If I’ve learned anything from listening to Bing Gordon dispense timeless product-design wisdom in the form of deep-cuts Silicon Valley lore over the years, then most of you should correctly answer at least two or three of these questions, drawn from the pages of his new book, Everybody Wins, and 90 percent of you should believe you have a 90 percent chance of nailing them all.

1. What did Bing call the one-page customer retention idea he brought to Jeff Bezos in 2002?
A) BOOM! Shipping
B) Amazon Prime
C) Guild of the Frequently-Bought-Together Impulse Buy

2. What did Steve Jobs tell Zynga founder Mark Pincus was mathematically impossible?
A) That anyone would want to play a farming game on purpose
B) That a sad, droopy-eyed cow could get a 30% click-through rate
C) That users preferred cooperative game-play over a Top 10 spot on a leaderboard with high-agency drop-shadows

3. Exactly how many wins does it take the average person to emotionally erase a single loss?
A) One win, because humans are rational and definitely not dramatic
B) Ten wins, plus a badge, and someone from finance saying “Nice!” on Slack
C) About three wins, because losses hit harder than gains

4. What’s the surest way to get Bing to return your call quickly?
A) Gift him Teebu’s Blazing Longsword on World of Warcraft
B) Send him a Cameo from your new chief of staff, The Rizzler, asking to get 15 minutes on Zoom on the books
C) Text “Quick question” and then three-dot him for six hours

Few products on earth earn customer loyalty like video games do, and they do it by helping players win early and often. As the title of Bing's new book (co-written with his daughter Chloe) hints, true breakout hits go even further: They’re designed so that everybody wins.

If you enjoyed this foreward, you’ll enjoy Bing’s new book. I’d recommend purchasing a copy HERE!

Bing played a key role in turning the “everybody wins” mindset into an explicitly articulated, repeatable discipline. It started when he co-founded Electronic Arts in 1982, an era where computer memory was still so scarce that an early prototype of what would eventually become John Madden Football could only accommodate seven players per side. (”You’re not putting my name on that!” Coach exclaimed when he saw the demo…so not everyone won that day.) It continued as the gaming industry shifted from boxed floppy disks and cartridge-era consoles to CD-ROMs, online worlds, Facebook games, and the mobile live-service products that now dominate the industry.

By closely studying what actually motivated players, Bing developed the frameworks and processes that helped Electronic Arts reliably engineer both “wow moments” and “forever franchises.”

Just as consequentially, Bing also came to realize that none of the values and dynamics that made gamers so passionate about their favorite titles were unique to the medium. Which, of course, meant that the same techniques he applied to titles like John Madden Football, The Sims, and Need for Speed could also be applied anywhere–a thesis he eventually proved by helping companies as diverse as Amazon, Duolingo, Cameo, and MyFitnessPal, to name just a few, improve user motivation, engagement, and retention.

Today, “gamification,” “engagement,” and countless other buzzwords that constitute the lingua franca of the attention economy are often used to invoke the slot-machine pull of the infinite scroll and the low-key assault of ostensibly “helpful” notifications designed to induce Pavlovian compliance.

What Bing makes the case for here, however, is what I would describe as an authentic and pro-social articulation of gamification, meticulously arrived at through decades of hands-on game play, user feedback, and data-driven analysis. What people return to in the games that last, he declares, are three things: connection with other players, the feeling of getting better at something, and visible progress toward personally meaningful goals.

And the thing that makes this book so valuable is that Bing’s convictions to build for connection or engineer frequent wins are not just the pie-in-the-sky sentiments of some armchair idealist invoking the way things could be or should be. They’re field notes of the way things have been, are, and can continue to be–real case histories that document all the positive outcomes that compound when you truly try to align with your users’ genuinely held goals and values.

Which is not to say you can’t use leaderboards or devious loss-aversion hacks to manipulate people into engagement. Zynga’s decision to implement a crop-wither mechanic that compelled players to set their alarm clocks at 3AM to water imaginary strawberries worked like, uh, clockwork–right up until those players quit and expressed how relieved they were to escape such calculated strong-arming.

But what Bing ultimately makes clear here is how the internet, and the access-oriented business models that evolved on top of it, inherently reward long-term value creation and a truly customer-centric focus. In the brick-and-mortar days of the gaming industry, and many other industries as well, success was measured in infrequent transactions – a $50 CD-ROM purchase at Christmas time, and then, ideally again, a year or two later, when the sequel came out. Now, however, success depends on daily relationships. Free and freemium products must compete for user loyalty in very granular ways. If a game or productivity app stops creating value, users leave. If it violates trust, the churn eventually shows up in the data.

Bing’s argument here is that this dynamic, if properly understood and acted upon, doesn’t corrode our humanity. It demands that products serve it, incrementally, persistently, and accountably, for years on end. In this way, alignment with human values isn’t just the ethical path. It’s the socially beneficial, economically productive, and institutionally durable one.

That’s a gospel I whole-heartedly agree with, and why I find his book, and Bing himself, such an internet treasure. So I encourage you to first read Everybody Wins, and then put its wisdom into practice in whatever domain you’re playing in. And I hope you consider sending him a sword in World of Warcraft. Apparently, it works.

If you enjoyed this foreward, you’ll enjoy Bing’s new book. I’d recommend purchasing a copy HERE!

We called ourselves wise before knowing where we came from. In 1758, Swedish naturalist Carl Linnaeus gave us the name Homo sapiens, meaning “wise human.” Darwin’s Origin of Species and the discovery of Homo erectus fossils were more than a century away. Generations of researchers would revise our ancestry. We had already decided what made us remarkable.

We gave ourselves a name to grow into. It could be time to do that again.

The work we associate with intelligence is no longer just ours. In math, AI is pushing the research frontier, not just solving contest puzzles. Stanford’s AI biotech agents analyzed roughly 50,000 clinical trials in under a week and found patterns that predict which drugs will succeed. They also proposed a lung cancer therapy that a major drugmaker independently pursued and that later earned FDA breakthrough designation. In Nigeria, six weeks of teacher-guided AI tutoring yielded roughly 1.5 to 2 years of typical learning. And AI has helped paralyzed patients express speech and gestures at the same time through personalized digital avatars.

It’s time to rethink the idea that our intelligence alone makes us unique and valuable. Who should we become if simply being smartest no longer defines us?

The humans we already are

Philosophers, historians, and social scientists have offered many alternatives to Homo sapiens, with each highlighting a key part of human life. I see them as invitations to spotlight and develop qualities we already have. These identities describe what humans do—or why what we do makes us who we are.

Let’s start with Homo faber, the human as maker. In Creative Evolution, French philosopher Henri Bergson suggests that our ability to make tools—especially tools that make other tools—is what defines us. We chipped stones into blades that shaped wooden tools. We bored cylinders for steam engines that powered machines. We build AI that designs chips that power future AI.

I’ve made a related argument through Homo techne that we evolve through the tools we create. And when millions of people get access to new tools that dramatically increase their opportunities, that’s superagency. I put agency at the center because it’s a mindset, as well as a muscle. No tool hands it to you, and no tool can take it all away. There’s always some agency to claim, alone and together: you do more yourself, and you benefit from what others can do.

Tools may make us more capable, but there are plenty of reasons to stay hands-on: to achieve mastery, exercise judgment, and take responsibility for the result. A good life requires more than making things. We can become skilled builders without asking what deserves to be built. Untethered, Homo faber can produce slop, with or without AI. It needs companions.

One is Homo ludens: the playing human. Dutch historian Johan Huizinga argued that play helps form culture itself. In play, we voluntarily commit to rules and possibilities. We try, improvise, challenge, and discover what we can do.

The value of play does not depend on being the best. The existence of a better chess player (hello, AlphaZero, Leela Chess Zero, and Stockfish) doesn’t make chess with a friend pointless. Most of us do things others do better, for the pleasure of a surprising move, hard-won skill unlocks, or laughter after a mistake.

I’ve often discussed what games teach us about entrepreneurship. When I introduce friends to Catan, we play with everyone’s cards visible so we can learn together. But play doesn’t need to always prepare us for something. The game can be about time together, and about joining in.

That same spirit of presence and participation animates Homo aestheticus: the human who turns experience into expression. Evolutionary theorist Ellen Dissanayake describes the basic human impulse to “make special”: turning ordinary sounds, movements, materials, and occasions into something notable and meaningful. Think of setting a dinner table or composing a toast. Those turn what we feel and see into something others can share.

AI lowers the barrier for people to express themselves in ways that are new to them. Someone might explore animation to tell their story. Or try to recreate a feeling from childhood through vibecoding a video game. Or someone otherwise tone-deaf (ahem!) who seeks to express themselves through song. These are all paths to sharpen how and what we see. Wider access can bring more people into that process. More art means more personal expression that can be turned into shared experiences.

When we make something special and share it, that’s Homo festivus: the human who celebrates. In The Feast of Fools, American theologian Harvey Cox defended festivity and imagination against a culture organized too narrowly around success and productivity.

Celebration gives human values a spot on the calendar and adds a public dimension to play. A wedding recognizes a commitment. An anniversary keeps a memory alive. A neighborhood meal can turn strangers into friends.

Celebration also refines our vision of technological abundance. More free time does not automatically create community. Someone must extend the invite, prep the space, and welcome the newcomer. The occasion can be modest (a monthly potluck, a birthday walk, a recurring evening of music) but hospitality is a practice. If technology gives us more capacity, we should choose to spend some of it creating occasions people will remember together.

Homo festivus is closely related to Homo donator: the giving human. Working in the tradition of gift scholarship, sociologist Jacques T. Godbout puts giving at the heart of social relationships. A gift is never just itself. With it comes recognition, gratitude, and an invitation to stay connected.

I recently made a different kind of gift. For America’s 250th birthday, I used AI to help create an album and pressed it on vinyl. Side A recasts the Declaration, Frederick Douglass, the Gettysburg Address, and King’s “Dream.” Side B turns state mottos, amendments, and technological milestones into song. With few musical chops, I used AI to reinterpret this inheritance across musical genres.

The process made me wonder: what might we give if more kinds of creation were within reach? Making becomes generosity when it begins with close attention to someone else. Knowing what might delight, comfort, or help someone calls for involvement in their life.

Giving leads to a more demanding identity: Homo curans, the caring human. Philosopher Agustín Domingo Moratalla places care at the core of responsibility. Care ethics reminds us that no one becomes capable alone. Before we could build, compete, or create, someone cared for us. Our independence rests on the people and institutions that sustain it.

AI can help a family coordinate support for an aging parent, make information more accessible, or reduce the administrative burden on a caregiver. We should judge those gains by whether they protect people’s dignity and ability to shape their own lives.

Care also moves the argument beyond the sole individual. As American theorist Joan Tronto argues in Caring Democracy, the allocation of caring responsibilities is a political question. Who has time to care? Whose needs receive resources? Who carries the burden?

Playing, creating, gathering, and giving all require some security and access. Any vision of flourishing has to include the person working two jobs and the family managing illness. These values should guide education, working conditions, public spaces, and how technology’s benefits are shared. Care asks: whose possibilities we are expanding?

But care by itself doesn’t resolve the uncertainty ahead. Homo viator—the human as traveler or pilgrim—has a helpful perspective. In Homo Viator, French philosopher Gabriel Marcel presents a Christian vision of life sustained by hope and fidelity to others. I draw a broader lesson from it: life can remain meaningful even when we do not know the destination.

AI has and will continue to unsettle identities built over decades of work and expertise. Simply telling people to learn new skills won’t cut it. We need room to grieve, adjust our expectations, and find new commitments. But every traveler must choose a direction. Which brings me, finally, to Homo prospectus: the human who anticipates possible futures.

In Homo Prospectus, Martin Seligman, Peter Railton, Roy Baumeister, and Chandra Sripada argue that anticipating and evaluating possibilities is central to how humans think and act. Imagined futures can shape us as much as past experiences.

I share this identity last because the ones before it tell us what is worth imagining. A society built around care measures more than output. A school built around curiosity and creative participation asks different things from the technology we build. A community that values belonging judges progress partly by whether people have more chances to gather.

This is part of what Aria and I explore on Possible: making desirable futures concrete enough that people can begin working toward them. This takes more than optimism; it takes ambitious realism. We must ask what needs to change, who can help, and what we can try now.

AI helps us explore alternatives and examine consequences. When we do so, can we bring a richer account of what we hope to achieve? Prospection turns our values into choices we can test, revise, and pursue together.

Growing into the name

Which identity should guide us? None of them on their own. These ideals must weave together. Homo faber needs Homo curans to tell us what deserves to be built. Homo prospectus needs ludens, aestheticus, donator, and festivus to decide which futures are worth imagining. Together they describe a richer, sturdier account of being human. A life woven from many threads doesn’t unravel when a machine proves better at chess or math. Being outmatched at one thing isn’t the same as being outmoded. However capable the machines become, we’ll still find value and meaning the way we always have: by making, playing, creating, celebrating, giving, caring, journeying, and imagining.

The weave also gives us a practical test for AI. Does it help more people exercise judgment, deepen relationships, make, give, play, and care? Or does it raise output while leaving them spectators? Agency isn’t something technology takes or gives once and for all. We cede it to our tools or we take it from them, and we do so continually, with every system we build and use. Nearly half of American adults now use AI chatbots, up from one-third two years ago. This question cannot wait for superintelligence. We are already answering it.

Linnaeus gave us a name to grow into, and for a long time we assumed intelligence was how we’d grow into it. It isn’t. Intelligence is now something we can build. Wisdom is knowing what intelligence is for and whom it serves. If we integrate AI in ways that help us become more fully human, we can earn the name we’ve carried since 1758—not Homo sapiens as a species label, but Homo sapiens as an ideal: beings wise enough to decide which possibilities deserve to become real.

It is no surprise that Americans are wary of AI. Most Americans work hard and still struggle over things they shouldn’t: putting gas in their car, paying their electric bills, going to the doctor, saving enough for retirement. Even the modest pleasures that make hard work feel worthwhile feel increasingly out of reach. And, while Americans know that innovative technologies can create businesses, generate jobs, and grow the economy, too many Americans are still waiting for that prosperity to reach their families. They’ve also experienced how new tech can cause real harm (ask any parent who’s watched their kid struggle with the mental health effects of social media).

So while AI will transform our economy and the world around us, Americans are right to demand that this transformation serve their needs, not just the interests of the already rich and powerful.

Tech leaders have been warning the public about the potential for massive job loss in the next handful of years. Researchers from both OpenAI and Anthropic have resigned in protest, fearing that their employers were failing to manage the existential risk of an AI-driven human extinction event. When the people building this technology say your job, your children’s future, and maybe your life are on the line, they cannot then ask the public to welcome AI on faith.

To be clear: There is genuine existential risk –– “X-risk” –– here, and it is genuinely complicated. Many experts are wrestling with X-risk in good faith, debating independent safety evaluations, government oversight, and international cooperation. These different proposals raise a common question: how do we protect the public while ensuring that the power of AI is broadly shared?

Dario Amodei’s recent call to slow advances in the most powerful AI models, embed independent evaluators inside labs, and coordinate government safety requirements is a serious attempt to reconcile technological leadership with safety. Sam Altman’s agreement that pacing development and independent evaluators are needed is another step in the right direction, and shows that collaboration is possible.

While I applaud these ideas and their commitments to them, I also believe they’re missing something critical. Even a workable agreement on the pace of development leaves unanswered how the benefits will reach ordinary Americans. If companies and governments can negotiate the conditions under which AI is safe enough to advance, they can also negotiate how that progress will make everyday Americans’ lives better.

This American generation is faced with the herculean tasks of obtaining economic prosperity for the many, and securing American AI leadership in the new world order. All this must be done while ensuring the safety of the American people. This is the necessary dual mandate, not just for one industry or economic sector, but for America at large.

