.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Reid Hoffman, GreylockStarting a company is like jumping off a cliff and assembling a plane on the way down.
— Reid Hoffman, commencement address, Bologna Business School, September 8, 2023As an entrepreneur, investor and a perpetual philosophy student, I've always believed that any new venture must have a strong theory of human nature underlying it.
— Reid Hoffman, lecture at the London School of Economics, December 2024We shape our tools. Then our tools shape us.
— Reid Hoffman, HBR IdeaCast, April 2023One of the illusions that are sometimes promulgated is that technology is essentially value neutral.
— Reid Hoffman, to TechCrunch, January 2025Everyone, generally speaking, focuses way too much on what could go wrong, and insufficiently on what could go right.
Maxims
- Psychology before product. A venture is a hypothesis about human nature, and the product only tests it.
- Learn fast, don't fail fast. Failure is the cost of learning, not the goal, so go after your most dangerous assumption first.
- Titles are revocable; maps are not. When the org chart turns against you, take the job that touches the most outside parties.
- An invitation spends reputation. Every message sent in a member's name draws on their trust, so never send one they did not choose to send.
- The few pay for the many. Charge the minority for whom the network is a working tool, and let everyone else supply the value.
- Give the operator the keys. A founder's best contribution may come after stepping out of the daily cadence.
- Route, don't hoard. The introduction you make can be worth more than any check you write.
- Sell to the buyer who reads the map. Markets price your quarter, while strategic buyers price your graph.
- Disclose, then decide. Name your conflicts before critics do, and know in advance when disclosure is no longer enough.
- Your network is your record. Every connection lends you reputation and borrows from it, so audit the edges while you can still choose them.