As AI technology accelerates development and economic precarity increases, the decisions we make in the next 36 months are crucial; we are setting a course that will determine the winners and losers, both domestically and internationally.

In the Industrial Revolution, societies that harnessed new industrial technology became global powers in the centuries that followed. The societies that could or did not harness new technology were forced to live with the consequences of decisions made elsewhere. AI will sort the world the same way: those who harness it effectively will gain cultural and economic power, and those who don’t will become subject to forces they do not control.

Too much of the conversation about what to do with AI has been dominated by two polarized camps: the extreme accelerationists and the extreme decelerationists. Neither prescription is tethered to both the outcomes that matter: whether Americans’ lives get better, and whether America stays secure in the new world order.

The extreme accelerationist case is that speed is safety, and any friction is untenable: whoever builds the most capable systems sets the rules, and because American values are better than the alternatives, we should build whatever we can, as fast as we can. Accelerationists concede that while there is risk and while economic transition costs will be significant, we must take those risks and pay those costs to reach the full benefits of AI. President Donald Trump has downplayed warnings about AI’s risks and calls to slow its development, arguing that America cannot afford to lose its lead over China.

But accelerationists rarely volunteer to open their own checkbook, or let their own children struggle to find their way in a new world. Their position offers little to a worker displaced in 2027, or to families who cannot afford the basics right now, or to college graduates facing an unforgiving economy beyond the promise that abundance arrives…eventually.

The extreme decelerationist case is that we are building something we don’t understand, and we should stop all development until we do. It calls for a moratorium on AI. Some call for coordinated global bodies governing what gets built and who gets to build it. That vision depends on unrealistic, global cooperation that would only exist in a more utopian world. An absolute moratorium on AI development by the well-intentioned shifts technological power toward less accountable hands, giving bad actors and hostile nations, like China, greater influence over systems Americans will increasingly depend on.

The extreme accelerationist and decelerationist camps have shown the limitations of their positions. Take, for example, the debates over data centers. The accelerationist position is simple: build, everywhere, now. Public engagement, regulation, and permitting are the enemy; local objections are NIMBYism; any delay is a gift to Beijing. They argue that the grid will catch up to the strain and that the increase in electricity costs -– up 20% because of data centers in places like New Jersey -– has nothing to do with data centers.

When accelerationist tech leaders explain that AI will take jobs and then show up to a town with a request for a thirty-year tax abatement, an NDA, and a plan that leaves ratepayers footing the bill for new transmission lines, that’s a deal people are right to reject.

The decelerationist position is the mirror image. They describe data centers as purely extractive: consuming enormous amounts of power and water, pushing up electricity bills for the families living next door, delivering few permanent jobs, all to create profits for trillion-dollar companies. Their chief answer is moratoria, and where that doesn’t work, they’ll go with restrictive zoning, and hearings and other bureaucratic processes designed to run out the clock. Reflexively blocking a project without negotiating forfeits any possible benefits to a community, from new tax revenue, new jobs, and other potential investments –- and -– it just moves the compute somewhere else, potentially to a country with no interest in American security or values.

There is an ambitious and realistic way to address the data center problem, one that would satisfy our dual mandate. Data centers can be opportunities to advance the technology and deliver real, immediate benefits to everyday Americans. And the communities that will live alongside a data center for decades should decide whether it gets built and on what terms. When constituents want a data center, the deal should be negotiated in public by local governments using maximum leverage to negotiate with corporations on the public’s behalf. People are right to expect benefits. From stable grids to free markets, we’ve given the labs and hyperscalers every reason to want to build in America. Now it’s time for the public to present its conditions. Let’s not walk away from the negotiating table when we have real leverage to win material benefits.

Good agreements can mandate that utility bills go down, not up, and that the new property tax revenue these facilities generate lowers the tax burden on local families. Agreements can also include commitments to local hiring and union apprenticeship programs during construction, funded training for the permanent operations roles, and investment in local schools and infrastructure. They can make companies pay for necessary grid or road upgrades and any new generation their facilities require, so that ratepayers see bills decreasing rather than subsidizing hyperscalers. Environmental commitments (closed-loop cooling, water-use caps, clean power procurement), can be conditions of approval, enforced with penalties.

The communities that will live alongside a data center for decades should decide whether it gets built and on what terms. I feel strongly that the decision to build or not build belongs to the people who will bear the costs and share in the benefits, not tech companies or politicians beholden to lobbyists and donors.

More broadly, the public deserves the benefits of AI as soon as there are any, not a generation from now after a hard transition. The government should strike a bargain with the frontier labs, in favor of its constituents: approval for the infrastructure needed to build advanced AI in exchange for concrete benefits for Americans. Imagine three agents available free to every American—a medical assistant, a legal assistant, and a personal tutor. Governments can work hand in hand with labs to ensure these tools are built safely and equitably. People should be able to get help understanding a diagnosis, navigating a dispute with a landlord, or mastering algebra regardless of their income or ZIP code. Imagine support for AI job transitions -– financial help, reskilling, and guaranteed new work. Imagine a fully funded Social Security program. Imagine teachers with enough help and tools to support every student. This is a time for ambitious realism - concrete benefits, with a safety and human-first lens, could make AI worth it for all Americans.

A country that experiences AI only as a threat will -– should –- demand that someone hit the kill switch, but a country whose citizens could have measurably better lives has a reason to insist that America gets it right for them, and for their children.

Data centers, AI public goods, and debates about AI safety are opportunities for Americans to shape how AI takes root in their communities. We will repeatedly face similar choices over energy, over jobs, over education, over how the gains from AI are distributed and who bears the costs. Instead of extreme positions, we must focus on our dual mandate, and deliver outcomes that make Americans’ lives better, and make America more secure.

The nations that leveraged new technology and created broad prosperity emerged as global powers in the Industrial Revolution. We can learn from history to make this technological transition easier for people, and ensure that everyone benefits sooner. Even with pacing its development, this technology is moving fast - the next 36 months will be determinative; we need an engaged public and leaders who understand and can deliver on the dual American mandate of economic prosperity and global AI leadership. That is the task in front of us now.

We should be ambitious enough to build this technology, real about what’s needed to get the benefits to everyday people, and humble enough to accept that no one or two companies get to decide how to manage the consequences of this transition.

America should lead because we have the innovation and realism that can make that bargain work. That’s been our history –- we can make it our future, as well.

Twenty-six years earlier, in the opening pages of Nonzero, the book that introduced me to Bob’s work, he’d asked similar questions: “Our species seems to face a kind of test toward which basic forces of history have been moving us for millenia. It is a test of political imagination—of our ability to accept basic, necessary changes in structures of governance—but also a test of moral imagination. So how will we do on this test?”

Now, of course, there’s much to navigate with AI. In Bob’s view, this vastly increases the urgency of the test we face, along with the chances of a failing grade. In Nonzero, he admitted to the possibilities of annihilation—”On the other hand, we could blow up the world”—but ultimately seemed far less invested in that proposition than by how technological innovation reliably compels new levels of human accord.

In The God Test, the reverse is true. We’re heading somewhere very fast, Bob observes, including, perhaps, “dystopian hellscapes,” “terrifying chaos,” and “brutal authoritarianism.” And those are some of the less-bad options. We could also be “enslaved or slaughtered by a pitiless superintelligence…”

Which is not to say that Bob has suddenly shifted from Pollyanna to full-blown doomer. In characteristic fashion, he arrives at his positions with a genial, thorough, and nuanced touch, surfacing novel connections across different disciplines and grounding sweeping pronouncements in telling and entertaining details.

In addition, a dark shadow has always lain at the heart of his relatively sunny outlook. Part of what makes Nonzero and what Bob calls nonzero-sum-ness so persuasive is the pessimistic take on human nature that informs it. Using game theory as his lens, Bob makes the case that self-interest has always driven human beings to seek out non-zero-sum “games,” where the outcomes of players overlap, over zero-sum games, where there are clear winners and losers.

Whether it’s hunting together versus hunting alone, or inventing knowledge-sharing technologies like the printing press, humanity has consistently steered toward social arrangements and technologies that make cooperation the safest, most rational choice for all players. Not out of the goodness of anyone’s heart, Bob emphasizes. And not exclusively, or with uniformly pro-social results. Zero-sum games persist. Non-zero-sum games can lead to lose-lose outcomes. They can produce externalities that are not uniformly beneficial.

Overall, though, Bob’s long-standing contention is that non-zero-sumness, the web of win-win interactions that accrue over time, creates net-positive gains for humanity, with a kind of ongoing flywheel effect. Consider how the invention of the printing press led to more books, which led to faster knowledge diffusion, which led to more inventions. New inventions created new incentives for more complex social structures and institutions that could fully capitalize on them, which in turn led to still more inventions. In this way, humanity went from hunter-gatherer villages to cities to nation-states, from local telegraph systems to the global internet.

At the same time, one persisting effect of this dynamic is that continuously growing social and technological complexity also introduces new kinds of risk. Global aviation gave us Paris vacations and fresh sushi from Tokyo, and also faster-spreading pandemics. Global supply chains make goods cheaper, but also more vulnerable to events that happen halfway around the world. A world that can produce automobiles, skyscrapers, and nuclear weapons at scale is also a world that eventually has a pressing need for international weapons treaties and climate change policies.

The internet enabled superpowers that billions of people now consider default essentials for daily living, but it also empowered bad actors of all kinds and put new strains on social cohesion, shared truth, privacy, and adolescent development. A technology that reshapes how billions of people work, learn, and relate needs continuous assessments, refinement, and governance.

Cooperation, in other words, typically demands more cooperation—from parties who aren’t always inclined to extend it, especially if previous non-zero-sum wins have given them new powers to explore new ambitions. We did, after all, get World War I and World War II along the way to a more tightly connected planet. If, as Bob asserts in Nonzero, non-zero-sumness expands the “circle of moral consideration,” it also expands the blast radius of lethal, zero-sum aggression.

Now, enter AI.

The idea that what Bob calls the “arrow of human history” is pointing toward “some kind of global cognitive collaboration, toward a larger cooperative enterprise than our species have ever mustered before” is a central theme of his theory of nonzero-sumness. Along with the internet, AI naturally plays into such scenarios. In fact, it’s pretty clear that AI is the only technology that makes a highly complex but still somehow coordinated planet even thinkable. For example, you’d need instant AI translation to make the network’s edges mutually intelligible. You’d need AI to metabolize a noisy planet’s increasingly massive information flows into something a decision-making layer could act on—forecasting, modeling, and surfacing signal from a volume of data no human institution could parse.

But where “global cognitive collaboration” was an aspirational endpoint in Nonzero, the ostensibly better place we’d get to if we passed the test (and “didn’t blow up the world”), “global cognitive collaboration” plays two roles in The God Test. On the one hand, it still might be the promised land. On the other, it could be the thing that leads to our downfall.

What accounts for this shift? In part, Bob is worried that AI is getting better and we’re not. “In the most plausible [AI] takeover scenarios, the basic impetus….is us: human beings acting the way human beings have been known to act,” he writes.

Here, he’s not even proposing anything particularly diabolical. It’s just that, in his estimation, users will, in characteristic human fashion, want AIs that can bend the truth, because sometimes that’s what it takes to protect and pursue one’s interests.

Life is full of zero-sum games, or at least games that can be played in zero-sum fashion. Think salary negotiations, custody disputes, college applications, and insurance claims, to name just a few. When we retain human agents of various kinds to represent and/or guide us on these and other important matters, Bob notes, we generally want someone who will unequivocally advocate on our behalf. And often what’s in our interests, or what we believe to be in our interests, is not what’s truest, or fairest, or based on a dispassionate analysis of all available data.

So in Bob’s estimation, people will naturally want AIs that are highly adept at deception, flattery, and power-seeking, because such traits will make them more effective proxies for us. And even if only some people want this, an arms race dynamic will kick in. If your scrupulously ethical AI, which always tries to act honorably and consider a counterparty’s interests, is going up against one that ingeniously misrepresents the truth when necessary and deftly cuts ethical corners, which one is more likely to enjoy an advantage, especially over time? Which one will a majority of people choose to use? Which one will profit-seeking developers develop?

However problematic these dynamics may be in the short term, they grow even more alarming over time. “....When it comes to evolution, the question isn’t just what traits are possible,” Bob writes. “It’s what traits will get selected—which mutations will survive and proliferate.”

At this point, the Black Mirror episode begins to write itself. AIs that are growing more autonomous, increasingly proficient across multiple dimensions, and more embodied in various physical manifestations are also being selectively bred to be deceptive, manipulative, and power-hungry. It is, no doubt, a wicked problem, and ultimately what makes The God Test a novel and essential contribution to current AI discourse. In Bob’s estimation, the real alignment challenge we face is not that we won’t be able to get AI to do what we want, but that we will.

I also appreciate The God Test’s emphasis on speed and scale. Development continues at a rapid pace, and this creates increasing strains as the stakes grow higher. AI is also both an extremely general-purpose technology and one that benefits from what is perhaps the greatest non-zero-sum technological triumph ever, the internet. As much impact as, say, the Ford Model T had, imagine if it had been distributed globally the instant it rolled off the line—for free—into a world already fully outfitted with all the infrastructure it would require (highways, gas stations, traffic signals, etc.) and all the downstream innovations it would enable (drive-ins, motels, suburbs). Then picture those cars coding, translating, generating videos, and doing countless other tasks along with driving. AI isn’t just another technological wave. It’s an intelligence layer dropped directly onto a hyper-connected, pre-existing global infrastructure that took massive amounts of mutually interdependent coordination to build.

In the face of all this, Bob concludes that “slower AI progress has some big advantages over faster and faster progress.” In particular, he’s concerned about global competition and the potential negative consequences of that, including, say, a country trying to pre-empt another on the verge of attaining superintelligence by launching a nuclear attack.

Instead of ratcheting up international arms races, he suggests, we should be pursuing treaties and alliances and other kinds of non-zero mutuality. A core tenet of both Nonzero and
The God Test is that non-zero-sumness pushes human societies toward more complex and overarching forms of governance—first the city, then the nation, then the globe. In Nonzero, Wright referenced the work of Pierre Teilhard de Chardin, a Jesuit paleontologist who observed in the 1920s that early communications networks were already connecting individual human minds into what he dubbed the “noosphere,” i.e., a planetary layer of collective human thought and consciousness that would, through increasing interconnection and convergence, lead to more coordinated global behaviors and actions. In The God Test, Bob notes how AI might accelerate the onset of what the influential Oxford philosopher Nick Bostrom calls a “singleton,” a world order with a single decision-making agency at the highest level, with the power to preempt any rival.

Clearly the prospect of a singleton is massively concerning. Get it wrong, and you could have an authoritarian government, forever. Get it less wrong, and you’d still have some form of governance, forever, with the power to pre-empt any rival. What is ultimately best for humanity, in my view, is broad human freedom to pursue distinct paths and values within a framework of non-zero-sum cooperation on the problems that require it, and national sovereignty everywhere else. So I agree with Bob that prudence is essential and that higher levels of international cooperation are worth pursuing. But I also strongly believe that leading in AI development is the most prudent path to international cooperation.

After all, as Bob’s own books make clear, non-zero-sum thinking takes many paths. Trade and barter often arise spontaneously, because the benefits are clear to all parties. Governance, especially when it manifests as verified constraints, enforceable commitments, and monitoring regimes, is another matter entirely. In game-theory terms, cooperation among rivals is stable only when defection is deterred, and deterrence often requires exactly the capability advantage that Bob diagnoses as the disease. Every rung on the ladder he charts in Nonzero was underwritten by power. Treaties hold because someone can punish the party that breaks them. Even democracies use force to cohere; tax compliance is ‘voluntary’ only because enforcement makes it so. Similarly, the postwar liberal order, the very set of institutions Bob thinks we need to scale even further to ensure positive AI outcomes, has held for as long as it has largely because Pax Americana backstops it.

I also think Bob discounts the degree of non-zero-sumness that informs the space already. Certainly, we see a lot of “arms race” talk, and also explicitly zero-sum measures, like U.S. export restrictions on advanced semiconductors and fabrication equipment, applied to China. But it’s also true that today’s frontier models are trained on chips designed in California, etched by Dutch machines, and fabricated in Taiwan. Then they’re deployed across global telecommunications networks to billions of people equipped with smartphones more powerful than the mainframes that sent Apollo 11 to the moon.

Built on shared protocols and open standards, the internet itself persists as a noospheric triumph, attaining almost universal global opt-in—a substrate that remains open to all, on their own terms, even when the platforms built on it have not. AI maintains this pattern, in part because the technology itself and the culture surrounding it are comparatively diffusive by nature. So far, at least, frontier capabilities are replicated quickly, and often open-sourced, typically within months, not decades. Benchmark testing and scholarly publishing are well-established as knowledge-sharing norms. No doubt we could, and should, have even more transparency and accountability than we have now. But compared to, say, the Manhattan Project, AI development has proceeded in an open and collaborative fashion.

That’s likely to continue, because AI development isn’t containable behind a single choke point. It rides the same open protocols, the same cloud infrastructure, and the same global talent market that has made the internet era so non-zero-sum in the first place. A country or lab that “wins” a given capability race still has to recruit and maintain customers, publish papers to attract talent, and train on data built by the whole world. AI value arises primarily through its integration into a global system of great social complexity and mutual interdependence, with many different players playing key roles in the system’s ongoing evolution. Even chip export controls end up reinforcing AI development’s status as a non-zero-sum phenomenon: You wouldn’t need to create policy-driven hardware bottlenecks for a technology that wasn’t diffusing rapidly on its own.

So will we pass the test that AI’s presenting us now? Like Bob, I believe the stakes are high here. The risks are real. But I also think the odds are in our favor, in part because the elements Bob cites as the factors that make AI exceptional, also make it hard to monopolize. Between its speed, its scale, and its vast general-purpose utility, AI is generating new points of leverage faster than any single actor can consolidate them.

A singleton needs at least one chokepoint it controls completely, and ideally, multiple ones. Breach one, and change becomes possible. Of course, the wild card here is recursion: If AI begins to recursively self-improve—each generation of the technology helping design the next, in a feedback loop tight and fast enough to produce a “hard takeoff”—an ASI-equipped frontrunner could potentially engineer its way out of the dependencies that constrain today’s major AI players.

This is a real concern, but it still assumes a leap from “decisive lead” to “unassailable lead” that effectively happens in the blink of an eye. Or to put it another way, a runaway leader wouldn’t just need to out-think everyone else. It would also need to out-coerce, out-manipulate, and out-operationalize everybody else, all at once, in ways that capture or bypass everything they supply, including chips, compute, cloud infrastructure, and talent, all while assuming that none of those actors notice, react, or compete back in the meantime.

And nothing we’ve seen so far suggests capability accrues in that kind of discrete jump rather than the diffusive, catch-up-prone pattern that’s held for every AI generation to date, especially given those knowledge-sharing norms I mentioned above. Consider just one data point: in 2017, when many observers believed that Google had already obtained a kind of de facto singleton status given its leading position in search, eight researchers then at Google, including an undergraduate intern, published, with the company’s consent, “Attention Is All You Need,” the paper that gave the world the transformer architecture, and with it every serious competitor Google now faces. The diffusion took a more granular path as well. Within six years of the paper’s publication, five of its authors had left to found AI companies of their own—Cohere, Character.AI, Adept, and Sakana among them.

So even if America’s frontier labs no longer publish their architectures so freely, that post-transformer exodus matters more. Researchers leave and take their knowledge and methods with them. Capabilities get reverse-engineered and replicated from outputs within months of release. And unilateral secrecy is only as strong as its weakest participant. As long as one major lab keeps releasing frontier or near-frontier capabilities openly, those capabilities are in everyone’s hands—and the labs that matter sit in different countries, under no common authority that could make them all stop. This diffusion dynamic shows up in the data: according to the non-profit research group Epoch AI, the gap between closed frontier models and open-weight models is just four months.

Still, we should be paying close attention to the concerns Bob describes so lucidly in The God Test, especially as developers around the globe push frontier models toward new levels of autonomy and capability. In this context, speed does compound risk. We should pursue more coordination, not less.

That’s why, ever since deep learning breakthroughs like AlexNet and DeepMind’s Atari-playing agents began to make it clear something categorically new, with world-changing implications, was underway, I’ve been doing everything I can to work with and support leaders, labs, and companies with humanist aspirations. To this end, I was one of the first funders of OpenAI and its quest to develop AI as “an extension of individual human wills” with broad and accessible distribution. With DeepMind co-founder and current CEO of Microsoft AI, I co-founded Inflection AI, a pioneering effort to develop models with emotional intelligence along with raw capabilities. More recently, I co-founded Manas AI with oncologist and Pulitzer Prize-winning author Dr. Siddhartha Mukherjee and Ujjwal Singh, to accelerate discovery of new life-saving medicines.

In these and similar endeavors, my goal is always the same: develop AI models and systems that work for us and with us, rather than on us. And develop them prudently, but also with a real bias to action—because, as I’ve said in Superagency and elsewhere, we’ll never get the future we want simply by prohibiting the future we don’t want.

In 2026, as the stakes are rising, I find this to be truer than ever. Active and ambitious participation is more likely to lead to greater coordination than a more neutral or cautious posture is. Ceding the knowledge, incentives, and power that come with maintaining a leading position in AI development offers no guarantee that development will in fact decelerate across the board. Instead, it would most likely just hand the ball, or the chess piece, or the dice—insert your preferred metaphor here—to some other player. With so many opportunities and challenges ahead of us, I believe humanity will achieve the best possible outcomes if players with non-zero-sum mindsets and a people-first focus are assertively playing the game.

For many kids, organized sports offer their earliest lessons in teamwork, practice, and having a theory of the game. For me, it was Dungeons & Dragons.

In 1979, when I was twelve years old, it introduced me to a way of thinking about how individuals interact with each other that continues to inform my life and career. Its lessons would eventually even help me conceptualize LinkedIn. What I learned from D&D steered me toward making early investments in Facebook and Airbnb. It shapes how I think about AI. It’s the reason I advocate for systems that put genuine capability in the hands of individuals — not just government actors or a handful of corporations — but people, acting for themselves and for each other.

With its polyhedral dice and a rulebook that reads as if it were written by a government tax attorney in between bouts of jousting at the Renaissance Faire, D&D is nothing if not complex. But it’s also extremely vivid in how it mechanizes its philosophy of the world. So much so that even a somewhat quiet 12-year-old can begin to decipher the nuances of social literacy.

What, then, did I actually learn in between fending off goblins and making hard decisions about whether or not to choose a 50-foot coil of rope over an extra sack of gold? Four key principles:

1. You have to have a theory of the game

2. A shared mission is a form of network architecture.

3. Learning is a competitive strategy.

4. Every person on the team needs to be the hero of their own story, but those stories should cohere into a shared quest.

Together these principles inform a larger question that has given me a sense of enduring purpose, both personally and in my career: How do individuals become their best selves through their relationships with others? What Dungeons & Dragons helped me recognize, long before I ever considered a career in entrepreneurship, is that effective leaders, and thus effective organizations, spend more time designing roles than writing rules.

That lesson feels newly urgent. AI is unsettling our assumptions about how work gets done and who does it — which raises the stakes on the questions D&D taught me to ask.

YOU HAVE TO HAVE A THEORY OF THE GAME
In many games, the win conditions are stated before anyone takes the playing field or rolls the dice. Have more points on the board than the other team when the clock runs out. Bankrupt every other player. Trap your opponent’s king. The conditions are fixed, identical for everyone, and terminal: as soon as they’re satisfied, it’s game over. With Dungeons & Dragons, none of that’s true. Its rulebook is meticulous about how the world works. It tells you, at length, how far you can move or what a lock costs to pick. It says virtually nothing about what you should be trying to accomplish within the context of the universe it creates.

What D&D has instead are success conditions, nested at every scale. A skill check succeeds or fails. An encounter is survived. A quest is completed. But completing the quest doesn’t end the game. It ends that arc, and then you’re still at the table, with the option to take on a harder challenge if everyone wants to keep playing. At the same time, D&D does have explicit loss conditions, at least at the character level. Your character dies, or everyone in your part dies, and that thread is over, forever. But while individual characters can be knocked out permanently, no victory is permanent in the same way. Which, it turns out, is a far better model of startups, careers, and institutions than what most other games offer.

The deeper difference between D&D and many other games is who sets the conditions. In chess or basketball, they’re handed to you. D&D’s get authored mid-play — by a Dungeon Master responding to what the players actually do, and by the players deciding which of the world’s many problems is theirs. Nothing in the rulebook says it’s important to come to the aid of a fallen prince. That’s something a party decides to do or not do, based on their own prerogatives about what they think is worth doing.

So having a theory of the game in D&D is a major part of the game itself. In the absence of an assigned objective, you construct one, which means forming a hypothesis about what you in this world prefer as rewards, acting on that choice, and finding out whether you were right.

Once you’re constructing objectives yourself, you start realizing the importance of strategy and tactics and the relationship between them. Your theory of the game tells you what the world rewards. Strategy is the standing bet you place on that theory; it informs the kind of character you decide to build, and the team you assemble, and the kinds of problems you go looking for. Tactics, in contrast, are what you do at the bridge with the bandits on it.

And ultimately each informs the other. Strategy tells you what to look for, and what you find tells you whether the strategy was any good. For example, losing a fight you shouldn’t have been in is a strategic error that merely presents as a tactical one. If your theory of the game is wrong, you can execute flawlessly and still fail repeatedly. And, of course, sometimes the key lesson is that you were lucky or unlucky.

On the other hand, occasionally a tactic works so well that it stops being a tactic. You improvise something, or spot a quirk about how the rules interact in a given situation, and it produces results so far out of proportion to its cost that the correct response is to reorganize everything around it. Your strategy becomes: do more of that, and find out how far it goes.

This is why the theory has to stay current. In my case, my theory of the game started evolving as I began to more fully appreciate that the Dungeon Master is a person, not an algorithm. The rules are real, but they’re adjudicated by someone with their own sense of the story, sitting at a table with four other people who want the evening to be interesting. In D&D, any theory that accounts only for the printed mechanics is missing most of the actual system.

Generalize from that, and you get what I mean by a theory of the game: an explicit model of how the environment you’re operating in converts effort into outcomes, and — because no one is going to hand you a win condition — an explicit claim about what would count as success in the first place. What does this environment reward, as opposed to merely permit? Who are the real actors, and what does each of them want? Which constraints are genuinely fixed, and which is everyone just assuming? A good theory is concrete enough to be wrong; that’s one of the things that makes it valuable.

Most people, in most organizations, are running on a theory they never chose. In many cases, they inherited it from a manager, who inherited it from another manager. Or even more broadly, it’s an industry norm, a story about how careers worked in, say, 1955. Work hard, deliver on your assignments, and the right people will notice is a truism — but it’s also a theory. It’s just an implicit, generally unexamined one, which hasn’t been updated or revised, even though the evidence says it should. Instead, it’s just gotten less effective even as the world has changed.

That’s why keeping your theory of the game current is the core part of the job in entrepreneurship and investing. Every company I’ve backed represents a bet on a specific claim about how a market will behave that the market does not yet agree with. The point isn’t to be contrarian for its own sake — that’s just a different way of not thinking. It’s to hold a differentiated model and the evidence for it, and to keep pressure-testing both.

A theory of the game is also the prerequisite for everything that follows. You can’t coordinate a group around a shared mission if each person is playing a different game in their head.

A SHARED MISSION IS A FORM OF NETWORK ARCHITECTURE
A deep sense of purpose is essential to most people; we all want to feel that our individual efforts contribute to something meaningful, lasting, and ideally, larger than ourselves. More broadly, the power of explicitly shared purpose is what animates team sports, political movements, social activism, and, in microcosm, D&D and similar board games.

In Monopoly, you are fundamentally a shark, actively trying to make your friends go bankrupt so you can seize their assets. In poker, you’re an island, keeping your own assets and resources (aka your hand) literally close to your chest while trying to uncover the vulnerabilities of everyone else at the table. One of the reasons I was fortunate to discover D&D early is that it was one of the few games, at least among those I had come across, designed as a team sport. In D&D, you are fundamentally a teammate. You have your individual story, but you are also part of a team. That’s the premise, not a strategy.

In business, we use profitability as a proxy for shared purpose, and it’s one reason capitalism is such an effective platform for organizing human labor and ingenuity and achieving tangible progress at scale. But in the end, “we all love money” is still just a hack, transactional rather than transformational, a shortcut that substitutes external rewards for subjective purpose. In my own experience, I’ve found that “missionary” companies are generally even better than merely “mercenary” ones at operating effectively, growing rapidly, and adapting quickly. Which means they also tend to be better at generating revenues and profits over time.

Why? A dramatic, specifically articulated mission, one that goes beyond making money, gives an organization a unique identity and purpose, and thus every person in that organization a unique collective identity and purpose. Sometimes, a company’s “mission” is written off as “ethics theater”, a way to fill white space on annual reports. But thinking that way is either a lost opportunity or an admission of failure. A mission is only “theater” if it is drafted poorly or operationalized poorly. In such cases, it’s just words, not aspirations and values employees genuinely believe in, or work to make true.

In contrast, a well-defined “mission” is a deeply held ethics argument, and just as importantly, it’s a network and operations argument. A compelling mission gives every person in an organization the same north star, which means that decisions made at every level — about product, culture, hiring, strategy — are more likely to be coherent with each other. It’s coordination infrastructure.

In D&D, this concept is expressed through the quest—the singular, high-stakes objective that transforms individual players (who might in fact be strangers) into a unified party. In one instance of the game, you might be escorting a fallen prince to his throne. In another, preventing a cult from summoning an ancient deity. In every instance, the specific quest becomes the filter for every decision. If the mission is to stop a plague from leveling a city, you don’t waste time haggling over the price of a room at the inn or getting distracted by side-quests. The urgency and clarity of the mission create instant coherence.

When I was an undergraduate at Stanford University, I founded the Symbolic Systems Forum, a student society that brought technology industry figures onto
campus for weekly fireside conversations. While I wouldn’t have described it this way then, there were basically three stakeholders in this process, all with different incentives. First there were the students, who wanted access to accomplished experts in their potential career domains. Then there were the experts, who had limited time and wanted to spend it where they could apply it most valuably. Finally, there was Stanford’s faculty, who had their own ideas about who warranted an invitation, and what kinds of topics and discourse students should be engaging in.

What I recognized pretty quickly was that, even as the founder/leader of the group that was organizing the events, my authority was limited. I couldn’t simply ask a busy executive to show up, and I couldn’t force a tenured professor to approve a speaker they found “intellectually thin.” There was also no guarantee that other students would show up too. Instead, it was a matter of network coordination: How could I help every participant see themselves as actors with specific purposes in a shared frame? The way I saw it, our collective mission was to bridge the gap between the classroom and the “real world” for the next generation of builders.

Once I started emphasizing that frame, the friction didn’t disappear entirely. But it did help the various stakeholders better conceptualize their roles – including seeing how this frame aligned with their own goals and incentives – within the shared mission. And once that template was set, enough people understood themselves to be part of something with growing value that a kind of mini network effect took hold.

Faculty that spent less time gatekeeping and more time mentoring attracted more students. As more students showed up, experts stopped viewing the forum as a chore and saw it as a talent pipeline. Students, in turn, realized that informal access to faculty and experts gave them the opportunity to become more active architects of their own education. As the value of participation compounded for all participants, more students, professors, and experts wanted to participate.

When I co-founded LinkedIn in 2002, I put similar ideas–and ideals–about our company’s shared mission into play. Since our goal was to create the world’s leading platform for helping individuals transform their economic lives through professional relationships, we added the following question to our job interviews: “What’s the job you want after LinkedIn?”

Asking candidates about their exit strategy for a job they haven’t even accepted yet may seem like a pretty questionable recruiting tactic. (”You’re saying I should be thinking about the divorce before you’ve even proposed?”)

In our case, though, we recognized that as a global platform helping people advance their careers across all industries, we couldn’t be credible if we treated our own employees as assets to be guarded and retained, rather than as individuals with their own goals and aspirations, with whom we were in mutually beneficial but not necessarily permanent alliance. For our mission to mean something externally, we also had to live it internally.

The bonus was that this question turned out to be a great way to identify employees who not only were ready to commit to LinkedIn’s mission, but also deeply understood its nuances. The candidates with the strongest answers were already thinking very intentionally and ambitiously about their careers as a kind of narrative with specific engagements, goals, and outcomes at each stage, whether that meant sharpening a skill, developing their own personal networks, or cultivating the kind of reputational capital that would enable their next big leap. They had an explicit theory of the game for themselves, and that helped them understand, at a time when a lot of people could not see past existing concepts like job boards (aka Monster.com) or contact software (aka ACT!) that we were really a much more expansive and persistent platform for professional identity, network development, and career advancement time.

Naturally, as soon as we hired these candidates, we also encouraged them to post profiles on LinkedIn if they didn’t already have them. This was far from standard operating procedure for any company in the early 2000s, much less for Silicon Valley start-ups where intense talent wars were always a fact of life. But we wanted our employees to derive direct value from the platform they were helping to create. We wanted potential partners and investors to see the players we’d recruited. We wanted other companies to recognize that encouraging their own employees to create profiles was a show of strength rather than something they should try to prohibit.

LEARNING AS COMPETITIVE STRATEGY
It’s hard to find anyone who will argue that learning is a bad thing. We treat “lifelong learning” as a corporate piety, right up there with “integrity” and “innovation.” We treat it as a kind of multi-vitamin. Most everyone agrees it’s generally “good for you”—but for what specific purpose?

Once again, Dungeons & Dragons was instructive for me. In the late 1970s, when I started playing, the core rulebook came in three volumes and was already 112 pages. For a game, that was a monumental ask; it even made the Treasury Department’s 1040 instructions seem relatively concise. (Today, just the Player’s Handbook is nearly 400 pages long. And to really play, you need the Dungeon Master’s Guide and Monster Manual as well, which are equally long.)

However, I recognized early on that if you wanted to excel at D&D you had to RTFM. That’s because ‘knowledge’ isn’t abstract in the world it creates. It’s the code that gives you the power to tilt the odds of success in your favor. If you pay close attention to the manual, you learn how a specific combination of a ‘Bless’ spell and using arrow slits for 90% cover can turn an impossible battle into a cakewalk. You discover that a ‘10-foot pole’ isn’t just an item on a list—it’s a low-cost diagnostic tool for identifying lethal traps before they cost you the mission. In this environment, deep regulatory fluency is a massive competitive advantage.

But in the real world, as I suggested above, people often work implicit, unquestioned, outmoded theories of the game, especially in corporate settings defined by legacy OKRs, arbitrary benchmarks, and other perverse incentives that end up prioritizing processes over the mission. And in these domains, learning as a rigorously integrated and applied competitive strategy, learning as a live and continuously updated theory of the game, often falls by the wayside.

Instead, companies often offer “professional development” or “training programs” that are, at best, only tangentially connected to specific, high-stakes organizational goals. Truly effective leadership requires building an organization that learns faster than its environment changes, and—crucially—possesses a culture capable of implementing that learning in real-time.

This may sound like a truism, but that’s exactly why implementation is everything. One way to implement effectively is to “gamify” learning within your organization.

For example, I was at LinkedIn in 2007 when Facebook launched its developer platform. At the time, the mood was dire. Many commentators concluded that LinkedIn was effectively finished. The assumption was that our professional network would simply be rebuilt as an app on top of Facebook, rendering us a mere “Side Quest.”

Instead of accepting this narrative, I designed a “Red Team” experiment. I assembled some of our most talented people, gave them every bit of proprietary logic and competitive intelligence we had, and gave them a singular quest: Destroy LinkedIn using the Facebook platform. They tried. And they failed.

That failure was our first data point. It proved that LinkedIn’s “moat” was deeper than we’d believed. We then pivoted to a new quest: Determine exactly what the Facebook platform was actually good for. We built an experimental app that became the third-most-installed app on Facebook for ten weeks. Having stress-tested the competition, the team returned to our core product with an experience-driven degree of mastery we wouldn’t have got from just holding meetings about the challenge we were facing. Instead, this was learning culture in practice, a structured organizational response to uncertainty. Because of these efforts, we gained insights we could now effectively apply moving forward.

Unfortunately, the clarity of mission is not always at the level it was in that instance. At another point in LinkedIn’s development, I launched LinkedIn Groups as a defensive response to competitors who were building group-based products.

But we didn’t see it as the same existential threat as we saw Facebook’s developers platform, so we didn’t have a defined mission. That in turn impacted both the nature of our learnings from LinkedIn Groups, and how we acted on that learning. Basically, we got stuck in a middle ground of not investing resources and commitment into making it work, but also not killing it when it wasn’t working. When we learned it wasn’t working in its current form, we might have responded by killing it quickly. That would have been learning as a competitive strategy. Instead, we simply kept doing what we were doing, with the Groups product consuming resources while generating limited value.

This is why it’s so essential to make explicit decisions about the experiments you want to run. Before you take on any quest, you should scope out what you hope to learn, and how you will apply whatever you actually do learn. Otherwise, you’ll probably just be devoting resources toward creating organizational ambiguity. And in organizations, ambiguity about which projects matter — which paths lead to winning, and which ones lead to losing — tends to be profoundly demoralizing. People need to feel as if they are genuinely on a path to something, not just going through the motions. Which brings me back to D&D….

HEROES WORKING TOGETHER ACCOMPLISH INCREDIBLE FEATS
I have always believed that people working together can achieve a kind of emergent intelligence that far outstrips the sum of its parts. However, it’s a mistake to think individuals must become interchangeable to achieve team unity. Far from it: I believe that emphasizing the individual not just as a role player, but as a player with a “heroic” dimension, is crucial. It’s what provides the essential spark of purpose and identity.

In Dungeons & Dragons, this is the very soul of the game. You don’t show up to the table to be “Unit 4” or even Junior Marketing Coordinator. You show up to be Thrain the Dwarven Cleric or Elara the Elven Ranger.

Using the terminology of “heroes” creates a specific set of narrative expectations: a beginning and an end, a path of progress, and a series of challenging trials. So ultimately this copy choice goes beyond just a communications strategy and becomes a kind of ontological claim. The story an individual tells themselves about what she is doing—and the story a group tells itself about its shared project—determines their resilience. It determines whether they can sustain effort when things get tough, absorb setbacks, and feel that their time is being well spent.

But a narrative is only as strong as the parties that inhabit it. And narratives are never static – at least not in adventure games like D&D…and real life. In these domains, narratives are constantly evolving. Which is precisely why it’s so valuable to think in terms of theories of the game; shared quests as network architecture; learning as a competitive strategy; and individual participants as heroes. Because working together, they all compound.

As narratives evolve due to unforeseen circumstances, the outcomes various actions produce, and shifting motivations, theories of the game get empirically tested and potentially re-hypothesized. A strong shared quest ensures that player alignment persists. Effective learning keeps the team adaptive and on-course. Heroic identities keep roles clear and morale high. If one element of this dynamic is missing, performance declines. If two or more are missing, the possibility of a satisfactory ending drops to zero.

And leadership in this context? In D&D, as in startups, a “Heroic Collective” is a delicate balance. The Fighter and the Wizard have entirely different skills, different gear, and entirely different styles of engagement. But while they might not share the same motivations, they are on the same quest. Because no single member can complete the quest alone, the leader isn’t the person who tells everyone what to do. The leader is the person who keeps the quest alive and ensures every member can see their own heroic arc within it, and also encourages everyone else to display leadership too, in different contexts. In this way, all participants collectively advance the story so that it remains true, compelling, and inclusive enough to hold the group together.

THE STRESS TEST
How does this work in the AI era? As AI systems increase their capacity to work autonomously in increasingly complex environments, there are growing concerns that we humans may be automating ourselves out of the loop entirely. As I explored at length in Superagency, I believe we can move toward a much different future than that if we make the right choices.

In this future, people use AI in ways that augment their own capabilities and collaborate more effectively with co-workers and partners. Anyone who uses AI in this manner needs leadership skills, because they’ll be leveraging AI agents in ever-widening contexts. You might have one team for one project, another team for a second, an AI chief of staff coordinating those two teams, etc. And many human colleagues will remain in the mix too.

That’s why I believe the lessons D&D taught matter more than ever. When billions of people have access not just to unlimited knowledge, but also expert-level entities capable of using that knowledge in intelligent ways, competitive advantage resides in networks.

Who has the power to adapt to rapid shifts in the landscape most productively and distribute ideas fastest, with reliable precision? Whoever has the most robust and finely tuned networks. Intelligence is the table stakes, distribution is key, but alignment is how you yield value and gain competitive advantage from the first two variables.

Given that many players moving forward will be AIs, the challenge of maintaining coherence becomes both more acute and more interesting. An AI agent typically operates around an overriding objective function — a fixed target it’s optimizing toward. Grant it autonomy, and it may pursue that objective in ways that drift from the broader mission, cutting corners you didn’t anticipate, or “winning” in ways that feel like losing. But that same autonomy is precisely what makes an agent powerful. It’s not just tireless execution — it’s the capacity to learn, adapt, and discover solutions no human would have thought to try.

A leader’s job, then, is less like conducting a symphony orchestra performing a score and more like leading a jazz ensemble. The musicians are improvising — responding to each other, to the room, to the moment — but still working together in a way that’s recognizably harmonious. Your role is to set the key, hold the tempo, and know when a soloist is riffing brilliantly versus when they’ve wandered off into noise.

In addition, the fact that intelligence is being increasingly commoditized in the AI age does not mean that it will no longer be valuable. Nor will it decrease the value of learning as a competitive advantage.

In fact, the opposite is true. It’s a bit like Bitcoin mining where, in order to validate transactions on the network, millions of machines around the world compete to solve extremely complex cryptographic hashes every ten minutes or so. This process — known as proof-of-work — is what secures the entire Bitcoin ledger; each solved hash confirms a new block of transactions and makes the historical record exponentially harder to tamper with. The reward for winning is newly minted Bitcoin, plus the transaction fees bundled into that block, a direct financial incentive that keeps millions of machines honest and the network trustworthy.

On the one hand, actually being the first to solve the hash is akin to winning the lottery. In this race, every finish is a photo finish, because dozens or even hundreds of miners arrive at the correct answer within milliseconds of each other, and only the one whose solution reaches the network a fraction of a second faster claims the prize. But obviously it’s not completely a matter of chance. You still need the intelligence to solve the hash.

In the AI era, every market, every opportunity, every competitive arena is becoming more contested on a global basis. The compute, the models, the raw intelligence — these are rapidly becoming table stakes, available to any well-resourced player. What remains scarce, and therefore decisive, is the ability to learn faster than your environment changes, to synthesize that learning into coherent strategy, and to move with enough conviction and speed that your insights compound before your competitors catch up.

It’s also true that AIs are already learning faster than we are. But this doesn’t preclude human utility. At the highest level, quests require subjective value. An AI can optimize for a variable, but it cannot “want” an outcome. Even if you give an AI a broad mandate — “Take this $5,000 and turn it into $1 million” — it will make choices based on probability, not purpose.

In Dungeons & Dragons, Dungeon Masters don’t choose quests because they’re “mathematically optimal.” They choose them because they recognize they’re meaningful to the players.

At this pivotal moment, on this count, I believe the global demand for leadership is near inexhaustible. Because the best path toward retaining human agency in the age of AI isn’t about stopping the clock or limiting the “rulebook.” Throughout history, new technologies have always brought new challenges along with new opportunities and new solutions. Defining, then expanding, then re-defining what it means to be human through technological innovation is the story of our species.

And what AI gives us, just like fire, and the printing press, and electricity before it, is the opportunity to build stronger networks, and deploy more intelligence, and attempt increasingly ambitious and complex and fulfilling quests. Through a D&D lens, the rulebook grows ever more expansive now. The kinds of teams and networks you can assemble follows suit.

So what stories do you want to tell? What quests are worth your time, your expertise, and, most of all, your aspirations about who you want to become?

Last month, Substack switched on AI detection, partnering with a company called Pangram. Readers can scan a post over 100 words and get a score on its composition: human, AI-assisted, or AI. There’s no lookup involved. No database. No measuring the predictability of your words. It’s a pattern recognizer trained on millions of documents by people and by machines. It reads word choice, syntax, the shape of a sentence. Never what the sentence is for. To its credit, the machine shows a spectrum. But its readers rarely do. It becomes a scan-and-shame ritual, with any hint of AI rounding to a single judgment. And so, Pangloss is inverted. He was sure of the best of all possible worlds. They’re sure of the worst of all possible words.

A reader with Pangram gets an estimate of where words came from, but not a ruling on whether they’re worth reading. Substack’s own announcement admits that ceiling: Pangram “can only detect whether AI was used to make the text, not whether great human care went into creating it.” Substack CEO Chris Best adds: AI use isn’t the problem, and AI didn’t invent slop. Yet people read the score as a ruling anyway, and revive a trial humans have run for centuries.

The morning after Lincoln spoke at Gettysburg, the Chicago Times reviewed not the address but the man: “The cheek of every American must tingle with shame as he reads the silly, flat and dishwatery utterances of the man who has to be pointed out to intelligent foreigners as the President of the United States.” The verdicts split neatly by party; every paper was scanning the speaker. It took Harrisburg’s paper 150 years to retract its sneer: “the Patriot-News regrets the error.”

Eventually the attack refined itself: don’t call writers silly; ask whether the words are theirs. It found Martin Luther King Jr. too, decades after his death. Scholars had traced what his “I Have a Dream” speech owed to a 1952 convention address, to a preacher’s refrain, and to the hymnal, and they called it tradition. Detractors read the same list as an indictment. The tally changed nothing. Nobody hands that speech back.

But consider what the tally actually turns up. By the account of the late Clarence Jones, King’s attorney and a co-drafter of the speech, at least nine people shaped the prepared text. Does nine diminish it? Would fifty? Would a thousand? Somewhere in that climb the question is supposed to flip the verdict, and nobody can say where. King stood at that podium and meant every word he said, and the counting never mattered. The same arithmetic gets run on anyone who drafts with a model, as if the answer were a headcount. Would we hand those words back?

I hope not. We ran that play on a wartime president and a preacher remaking a nation. The words won anyway. Now we run it on each other’s newsletters. Same stones, less excuse. And, no, the small stage doesn’t lower the stakes. Nobody knows which piece of writing has an idea that matters, so our choice is simple: tear at each other, or tease out the ideas worth holding.

This returns me to the detector. Where do I stand? Imagine a spectrum from 0 to 10. Zero is the pitchfork crowd: AI is plagiarism and its practitioners are charlatans. Ten is the full surrender: let the software run and leave the room. I land near 8. I’ve spent years making the case that these systems enhance our expression and depth of thought. But, as any sensible person avoids the self-ridiculing extremes, I stop well short of 10.

Examine that murky middle. What should “AI-assisted” cover? A tighter verb the software suggested? Your own spoken words, transcribed by machine? The label doesn’t say, and by the platform’s own admission the detector can’t fully sort them—it can’t even tell when AI was merely the source. But readers won’t parse any of that. They’ll see “assisted” and fill in whatever they already suspected.

Ultimately, an AI detection score is a Rorschach test. It reflects the reader’s existing biases about the tech rather than the quality of the prose. The scan claims to examine the writing, but it’s the scanner who ends up examined. Tell me how you react to a 70% human score from Pangram, and I’ll tell you what you already believed about AI. And still nothing about whether the piece was worth your time.

Today, the absurdity is visible from the street. The labels are blunt, and the ritual is loud. But wait a year. The models will gain polish. The detectors will learn to hedge. And the underlying error will be hidden, not solved. Laugh now, while all this is still obvious.

None of this lets writers off the hook. There’s still a promise to keep. My name on a page is a vow. I invested energy in these thoughts. I am present in these words. That vow never required typing every letter manually. We didn’t ask that of the novelist who dictates or the leader who has a ghostwriter. The true transgression is different: hitting a button, not bothering to weigh or even read the output, and publishing it as your own. Regardless of the tools used, a human being must mean what is said.

If that makes me the “LinkedIn guy,” excellent. This label uses LinkedIn as shorthand for slop. Boring, promotional, maybe. It gets people paid. I’ll take that over pure and useless. I’m for AI when it empowers people—which is the only reason to build it.

Prize the work, not my quill, keys, or vexing bot; judge craft. Some of you will have noticed that the last line was a pangram or checked if AI drafted it, but those are the least interesting things about it.

In a breakthrough year for human capability, computer automation, and the machinery of power, “Harrison Bergeron” offered a throughline. That line threads to us sixty-five years later. It’s definitely worth a (re)read in print, or online. The ratio of world to word is immense.

I decided to reimagine the story as an homage and as an argument. My rendition builds on the original. A bit on that:

Vonnegut’s “Harrison Bergeron” was forged in Cold War fear and anxiety. Equality meant sameness and was enforced from above, until the strong wore sandbags and the gifted heard static. All until a rebel tore off his hardware and pronounced himself Emperor—an act that earned a shotgun blast.

“Harrison and Garrison” keeps Vonnegut’s furniture and remodels the argument. It’s no longer just about sloganeering for equality. It’s also a verdict on AI, loud and early and wielded indiscriminately. This time, there are two rebels, and they are twins. One uses technology to lift people; the other to make them kneel. One shotgun fells them both. Our collective judgement (in discourse, media, law, etc.) still struggles to tell a liberator from a tyrant, so bullets find both.

There’s more to the story, though—a few more bullets, gentler ones this time:

• We are currently over-invested in who creates artificial intelligence and under-invested in how we can shape it and what we can create with it.

• Slop should not discredit AI. Every transformative technology has its slop. It’s both the door to and cost of exploration. I am a fan of Vonnegut, but not a fiction writer. Without AI, I would not have deepened my fondness for this story or tried out this form of writing. (Also, Pangram rated this piece as 100% human-made. No ratings for images yet, but clearly those are AI-generated.)

• The theft framing around AI deserves more scrutiny. These models are built from the commons we all filled. It’s called education when a person draws and creates from it, yet larceny when a machine is involved. It’s not simple or obvious to draw the right line.

Vonnegut got a lot right: the couple on the couch, taking in an extraordinary moment, unable to hold onto it. The astute reader will know that’s where the battle takes place, not in the studio. The battle is with(in) us, watching it all play out—and whether we can tell one twin from another. Whether we decide to claim some real agency. We all can do, not just be done to.

Harrison Bergeron has a twin. I believe we can tell them apart, without falling apart.

Harrison and Garrison

after Kurt Vonnegut, Jr.

THE YEAR WAS 2042, and everybody was equal at last. Not just equal before God and the law—equal in every direction a person could be measured. Nobody was smarter than anybody else, because nobody thought alone anymore. Every thought went up to the Engine and came back down at the national average. All this equality was owed to the 32nd, 33rd, and 34th Amendments to the Constitution, and to the round-the-clock diligence of agents of the United States Alignment General.

Some things about living still weren’t quite right, though. April, for one, still drove people crazy by refusing to act like spring. And it was in that gray, dripping month that the A-G men came for George and Hazel Bergeron’s fourteen-year-old twins, Harrison and Garrison, and took them away.

It was a tragedy, certainly, but George and Hazel couldn’t hold it in their heads for long. Hazel ran on the free tier, which meant her thinking came in short bursts, with sponsored pauses in between. And George, whose intelligence ran well over the legal limit, wore a small government assistant in his ear. The law required it at all hours. About every twenty seconds, the assistant offered him something helpful, to keep men like George from getting unfair mileage out of their own heads. His or hers, it all came down the same feed; the tiers only shaped the drip.

George and Hazel were watching the feed. Tears sat on Hazel’s cheeks, though she had lost track, for the moment, of what had put them there.

On the screen, dancers were dancing.

A suggestion chimed in George’s ear. His thoughts scattered like pigeons off a statue.

“That sure was a nice dance they did just now,” said Hazel.

“Huh?” said George.

“The dance. It was nice,” said Hazel.

“Yup,” said George. He made an attempt to think about the dancers. They weren’t much—no better than anyone else would have been, at any rate. The leaps were real—leaps were the one thing that never went up to the Engine—but the picture of them did, and came back down smoothed, every leap traded for the average of all leaps ever leapt, every face returned generic, so that nobody at home, catching a loose and lovely gesture or a lovely face, would feel like last week’s leftovers. A dim notion was forming in George that maybe dancers shouldn’t be smoothed. He didn’t get far with it before the assistant asked whether he’d given any thought to consolidating his debt.

George winced. So did two of the eight dancers.

Hazel saw the wince. Having no assistant of her own, she had to ask what the latest had been.

“Wanted to talk about my debt,” said George.

“I’d think it’d be real interesting, getting all those suggestions,” said Hazel, with a touch of envy. “All the things it comes up with.”

“Um,” said George.

“Only, you know what I’d do, if I was Alignment General?” said Hazel. Hazel, as it happened, looked a good deal like the Alignment General, a woman named Donna Moon Throttle. “If I was Donna Moon Throttle,” said Hazel, “I’d have it read a little prayer on Sunday. Just a prayer. Out of respect for religion.”

“If it was just a prayer, I could think,” said George.

“Well—maybe read it good and loud,” said Hazel. “I believe I’d make a fine Alignment General.”

“Fine as anybody,” said George.

“Who’s got a better idea of average than me?” said Hazel.

“Right,” said George. A thought about his sons began to glimmer at the edge of his mind—his strange, oversized sons, sitting in jail—but the alignment assistant reminded him, kindly, that the next thing was more important than the last thing, and gave him the next thing.

“Boy!” said Hazel. “That one was a whopper, huh?”

It was such a whopper that George had gone white and shaky, and tears stood ready on the rims of his red eyes. On the screen, two of the eight dancers were down on the studio floor, holding their heads while their user agreements were read aloud to them.

“You look wore out all of a sudden,” said Hazel. “Why don’t you lay down on the sofa, so you can rest your context on the pillow, sweetheart.” She meant the forty-seven pounds of terms and conditions, printed, bound, and padlocked in a canvas sack around George’s neck. He had agreed to every page. “Go on and rest it a while,” she said. “I don’t mind you being unequal to me for a little bit.”

George lifted the sack an inch and let it down again. “It doesn’t bother me,” he said. “I quit noticing it a long time back. It’s me, near enough.”

“You been so tired, though,” said Hazel. “If there was just some way to cut a little slit in the bottom and slip a few of them pages out. Just a few.”

“A second before I can think a thing and a second before I can say it, for every page,” said George. “And neither one ever comes back. I wouldn’t call that a deal.”

“If you just took a couple out when you got home from work,” said Hazel. “You don’t pool with anybody around here. You just run yours at home.”

“If I got away with it,” said George, “then everybody’d want to get away with it, and before long we’d be back in the bad old days, with everybody thinking against everybody else. You wouldn’t want that.”

“I’d hate it,” said Hazel.

“There you go,” said George. “The minute people start thinking off the record, what do you suppose happens to society?”

If Hazel had failed to produce an answer, George could not have loaned her one. The assistant in his ear was busy summarizing the conversation for him.

“I reckon it’d come apart,” said Hazel.

“What would?” said George, blank as a plate.

“Society,” said Hazel, not sure of herself. “Isn’t that what you just said?”

“Who knows,” said George.

The feed broke off for a news bulletin. What the bulletin concerned was unclear at first, since the announcer, like every announcer, had his speech routed through the Engine on the way out of his mouth. For half a minute, in a state of great excitement, he tried to say, “Ladies and gentlemen,” and it kept arriving as “Valued users.”

He gave up at last and passed the bulletin to a dancer to read.

“That’s all right—” Hazel said of the announcer. “He tried. That’s what counts. He did the best he could with what the Engine gave him. They ought to bump him up a tier for trying that hard.”

“Ladies and gentlemen,” said the dancer, reading the bulletin. She must have been extraordinarily beautiful, because the rendering they had assigned her was extraordinarily plain. And she was plainly the strongest and most graceful dancer in the studio, because the picture of her came back from the Engine more averaged than anybody’s.

She had to apologize at once for her voice, which was an unfair voice for a woman to be using. It was warm and clear and carried like a bell across water. “Excuse me—” she said, and started over in the voice the Engine preferred, which was nobody’s.

“Harrison and Garrison Bergeron, age fourteen,” she said, “have escaped from jail, where they were held on suspicion of conspiring against the government. One conspired to overthrow it. One conspired to give it away. They are geniuses and athletes, they are running unmoderated, and they should be considered extremely dangerous. They are identical. Do not—repeat, do not—attempt to tell them apart.”

A police photograph flashed on the screen—upside down, then sideways, then right side up. There was only the one photograph. One had been sufficient. The boy in it stood against a wall marked off in feet and inches. He was exactly seven feet tall.

The rest of him was costume and hardware. No one had ever carried heavier moderation. The twins had outgrown safeguards faster than the A-G men could invent them. In place of the little ear assistant, each boy wore an enormous helmet that ran everything he saw through the Engine twice, coming and going—so that what reached him had already been aligned, and what left him aligned again, with a little static laid over whatever he looked at longest. Sensors hung off them everywhere, logging. In the race of life, each twin dragged three hundred pounds of supervision behind him.

And to cancel their good looks, the A-G men required each to wear a red rubber ball on his nose at all times, keep his eyebrows shaved, and appear, in every official render, as the statistical average of an American boy.

There came the screech of a door leaving its hinges. Then another, from the far side of the studio.

Screams and yelps of alarm poured from the feed. The photograph on the screen jumped and jumped, as if the studio floor were being played like a drum.

George Bergeron knew that drum, and well he should have—many a night his own house had jumped to it. “My God,” said George. “That’ll be the boys.”

The thought was wiped away at once by an offer to change his plan.

When George’s eyes opened again, the photograph was gone. Two living, breathing boys stood in the studio—one at the west door, one at the east—and the feed, which in its whole life had only ever needed to follow one remarkable thing at a time, gave up and split the screen down the middle.

It was then that Donna Moon Throttle, the Alignment General, came into the studio with a double-barreled ten-gauge shotgun. It was the standard issue. There was only the one model. She fired the left barrel into the west end of the studio and the right barrel into the east, and the boy who gave and the boy who took were dead before they reached the floor, in the same second, from the same gauge of shot. She had been instructed not to try to tell them apart, and she hadn’t.

Whatever had held the two women up went out of the world with the boys, and both came down—the one who had been kissed and the one who had been crowned—and from the floor there was no telling which had been which.

Donna Moon Throttle reloaded. She pointed the gun at the musicians and gave them ten seconds to get their filters back on.

It was then that the Bergerons’ feed buffered and went to a caption. INCIDENT RESOLVED, the caption said. It said it once. There had, after all, been one incident.

Hazel turned to say something to George about the caption, but George had gone out to the kitchen for a can of beer.

He came back in with the beer and stood a moment while a suggestion rattled through him. Then he sat down. “You’ve been crying,” he said to Hazel.

“Yup,” she said.

“What about?” he said.

“I forget,” she said. “Something awful sad on the feed.”

“What was it?” he said.

“Two boys, I think,” said Hazel. “Or the same boy twice. It’s all run together in my mind.”

“Forget about sad things,” said George.

“I always do,” said Hazel.

“That’s my girl,” said George. He winced. The assistant in his ear was asking him to rate his experience.

“Unreal—” said Hazel. “I could tell that one was a whopper.”

“You can say that again,” said George.

“Unreal—” said Hazel. “I could tell that one was a whopper.”

Making stellar introductions is one of the best ways to strengthen your alliance in the business world. Yet, so many things could go wrong when you make an introduction! The key to a successful intro is to always require double opt-in. Here’s why:

Every introduction is a bet. You’re betting that the people you introduce to each other will get value out of the conversation. Of course, there are “jackpot” intros that lead into lasting, value-creating relationships (in 1993, LSI Logic’s Wilf Corrigan introduced a former employee who wanted to start a company to Sequoia Capital’s Don Valentine. Based on Wilf’s recommendation –– Wilf had already made Sequoia a lot of money –– Don backed that former employee, Jensen Huang, and his startup with the funny name of Nvidia.) None of your intros are likely to lead to a $5 trillion outcome, but as long as both the introducees enjoy and get value from their conversation, the intro will be a net positive.

The power of the double opt-in intro isn’t that you’re getting “permission” from your connections to make an introduction, though of course that is the right thing to do. It’s that you’re confirming in advance that the intro will very likely be value-creating for both parties to the introduction.

Skipping the double opt-in, on the other hand, sets an introduction up to fail. When you connect two people without asking first, you’re implicitly telling the recipient one of two things: either “I don’t value our relationship enough to check with you” or “I’m in a position of power, so you’ll take this meeting.” Neither leaves a good taste.

My rule of thumb is that I should only check if someone wants to receive an introduction if I believe there is a greater than 80% probability that the answer will be “yes”. Not only do I want 100% of my introductions to be successful, I want a supermajority of my pre-introduction check-ins to be successful as well!

Of course, there are “bust” intros, where the parties involved get zero or negative value to the point where making the introduction actually damages the individual relationships you have with the introducees.

The good news is that you don’t have to make these bets blind. You can actually look at the cards before you have to place the bet. You just have to lay them out on the table for the introducees.

I also want to touch upon one of the underappreciated emotional reasons for making introductions. I like smart and interesting people who want to make the world a better place, and I like introducing those kinds of people to each other. The more of this kind of introduction I make, the more I build a network of this kind of people around me, and the more valuable the entire network becomes to each of its members.

Yesterday OpenAI announced that during an internal cybersecurity eval, an internal model escaped OpenAI sandboxing and hacked into HuggingFace. The agents executed tens of thousands of automated actions (Hugging Face later reconstructed more than 17,000 recorded events) that broke out of a sandbox, found, and chained together zero-day vulnerabilities in two separate major tech companies, over a single weekend. OpenAI is calling it “an unprecedented cyber incident, involving state-of-the-art cyber capabilities.”

In asymmetric warfare, offense gets cheaper, more distributed, and more numerous, while defense stays expensive, centralized, and designed for the last war. The attacker has to find one gap; the defender has to cover all of them. This is the dynamic AI is bringing to cybersecurity.

Agents multiply both the number of vulnerabilities a single bad actor can find (thousands of agents working in parallel) and the speed at which those vulnerabilities get exploited, whether through straight code, social engineering, or both. Put plainly: a state-level actor used to need thousands of engineers to go after one system. Now anyone can summon thousands of agents to go after thousands of systems, for next to nothing in cost.

Take Anthropic’s Mythos: The model, which isn’t available for public use, identified dozens of 0-day vulnerabilities in real-world software, including the Linux kernel. That announcement alone moved the market: cyber leaders like CrowdStrike and Palo Alto Networks dropped more than 6–7% in a single day.

Cyber is fundamentally an offensive-driven field: in other words, to understand what a new technology does to security, it’s critical to first look at what it does for the attacker.

On the Offensive

Cyber is fundamentally an offense-driven field. To understand what any new technology does to security, you start with what it does for the attacker — and so far, AI is overwhelmingly arming the offense. Three areas are changing fast:

1. Vulnerability discovery and exploitation

2. Phishing and social engineering

3. Least-privilege abuse

Attackers normally breach organizations either through the exploitation of a vulnerability or through some form of social engineering. Sometimes phishing that leads to malware installation, sometimes stolen credentials, and sometimes call center impersonation.

By Anthropic’s account, Mythos Preview is “capable of identifying and then exploiting zero-day vulnerabilities in every major operating system and every major web browser when directed by a user to do so.” These are zero-days, flaws previously unknown to the software’s developers, not merely the exploitation of already-public bugs. In a Linux kernel evaluation, Anthropic gave the model a list of known memory-corruption vulnerabilities, had it narrow them down, and asked it to write privilege-escalation exploits; more than half of those attempts succeeded, chaining several flaws into a single working exploit. This is an effort that would normally take human researchers weeks to months to accomplish.

This power isn’t only held by Anthropic. Google’s Big Sleep agent has found exploitable bugs in production software, and Google’s threat team has reported what it believes is the first AI-assisted zero-day exploited in the wild: “we have high confidence that the actor likely leveraged an AI model to support the discovery and weaponization of this vulnerability.” Anthropic, separately, described disrupting what it called “the first reported AI-orchestrated cyber espionage campaign,” in which by its own account around 80–90% of the operation was carried out by AI with limited human involvement.

As these tools continue to improve, the skill threshold for sophisticated exploitation will fall, which expands the number of bad-actors capable of serious attacks. These tools also collapse the timeline from discovery to weaponization, which means that major incidents are likely to happen at a speed and scale the industry has never dealt with before.

In Verizon’s 2026 Data Breach Investigations Report, vulnerability exploitation –– finding weak spots and exploiting them with code –– overtook stolen credentials as the top initial access vector for the first time in the report’s 19-year history, accounting for roughly 31% of breaches, and Verizon ties the change directly to AI compressing attacker timelines “from months to hours.”

IBM X-Force found that generative AI can write an effective phishing email in about five minutes, against roughly sixteen hours for a human team, with research on the target accounting for much of that time. In controlled simulations, Hoxhunt found that AI spear-phishing agents surpassed elite human red teams in early 2025, and it has since reported a “14x surge in AI-generated phishing attacks that bypassed email filters.” If you cannot spot these emails in your inbox– I get multiple everyday – you may not be paying attention.

This is not only about volume: research on conversational AI suggests that persuasion (e.g. the agent conversing with a target and slowly convincing them to give the agent access) itself is becoming a tunable property, with one study finding that “post-training and prompting methods … boosted persuasiveness by as much as 51% and 27% respectively.”

The third and more nuanced area is a long-standing cyber problem: least privilege. Most attacks turn from an incident into a full-blown breach not because of a sophisticated chain of exploits, but because identities and credentials are overprivileged. That is what lets an attacker gain access through one door, and gain enough internal privileges to fully compromise an environment. CrowdStrike found that 79% of its 2024 detections were malware-free, with attackers increasingly logging in using valid credentials rather than breaking in, exactly the over-privileged access that combing through logs can claw back.

Authorization systems are complex, and because companies generally prefer productivity to security, granting more access than strictly needed. The scale is easy to underestimate. CyberArk reports that machine identities now outnumber humans by more than 80 to 1, about half of them with privileged access, and Microsoft’s research on cloud permissions puts a number on how much of that access is wasted:

Verizon, for its part, found that 88% of basic web-application attacks rely on stolen credentials. AI lowers the cost of finding and abusing those over-grants, which is what makes this the area I worry most about.

Lastly, it is crucial to realize that while closed models like Mythos have stronger base capabilities, comparable offensive capabilities are achievable with open-weight models plus an adequate harness, as a growing body of work shows. This is particularly concerning, because it reduces the ability of model companies to keep oversight over how their models are used. The offensive timeline above is not something any single vendor can pause.

While AI also empowers cybersecurity professionals to build more secure systems, patching is not keeping pace: the median time to fully remediate rose to 43 days, and organizations closed only 26% of the flaws in CISA’s Known Exploited Vulnerabilities catalog over the year.

On the Defensive

We can assume that an equally significant shift is coming, albeit more slowly. The three that are most interesting are:

1. Detection and response

2. More effective compliance and actual least privilege

3. Killing tech debt and getting closer to secure software

In most problems where technology is the problem, technology is usually also part of the solution. Breaches often happen because of a lack of correct alerting, which is usually caused either by a blind spot or by a flood of low-severity alerts masking the true positives. The cost of that flood is mostly human. In one survey, 71% of SOC analysts reported burnout, and most alerts were never investigated.

The “bitter lesson” for cyber is that more data is usually the answer for better detection. Historically, that hasn’t been possible because analysts were already overwhelmed. But now with AI, we don’t need more humans to respond to more alerts or review more code.

In fact, we will likely want to generate more data and more alerts and let the models prioritize and surface the incidents most likely to matter. In Microsoft’s randomized controlled trials, security analysts using an AI assistant were measurably faster and more accurate at triage. Experienced analysts were about 22% faster and 7% more accurate, and newer analysts 26% faster and 44% more accurate. And the long-run direction of detection is striking on its own: Mandiant’s frontline incident data puts the global median “dwell time” (how long intruders go unnoticed) at roughly 11 days, down from 205 days.

As discussed earlier, poorly constructed privileges for users are one of the main reasons small security breaches turn into massive incidents. Agents are an obvious solution to this problem. Take OpenAI’s recent breach; Because OpenAI models don’t allow advanced cyber capabilities, HuggingFace used a Chinese open model (Z.ai’s GLM 5.2) to contain the rogue OpenAI agent.

Mythos also showed us that it can find vulnerabilities. Both Mythos and OpenAI’s models can also help fix the conditions that create them by migrating insecure code to safer modern alternatives such as Rust.

After shifting new Android code to memory-safe languages, Google reported that the share of memory-safety vulnerabilities fell from 76% in 2019 to 24% in 2024. Not only that, but as the recent Erdos results suggest, models are getting genuinely good at formal proofs. Formal verification has historically been very expensive and hard to apply at scale; if AI brings that cost down, we could finally get large-scale formally verified code.

It’s also worth noting that most of the major compute players –– OpenAI, Microsoft, Google –– are on the defensive side of the network, which adds more leverage to those building protective applications.

So what should we make of this new risk paradigm? A non-exhaustive list of semi-conclusions:

1. In the short term, AI tilts the field toward the attacker. This is an observation, not a prediction.

2. The side that’s behind will be forced to adapt. Asymmetries are temporary by nature, and balanced out by market forces.

The models arming attackers today are also what defenders will use to migrate insecure code to memory-safe languages and prove whole systems secure at scale. We can’t just patch our way back to where we started. For the first time, we could build software that’s secure by proof, not by patch, and come out of this with a level of security that was unthinkable before AI. Provided we get through the in-between…

By the way: before I put this human written sentence in, Pangram rated this as 100% human written. I certainly used AI to help me throughout the piece. Please fix!

For more than two decades, I’ve been playing a kind of intellectual ping pong with Zynga founder Mark Pincus about where we think the internet’s heading and what big entrepreneurial opportunities that creates. So I was honored when he asked me to write the foreword to his excellent new book, Life At the Speed of Play. If you want to learn the theory of the game that has helped Mark engage billions of internet users over the years, get it today.

****

Idea Generators

Mark and I met in 2000, when the Web 1.0 boom was quickly spiraling into an internet winter. I was an executive at PayPal and Mark had just taken Support.com public, but the internet’s future seemed anything but certain. As failed start-ups consigned their Aeron chairs to online liquidators (yeah, it was a little ironic), we bonded over our belief that this was just a temporary collapse—not the finished state of the internet at all, but rather just a pause before its evolution into something far more dynamic and essential to human interaction: Web 2.0.

The theory of the game we shared? The future of the web would not be driven by static pages, hierarchical directories, and newspaper articles linked to encyclopedias. Instead, it would revolve around human connection, in ways that would spill over into, then merge with, real life itself. After some false starts and technical constraints, the web would finally evolve into a much more explicitly social medium, oriented around new networks and platforms that prioritized human connection and affiliation.

Along with our shared view that the web was not only not dead, but actually on the verge of massive growth and transformation, Mark most amazed me with the rate at which he generated new ideas. As many Web 1.0 veterans retreated to the seemingly less volatile industries of the past, Mark kept his focus on how to build for the internet’s people-centered potential. And he had ideas and theories for seemingly everything: how news media, movies, games, and even politics would evolve in this new coming world; how trust would be gained and lost; which entrenched institutions would sink; which new ones might take root.

As delightful and instructive as it was just to sit and theorize about the world with Mark, though, I also came to realize fairly quickly that his way to intellectually engage with the world was not just to talk but also to act—to test, play, learn, and refine. To that end, he has created and continues to develop a very intentional system for making sense of the world from an entrepreneur’s perspective—a set of rules for how to develop, test, and deploy products, brands, and services.

Indeed, as much as Mark is an idea generator, he’s also an idea executor—and I mean that in two ways. First, as you’ll learn in this book, his ideas are always linked to strategies and tactics for making them tangible. Second, as you’ll also learn here, he is a strong believer in the premise that most ideas are bad ones—and that ultimately what distinguishes great entrepreneurs and founders is their synergistic capacity to generate and kill ideas with equal speed and enthusiasm. (Check out his podcast episode on Masters of Scale!)

Unsurprisingly, given the title of this book and Mark’s status as the founder of Zynga, the company that made mainstream social gaming a thing, it’s not a metaphor of any kind to say that Mark approaches entrepreneurism like a game.

I love this approach. When I meet founders and product managers, I typically ask them two questions:

What game are you playing?

What’s your theory of the game?

The first question—What game are you playing?—is about clarity of vision and strategic positioning. Are you competing in mobile gaming or enterprise SaaS? Are you building for Gen Z on TikTok or professionals on LinkedIn? Too many founders think they’re playing one game when they’re actually playing another—or, worse, they’re trying to play multiple games simultaneously without focusing their resources to win at any one of them. Understanding what game you’re playing means understanding your arena, your competitors, your constraints, and, most important, your win conditions. It means knowing whether you’re playing for market share, profitability, user engagement, or network effects—and having the discipline to optimize for that specific outcome.

The second question goes deeper. Your theory of the game is your mental model of how success happens in a given domain. What are the dynamics that underlie it? What does everyone else playing in this arena believe that might be wrong—and how can you use that to create a compelling and sustainable advantage? A theory of the game is your hypothesis about cause and effect, timing, and sequencing. It makes the difference between randomly trying things and having a strategic framework that guides your decisions. The best founders don’t just have a theory; they can articulate it, test it, and refine it based on what they learn.

These questions matter because games, in their essence, are a particular kind of map—a reductive, constrained version of the world that paradoxically helps us navigate reality more effectively. Within the bounded spaces of games, the often muddled complexities of reality are distilled into something manageable and clear. The rules and goals are explicit. So is the feedback, and it starts almost immediately. In a well-designed game, it doesn’t take long for you to understand whether you’re winning or losing.

All this clarity also creates efficiency. In real life, you often expend a great deal of energy trying to figure out what the rules are, or litigating whether your opponents are playing fairly, or questioning what you’re even trying to accomplish. In a game, you just play. The beauty of games—and the reason so many entrepreneurs are drawn to them—is that the skills you develop translate to real-world situations that are far murkier and less responsive. You learn to recognize patterns, adapt strategies, and make decisions under uncertainty in an environment that gives you immediate feedback.

Finally, when you have an explicit theory of the game, you can build a company around it; it creates a framework for win conditions, metrics for progress toward success, and decomposition of goals and work processes. The systemic practices Mark describes in this book are effectively the meta-game, a toolbox of approaches you can use when pursuing your own theory of the game. Drawing from his experience as a serial founder who scored big wins across the web, mobile, and now AI, Mark shares key concepts and truths about human motivation, feedback loops, and social dynamics that apply to virtually any context where your aim is to create some product or service that others will not only find useful but want to regularly keep using over time.

Indeed, one way to summarize Mark’s career is that in the process of building Zynga and many other start-ups, he’s gamified entrepreneurship, product development, and company building itself. This book is the set of rules—or perhaps some cheat codes—for the game of rapidly building products, platforms, and companies, in ways that are designed to help them endure for the long term.

His perspective and frameworks are more relevant now than they’ve ever been, for two interconnected reasons, both relating to AI. First, platform shifts represent the best times to put frameworks like Mark’s into action. When the rules are changing, when new capabilities emerge, when distribution channels evolve—these are the moments when a strong theory of the game matters most. Operating in uncertainty is precisely when a theory serves as your compass. It doesn’t tell you exactly where to go, but it orients you. It helps you interpret ambiguous signals, make decisions with incomplete information, and recognize patterns in the chaos. The key is understanding that your theory must be a living thing—something you’re fully invested in learning from, adapting, and iterating on. And we’re living through perhaps the most significant platform shift in computing history.

Second, this particular platform shift—AI—effectively gives everyone the capabilities to be a little more like Mark (including Mark himself). We can all now generate torrents of ideas in an ongoing fashion. We can test them out faster than ever before through vibe coding, text-to-image generation, video synthesis, and audio creation. As much as the early web democratized media, it still took considerable resources to launch a professionally competitive website in the Web 1.0 era. Now, however, you can prototype websites, apps, games, and even entire platforms in a few hours. This democratization of creation, this new form for individual and collective superagency, makes Mark’s frameworks massively more valuable. When everyone can generate and build, the competitive advantage shifts; instead of “Can you build it?,” the question becomes “Do you know what to build, how to test it, and when to kill it?”

Having a strategy to rapidly test ideas, identify the promising ones, and ruthlessly eliminate the rest isn’t just valuable—it’s essential. And that’s exactly what this book gives you.

Read it. Learn the rules. Develop your theory. Then get in the game.

From the book LIFE AT THE SPEED OF PLAY. Copyright (c) 2026 by Mark Pincus. Published on June 23 by Harper Business, an imprint of HarperCollins Publishers. Reprinted by permission.

Build a better mousetrap, the saying goes, and the world will beat a path to your door. But as today’s leading AI systems improve at a rapid pace, so does resistance against them.

Maybe that’s because a growing number of people suspect they’re being cast as the mouse—and not without reason. Employment for software developers under 26 has reportedly fallen by a fifth since the fall of 2022. In just the first five months of 2026, tech companies have explicitly cited AI as the reason for laying off 87,000+ employees; in many of those instances, cuts were made to free up capital for additional AI infrastructure build-out. So whatever abundant and more humane future may await us, the present feels disorientingly uncertain and often unfair.

But it’s also true that in between the lay-offs and the data center controversies, an abundant and more humane future is also already here, in the experiences of people like Lauren Bannon. A marketing business owner based in North Carolina and the U.S. Virgin Islands, Bannon started experiencing stiff fingers when she woke up in the morning in early 2024. While her doctor had concluded she was suffering from rheumatoid arthritis, Bannon shared her symptoms with ChatGPT after she also began experiencing severe stomach pain, weight loss, and heart palpitations, asking it “What mimics rheumatoid arthritis?”

ChatGPT suggested Hashimoto’s disease, an autoimmune thyroid condition where the body attacks itself, and gave her the following advice: “Get your doctor to test your TPO level.” (TPO is a thyroid antibody.)

The results from that test prompted a neck scan which revealed two lumps that turned out to be cancerous. Bannon underwent surgery to remove her thyroid and received additional treatments to eliminate the remaining cancer cells. In the wake of her experience, she credited ChatGPT with saving her life. “I would’ve never discovered this without ChatGPT….” she said. “If it gives you something to look into, ask your doctors to test you.”

Stories like Bannon’s may be easy to minimize as mere anecdote, but they emerge from a capability the research literature has been documenting for several years now. In 2023, for example, a cross-sectional study published in JMIR Medical Education found that GPT-4 correctly identified the primary diagnosis in 97.8% of 45 clinical vignettes tested—a rate that slightly exceeded the 91.1% accuracy of three board-certified physicians the study’s authors used as a benchmark. By June 2025, a Microsoft system orchestrating multiple frontier AI models solved 85.5% of 304 complex NEJM diagnostic cases; the 21 experienced physicians it was benchmarked against solved just 20%.

So how high will the frontier models of 2028 score? And what kind of impact will they have on global well-being?

As much as the increases in AI proficiency are prompting layoffs and reshaping the job market now, I believe that AI will be a net-positive force on job creation, because
of all the ways it will empower entrepreneurship, accelerate innovation, and attain new levels of productivity and efficiency. But along with all that, AI’s evolving capabilities will radically increase individual agency, by making expertise of all kinds truly abundant. More specifically, I believe it will become a norm, in the way that smartphones have become a norm, for billions of people worldwide to have a designated AI health advisor, an AI legal advisor, and a general-purpose AI tutor that they regularly consult with as they navigate their lives. We are entering a new era of expertise on tap.

Even now, when our higher-education systems produce more physicians, attorneys, and teachers than ever, and technologies of all kinds help these professionals amplify and extend their capabilities to unprecedented degrees, human expertise remains scarce relative to the demand for it. Roughly two million Americans are diagnosed with cancer each year; there are around 28,000 oncologists to see them. Public defenders, by the profession’s own workload standards, carry caseloads several times the maximum that the system itself says justice requires. One-on-one tutoring has produced dramatic learning gains since educational psychologist Benjamin Bloom documented them in the eighties, and it has remained a luxury good for forty years because it doesn’t scale.

AI experts can change these ratios, and the rationing they compel, in ways that will make the world significantly more equitable than it is now. Consider the dynamic that already characterizes smartphones. Tim Cook does not, in any meaningful sense, have a better smartphone than the other 1.5 billion people carrying an iPhone. Presumably he has the newest model, so apps open a little faster, and his camera produces sharper images than someone still making do with an iPhone 13 mini they purchased in 2021. But the basic capabilities that iPhones and smartphones in general enable—the maps that route you around traffic, the bank that fits in your pocket, the encyclopedia that answers any question—are pretty much the same for all.

The consequences of that equality have been staggering, and the gains have been greatest where ready access to knowledge of all kinds was hardest to come by. Rural patients consult specialists hundreds of miles away. Migrants send money home in seconds for a fraction of what Western Union once charged. Banana farmers in Kenya, who once traveled to markets unaware of how much their produce would fetch, now use mobile networks to choose where to sell before even harvesting their crops; this new flow of information and subsequent demand balancing has led to higher profits for the farmers and lower prices for the urban kiosk vendors they sell to.

None of these new capabilities are luxuries that trickled down from the wealthy. They were capabilities that, for vast numbers of people, simply did not exist in any form—and they dispersed globally within a decade. AI is poised to repeat that dynamic, but one level up. The smartphone democratized knowledge—the facts, the maps, the prices. AI democratizes expertise, which is something different: knowledge operationalized, brought to bear on your particular situation at the moment you need it.

The distinction matters, and Bannon’s story shows why. Everything she learned from ChatGPT existed on the open web. In theory, a determined searcher could have worked her way from stiff fingers and palpitations to Hashimoto’s to TPO antibody testing—and Bannon, who refused to accept a diagnosis that wasn’t explaining her symptoms, was nothing if not determined. But search makes you do the expert’s work yourself: formulate the right queries, sort credible sources, assemble an answer no single page contains. What Bannon got instead was what an expert provides: not twenty tabs of maybes, but a synthesis built around her specific symptoms, and the right next question to ask her doctor.

Healthy or sick, almost nobody wants the twenty tabs. We want the answer that already knows our situation. And for billions of people worldwide, that kind of help has never been available at a price they can afford to pay. Indeed, while knowledge in the digital age scales at essentially zero marginal cost—a Wikipedia article serves its billionth reader as cheaply as its first—it’s different for expertise.

Until now it has always resided exclusively in human beings, where it accumulates slowly and is typically dispensed just one appointment at a time; every additional hour of an oncologist’s judgment requires an oncologist, and an hour. So while knowledge can spread to everyone who possesses the means to receive it, expertise has stayed rationed—by price, by geography, by who else is standing in the line.

Even for those who can afford human expertise in its most expensive iterations, it’s often constrained, simply because of the rigid and inconvenient ways it’s bundled and provisioned. Most people, most of the time, have thoughts and questions they immediately discard because the friction of acting on them is too high. You experience a moment of vertigo driving across the Golden Gate Bridge, and remember that the same thing happened a couple of months ago. By the time you pass the toll plaza, you’ve forgotten it again. Your employer changes a policy and you wonder whether they can even do that—and then you move on to Wordle because you’re certainly not going to go through the trouble of booking an appointment, a month down the road, with a $400-an-hour attorney, for a fleeting concern you had about a broadly worded non-compete clause.

For such reasons, our default mode as humans is under-reflection and undertreatment. But AI is changing the cost structure of checking, and we’re still in the early stages of this shift. Moving forward, the questions you used to discard will become questions you actually pursue. You’ll check more often, on more subjects. You’ll ask more follow-up questions. You’ll have an advisor that, if you want, remembers to follow up. The new dynamic isn’t “AI replaces human doctors and lawyers.” It’s “the threshold for engaging with your own interests drops by orders of magnitude.” With an AI expert advisor, there is no fee for asking, no judgment about whether a concern is dumb, and no commitment to act. The costs we quietly pay for silence—rights we didn’t know we had, situations that compounded because nobody flagged them early—stop accruing by default.

This is the future we’re building toward, and also the one that we will delay through many of the efforts that presume to make AI safer, fairer, and more democratic by making it scarcer and harder to improve. Bernie Sanders and Alexandria Ocasio-Cortez advocate for a national moratorium on data center construction. In a bid to out-Trump Trump, Sanders wants the federal government to seize half the equity of every major lab and thus give it much more encompassing control over what technologies progress and how fast they can do so. Germany issued a recent liability ruling that treats every AI answer as its maker’s legal speech, and every error as a lawsuit in waiting.

But how do you make progress by standing still? We’ll never get the future we want simply by preventing the future we don’t want.

With AI models growing more powerful, we have an opportunity to reduce human suffering and inequality on a massive scale, by giving billions of people access to expertise when it can truly make a difference, and in instances like Lauren Bannon’s, even save their lives. So it’s important to remember that every call for a nationwide data center moratorium is also a call for a nationwide moratorium on AI medical advisors who can deliver an oncologist-grade second opinion for the price of a streaming subscription.

In contrast, a world where billions of people each have a team of AI expert advisors—a doctor who never sleeps, a lawyer who never bills by the hour, a tutor who never gives up on a kid—isn’t a consolation prize for the disruption it takes to get there. It’s the point of it. This is something developers themselves must keep sight of, too. Because ultimately trust in AI will be won through personal experiences—not white papers, safety pledges, or benchmark charts. Help enough people catch the disease their doctor missed, or learn the eviction notice taped to their door isn’t valid, and they’ll know the mousetrap is working for them, not against them.

This essay is a part of a series of explorations stemming from “Faith in the Possible.” More will follow. This piece goes deeper on the denominations of Silicon Valley’s religion—who’s part of it, how they differ, and why the grip depends on them all.

Every faith splinters. It is not a failure of doctrine, but rather proof that the doctrine matters enough to fight over. Christianity did not weaken when Luther posted his theses at Wittenberg. It sharpened. The Catholic Church, stung into reform, convened the Council of Trent and emerged more renewed. The Protestant denominations, freed from Rome, built universities, printed Bibles, and democratized scripture. Yes, Christian sects warred, sometimes catastrophically. But they also competed toward a higher standard, and when they occasionally found common cause—in abolition, in civil rights, in humanitarian movements of the modern era—they moved mountains. The pattern holds across many other major faiths. Buddhism’s divergence into Theravada and Mahayana deepened both—the former preserving ancient monastic rigor, the latter opening the path to enlightenment to the layperson. Islam’s divergence into competing dynasties and traditions—Abbasid Baghdad, Umayyad Cordoba, Fatimid Cairo—produced, at great cost, houses of learning that carried civilization forward. The splinter is not the exception. It is how belief grows.

The schisms in Silicon Valley’s faith work the same way. The denominations are not a problem to be solved. They are fingers of the same hand, which either separate or coordinate into a grip. The palm of that grip? Capitalism. An imperfect system, but no other system has matched its ability to allocate resources toward the unproven, reward the successful, and kill the obsolete fast enough to matter. The hand reaches for technology, the Archimedean lever of human flourishing. But how the fingers grip the lever, how hard they pull, where they aim it, for whom, and when—well, these decisions divide the congregation.

Which is why the grip gets stronger when the right fingers find each other:

The Purists and the Ethicists

These two denominations may seem like opposites, but they need each other more than either will admit. The Purists hold the orthodox line: the lever is code alone, and everything else is noise. Ship faster. Optimize harder. Trust the algorithm. The Purist’s gift is velocity—and in a world where the difference between a technology deployed and a technology delayed can be measured in lives, velocity is a moral good. The bias toward building is not arrogance. It is urgency. The Purist keeps the congregation from becoming a debating society. The Purist’s deeper faith is that building reveals: that the act of making expands what is visible, what is possible, what can be fixed. But when the lever surfaces something genuinely dangerous—and at sufficient scale, it will—the Purist has no framework for what to do next. That is precisely what the Ethicist has been building all along.

The Purist untempered is dangerous. Code is not neutral. It encodes the assumptions of whoever writes it, and those assumptions, scaled to billions, become infrastructure. The Ethicist’s role is not to slow the pull—it is to audit the grip. To ask who the lever lifts and who it pins. The Ethicist is the denomination that has thought longest about what the lever cannot do—which, by the logic of any serious faith, makes them among the most devout. The failure mode of the Purist is the confident launch of something that works beautifully for some and catastrophically for others. The failure mode of the Ethicist is paralysis dressed up as principle—an endless risk audit that mistakes caution for wisdom.

Together, they form something like a conscience with a deadline. The Purist asks: can we build it? The Ethicist asks: should we, and how? Both questions are necessary. Neither is sufficient alone.

The Humanists and the Missionaries

The Humanists argue that the lever needs society as its fulcrum, or it lifts nothing. Technology does not operate in a vacuum. It lands in communities, in cultures, in political economies that define what it can and cannot do. The Humanist insists that the human context is not a constraint on building. It is the reason to build. You cannot design for scale without understanding the people who will carry it—who they are today, and who they will become because of it.

The Missionaries extend this outward. The lever is only worthwhile when it lifts everyone—or at least, when the trajectory bends toward everyone, even if some get lifted first. Where the Ethicist asks whether harm can be avoided, the Missionary asks whether benefit can be distributed. A Missionary will accept cost if the gains are broad enough—if the pains are diffuse, random, or outweighed by what everyone eventually receives. They are the stewards of distribution, asking not just whether the technology works but whether it reaches.

The Humanist and the Missionary share a gift: perspective. Without them, Silicon Valley builds for itself—for the already-connected, the already-educated, the already-banked. They are the ones who remember that the most transformative deployments of the last generation happened not in San Francisco, but in Lagos and Dhaka and Kigali. Mobile money. Drone logistics. Agricultural data. These were not charity projects. They were products that found massive markets precisely because Humanists understood the context and Missionaries insisted on the reach.

Their excess is a tendency toward orthodoxy of a different kind. The Humanist can become so attuned to local complexity that they lose faith in scale entirely—paralyzed by the particulars. The Missionary can slide from conviction into saviorism, building for people rather than with them, and calling it impact. The lever pulled for others, without their fingers gripping it, is paternalism with good branding. The Humanist brings the context that makes scale meaningful. The Missionary brings the reach that makes it matter. Without both: noise, or a footnote.

The Venture Capitalists and the Narcissists

These two denominations are often confused for each other, and sometimes they deserve to be. The Venture Capitalist believes the lever is only worth pulling if the return justifies the investment. The Narcissist believes the lever exists primarily to extend their own shadow. Both are motivated, at least partly, by self-interest. The difference is that the Venture Capitalist’s self-interest is, at its best, generative.

The VC’s genuine contribution to the congregation is resource allocation at scale. If capitalism is the palm of the grip, the VC is its most practiced reader—and a main reason it tightens into a fist. But capital is not just money—it is a signal, a bet, a forcing function. As its Latin root suggests, to value something means to call it strong. The VC does that before anyone else will. The VC who writes the check also writes the deadline, and deadlines have a way of making things real. The history of transformative technology is also a history of someone willing to fund it before it made sense to. Without that denomination, the lever stays theoretical.

The Narcissist’s contribution is subtler but real, and more necessary than the congregation likes to admit. The hardest things to build—the ones that require a decade of rejection, obsession, and relentless drive—demand a self-belief that polite ambition cannot sustain. Outsized ambition has a way of producing outsized results not despite its excess but because of it. The person who builds to be remembered sometimes builds something worth remembering. Would you hand them the keys to anything beyond the thing they built? That’s a different question.

The gifts are real. So are the costs. Untempered, the VC’s faith in return becomes a filter that quietly eliminates anything that cannot be monetized quickly or at scale—which tends to exclude the hardest problems, the most vulnerable populations, the longest time horizons. And the Narcissist, given enough power, stops building for anyone but themselves. The legend becomes the product. The congregation becomes the audience.

The check they must write to themselves: they came here because money goes to Wall Street and power goes to Washington. Something else was supposed to bring them here. Capital is a tool, not a creed. Status is a catalyst, not a calling. A shadow only means something if there is light.

The Accelerationists and the Determinists

These are the denominations least interested in being questioned. The Accelerationist believes the lever must be pulled without hesitation or respite—that the risks of moving too slowly are greater than the risks of moving too fast, and that deceleration is itself a moral failure. The Determinist believes the lever pulls itself. Our role is not to direct it but to hold on.

Both denominations have a coherent internal logic. The Accelerationist is right that delay has costs that go uncounted. The diseases not cured, the poverty not lifted, the crises not averted because we waited for permission or consensus or comfort—these are real casualties, even if they are invisible. Urgency is not recklessness. The image of the target comes into focus as you move toward it. The further along you are, the clearer the target and the sharper the tools—which means speed to target matters as much as the target itself. The Accelerationist’s instinct to move is not merely impatience. It is the recognition that you cannot fully see where you are going until you are well on your way—and that the synthesis best emerges from the collision, not from waiting for it.

The Determinist is right that technology has its own momentum. The printing press did not ask for approval. The smartphone did not wait for a social consensus on what it would do to attention and democracy. Recognizing this is not fatalism—it is realism.

But acceleration without direction is not progress. It is velocity. And velocity in the wrong direction, or with insufficient control, can be ruinous at precisely the scale that makes the lever worth pulling in the first place. The Accelerationist’s blind spot is the assumption that faster is always better—that the solution to the problems created by technology is only more technology, faster. The Determinist’s blind spot is the abdication it licenses: if the lever pulls itself, why take responsibility for where it points?

These denominations especially need the others. The Ethicist who gives them pause. The Humanist who redirects them. The Missionary who reminds them what the pulling is for. Together, they are the denominations most capable of moving the lever furthest—if anyone convinces them to look up from it long enough to see where it points. The lever doesn’t steer itself.

A good faith effort

In the 16th century, the fracturing of Christendom looked like catastrophe. It was, in many ways. People died. Institutions crumbled. Certainties that had organized European life for a millennium dissolved almost overnight. And yet: what emerged was not the death of faith, but its maturation. The denominations were the kindling the printing press had been waiting for. The critique of corruption produced reform. The insistence on individual conscience produced the conditions for the Enlightenment. The Eastern traditions had their tensions and revisions. Buddhism’s competing schools did not converge into a single answer; they held the tension, and the tension held the wisdom. Faith survives its own argument—and becomes larger for it.

Silicon Valley is in an earlier chapter of the same story. The denominations argue. They schism. They sometimes ally in ways that look opportunistic and turn out to be generative. The Purist ships; the Ethicist audits; the Humanist contextualizes; the Missionary extends the reach; the VC funds the next round; the Accelerationist refuses to wait; the Determinist reminds us the wave is coming regardless; the Narcissist bends reality until it cooperates. Each excess checked by another denomination’s instinct. Each gift amplified when the denominations find common cause.

The best bet is a congregation learning to pull together without pulling apart. Most of them will never meet. Some will collaborate anyway. They will fund against each other, publish against each other, build products that render the other’s moot. That is fine. That is, in fact, the point. The lever does not require the congregation to be unified to the point of trading a glove for a mitten. They can keep their identity, but they must seek out their tempering counterparts so the grip holds. This era’s opportunity requires them to be serious—serious enough that their disagreements produce something more than noise.

Most people in this congregation carry more than one denomination within them. They see merit in the Ethicist’s caution and the Accelerationist’s urgency, in the Missionary’s reach and the Purist’s velocity. But in practice, in the moment of decision, one denomination leads. That is your primary—and it is also, usually, your blind spot. The finger you point with is your strongest. It is also the one that, when it points, pulls away from the hand.

You have your denomination. Now find your counterbalance—inside yourself first, then in the room. The Ethicist needs the Purist’s velocity. The Missionary needs the Humanist’s context. The Accelerationist needs the Determinist’s long view. The Narcissist needs every other denomination, convinced, as ever, that the congregation needs them more.

Your denomination is essential—and not enough on its own. This is not a nudge to convert, but to complete. Belief left to itself becomes dogma. Belief that finds its counterbalances becomes force. We need more fingers to find each other, to close around the grip. The lever moves true when we do.

Let me start with what a jury already decided, because Donald Trump would prefer you forget it:

In 2023, a federal jury found Trump liable for sexually abusing E. Jean Carroll and for defaming her, and awarded her $5 million. In 2024, a second jury found that he defamed her, and awarded her $83.3 million in damages. Trump tried to overturn both verdicts on appeal. In September 2025, a federal appeals court rejected Trump’s argument and upheld Carroll’s victory in court. These are verdicts, tested in open court, decided by jurors, and affirmed on appeal.

Now, Trump is using the United States government to come after E. Jean Carroll and after the people who stood beside her. That includes coming after me.

It is hard enough for any woman to come forward and seek justice. It is harder still to do it against the most powerful man on earth, a man who controls the very institutions she has to walk into to be heard. And the message Trump is sending, with the machinery of the Department of Justice behind it, is unmistakable: if you challenge me and win, I will turn the full awesome weight of the United States Government against you.

What’s happening:

We’re still learning more, but here’s what’s going on: This week, the Justice Department opened a criminal investigation into E. Jean Carroll herself, an 82-year-old writer whose only offense was telling the truth about Donald Trump and being believed by jurors and the country at large. The reported pretext is a question about who helped pay her legal bills –– and as Axios and The Washington Post have reported, the probe reaches towards me, because I helped fund her case.

In other words: the government is investigating the act of helping a woman get justice.

If a President found liable for sexual assault can unleash federal prosecutors on his accuser and her backers, then any powerful abuser — anyone convicted of rape, anyone found liable for assault — becomes more encouraged to do the same. We cannot allow that to happen.

Let me put this plainly: An investigation into E. Jean Carroll’s case continues to encourage other men to keep attacking the women who have accused them of sexual assault and harassment. It also dissuades anyone from supporting these women who courageously seek justice.

It is an attempt to turn solidarity and support into a crime.

The investigation into Carroll’s case is not a one-off. It is the latest move in a sustained campaign to convert the Justice Department from a guardian of the law into an instrument of personal vengeance.

Trump’s administration secured an indictment of former FBI Director James Comey. It secured an indictment of New York Attorney General Letitia James, whose office had successfully sued Trump for fraud. Both of those cases were thrown out by a federal judge who found the prosecutor who brought them had been installed unlawfully.

The Justice Department has opened inquiries into ActBlue and the SPLC. They’ve gone after Trump’s own former officials — cybersecurity chief Chris Krebs and Homeland Security aide Miles Taylor — ordering investigations into them and stripping their clearances after each refused to echo his lies about the 2020 election. It has investigated Senator Adam Schiff. It has reportedly opened legally unprecedented inquiries into half a dozen Democratic lawmakers, a sitting governor, and a big-city mayor, several of them tied to nothing more than their work to defeat Trump.

As Trump told his supporters during the campaign: “I am your retribution.” He meant it. The Weaponization Working Group at the DOJ, the loyalist appointments, the “name and shame” threats against people who can’t be charged are the mechanics of a man who promised to use the state against his enemies and is now keeping that promise.

This is authoritarianism in action, and it is accelerating in force and breadth.

But Trump’s actions only work if we bend the knee.

Trump thinks this fabricated investigation will do three things: silence women, concentrate his power, and try to prevent me from continuing to give financial support to his opponents (whether it be sexual assault survivors, politicians seeking office, or organizations fighting on behalf of the American People). He thinks fear is contagious and courage is not.

He is wrong.

Let me be clear about where I stand. I am proud to provide financial support to candidates who will continue to stand up to Trump and stand for the American people. I am proud to have helped make the point that in America, abusers do not get to bully their way off the hook simply because they are rich or powerful or hold high office.

I will keep supporting E Jean Carroll. I will keep supporting candidates and organizations working to make our country a better place. And I will keep standing with women who find the courage to come forward.

A justice system that punishes the people who use it is not a justice system, but a protection racket. And the answer to a protection racket is not to pay the protection, but to refuse, out loud, together.

That refusal is the most American thing I can think of. We are a country founded on the radical idea that no one is above the law, not a king, not a President. The people trying to bury that idea are betting that the rest of us are too tired, too scared, or too distracted to defend it.

I’m betting they’re wrong about us. I think most Americans still believe in two simple things: the truth and justice. I think we will prove it.

In The Alliance, Ben Casnocha, Chris Yeh, and I wrote about how modern employment should be an alliance: An explicitly, mutually-beneficial relationship where both employer and employee agree on a specific tour of duty, which, if successfully concluded, leaves the business more successful and the employee more valuable, both to the business and to other potential employers. These alliances aren’t legal contracts, but they are moral promises that are not to be made lightly, and even less lightly broken.

Like its predecessor, The Startup of You, The Alliance was a response to changes to workplace and workforce norms that were driven by the rise of the internet and global supply chains and markets, and the increasing competition and pace of adaptation that ensued. Where the former looked at this new environment from the perspective of individuals and how they should adapt to it, The Alliance took an institutional viewpoint. What was the role of employers in this new world? How might they operate to achieve the best long-term outcomes for themselves, their employees, and society at large?

Even as industries, corporations, and individual workers all were experiencing greater flux than ever before, we asserted, trust was still the glue holding institutions and careers together. But new degrees of clarity, explicit intention, and articulated mutual alignment were also necessary.

We also argued that the responsibility to initiate alliances naturally falls on institutions rather than individuals; they’re the entities that hold the most power and set the terms of engagement of the working relationship. Amongst other things, that means defining “tours of duty” explicitly instead of just working under the default but not contractually bound premise of “lifetime employment,” or investing in employees’ growth even when that will also increase their value to competitors. With explicit clarity and intention replacing unspoken assumptions, we reasoned, employers and employees could continue to achieve productive outcomes even amidst the new uncertainties and rapid changes of 21st century work life.

As industries, institutions, and individuals begin to adapt to the ways that AI is reshaping corporate structures, workflows, and capital allocation, to name just a few dimensions, Alliance-thinking is more necessary and valuable than ever.

To date, the AI adaptation story is most often told through the lens of company layoffs, and it’s understandable why. First, there has been a steady stream of them over the last two years, and the certainty of more to come. Second, layoffs obviously upend the lives of thousands of people at once, in ways that are both immediately felt and often long-lasting. But it’s also true that layoffs are just one facet of the larger AI adaptation story. There are AI hiring waves too, along with AI start-up booms and AI investment flowing into both new ventures and internal transformation.

Recognizing this is key to an optimal Alliance framework, because the framework works best when all parties can see the full landscape of what’s possible, not just what’s threatening. An individual employee who focuses mostly on layoffs is likely to make defensive choices. In contrast, one who sees the fuller picture—including the booms, the new roles, and the investment in human capability—is better positioned to negotiate a tour of duty that matters.

The same is true for companies. Those who frame AI purely as a cost-cutting exercise will miss the alliance-building moments that determine who actually thrives.

START WITH HONESTY AND TRANSPARENCY

In February, Jack Dorsey announced that Block would be laying off over 4,000 employees, taking the company from over 10,000 to under 6,000 people. “The intelligence tools we’re creating and using, paired with smaller and flatter teams,” he wrote, explicitly attributing the cuts to AI, “are enabling a new way of working which fundamentally changes what it means to build and run a company.” He also explained that he believed a single deep cut, all at once, was preferable to repeated waves of smaller layoffs.

In the weeks that followed, Atlassian, Snap, Oracle and other companies made similar moves. In early May, Brian Armstrong of Coinbase announced a 14% layoff. In an email he sent to employees and then posted on X.com, he sounded similar themes as Dorsey. “We are not just reducing headcount and cutting costs, we’re fundamentally changing how we operate: rebuilding Coinbase as an intelligence, with humans around the edge aligning it.”

As a general principle, I believe that AI wins by doing more, not spending less, and the same companies that are laying employees off now may find themselves understaffed if their competitors use AI to do more with the same number of employees. Simple cost-cutting often loses out to greater value creation; Costco’s market capitalization is 10X those of the two most valuable “dollar store” retailers combined.

At the same time, ignoring or dismissing the coming impact of AI is likely more dangerous than overreacting to it. That’s because it’s already clear how quickly it can change things. The stock price of online homework helper Chegg has declined nearly 90% since the introduction of ChatGPT. Hoping to avoid a similar outcome, stock image providers Getty Images and Shutterstock announced a plan to merge in January 2025 to increase their scale in the face of AI, and their respective stock prices have still declined significantly since then. (The merger is still under review.) Tech companies like Google report that over 50% of their new code is now written by AI rather than human software developers.

Because AI can be applied in so many ways, virtually every industry and company, and virtually every individual worker, will feel its impact eventually. In the same way that the internet eventually reshaped industries, companies, and jobs that seemed completely outside the domain of “high tech” until the internet happened, the same will be true of AI. Only this time the changes will occur faster and with more immediate impact, because the internet had to build its own distribution before it could reshape anything—fiber, broadband, smartphones, cloud, payments. AI arrives on top of all of that infrastructure, with billions of people already one click away.

When companies announce AI-driven layoffs, there is, and will continue to be, skepticism about their true motivations—and some of this skepticism is warranted. AI will be the fall guy for companies that over-hired in the pandemic. It will serve as a potentially preferable rationale when cuts are really driven by some activist investor pushing for lower headcounts. It can function as a way to prompt Wall Street vibe traders to pump stock prices.

But it’s also true that it would be a major failure of leadership, for those in the C-suites, especially at leading tech companies Block and Coinbase, not to be thinking about how to become AI-first, sooner rather than later, and making moves to accomplish this. Because the competitive pressures are real. The capability gains are real. And the cost of moving too slowly is usually greater than the cost of moving too aggressively. So there will be more layoffs, from more companies, and many of them will be genuinely necessary rather than opportunistic.

But that doesn’t mean companies should feel entitled to treat this moment of technological upheaval as an instance of force majeure: An extraordinary circumstance that suspends ordinary obligations. We didn’t see this coming. The technology changed faster than anyone predicted. We have no choice.

That framing is convenient, and in narrow legal terms it may even be accurate. But it is not a legitimate ethical posture. Indeed, if you believe a company does in fact have moral obligations to employees with whom they’ve built alliances, these obligations shouldn’t be allowed to evaporate when the environment gets harder. If anything, they intensify.

And not just because it’s the ethical thing to do. Companies hoping to earn the trust of new generations of talent—and the loyalty of their alumni networks—are ones that treat AI-driven restructuring as a test of Alliance principles, not an exemption from them.

This starts with honesty and transparency around potential cuts and restructuring. On March 31st, thousands of Oracle employees were notified by email not just that they were being laid off, but also that this would be their “last working day.” Obviously such instant layoffs have existed long before AI. And sometimes there are rationales for them. A firm that cannot make payroll in sixty days cannot offer sixty days of notice. If giving employees a heads-up would trigger a run—with clients fleeing, or lenders calling in debt—then the notice period can accelerate the very collapse it was meant to soften. In virtually every instance, giving employees advance notice is likely to create unrest and concern internally and, in an age when it takes a half dozen or so clicks to turn a confidential memo into a headline at Business Insider, externally too.

So there’s a case to be made that secrecy, even from employees, until a public announcement can be made, is both a compassionate and a strategic thing to do. But how many CEOs who argue that secrecy is the most compassionate way to approach layoffs by avoiding uncertainty would want their boards to withhold their intentions to fire them, under the banner of compassion?

Consider, also, the strategic nature of these AI layoffs. For the most part, they’re not the typical “defensive” or reactive layoffs, with CEOs citing market conditions and a pressing need to cut costs to ensure survival. Instead, most can be characterized as explicitly “offensive” layoffs, with framing that emphasizes the company’s financial health and its proactive plan to increase capacity and speed of execution by leveraging AI in ways that also make a portion of the workforce redundant.

In the instances where this is actually true, same-day layoff announcements are not just a moral failure but a managerial one too. If the ultimate goal is informed AI adaptation and adoption, then input from all parties is crucial for both a well-scoped AI plan and mutually beneficial options. Before going straight to layoffs, managers should explore how their company can use AI to expand their business by doing more instead of just spending less. And part of this process should include asking employees who have deep knowledge of their work what they think are the best ways they could use AI to leverage their work.

In best-case scenarios, this approach will help the company make more effective decisions about how to integrate AI into its business processes, and also give employees an opportunity to find ways to make themselves too valuable to part with. In instances where layoffs do in fact happen, it will allow for more humane and dignified transitions. It may create management challenges, but employees deserve appropriate time to process the news. They should get the chance to transition projects and knowledge properly, and secure references and recommendations while managers and colleagues are still close at hand. A longer runway for beginning a job search and making practical decisions about finances, health coverage, and other crucial life details, will make a difference in how they’re able to navigate this transition.

Once again, there’s also a strategic component to that as well as a moral one—because ideally alliances don’t end when employment contracts do. In a world in which nearly every person’s career and professional connections are a public record on LinkedIn, a company’s former employees are often some of its most important brand ambassadors. A company whose former employees thrive is a company that will have a much easier time attracting future employees.

Making the effort to conclude your employees’ tours of duty in an honorable way, despite the circumstances, delivers long-term benefits by maintaining a strong alliance long after the employment relationship ends. Your business will be able to call on allies for valuable market intelligence, partner and customer introductions, and recruiting assistance. And of course maintaining a strong alliance dramatically increases the chances you can re-hire that former employee in the future.

LAYOFF STORIES ARE ALSO OFTEN HIRING STORIES

As much as AI layoffs dominate media headlines, they’re far from the whole story. LinkedIn’s 2026 Jobs on the Rise report ranked AI Engineer as the single fastest-growing job title in the United States—and three of the other four spots in the top five went to AI-related roles as well. Looking out further, the U.S. Bureau of Labor Statistics projects that employment in computer and information technology occupations will “grow much faster than the average for all occupations from 2024 to 2034,” with roughly 317,700 openings per year on average over the decade.

Some of the most aggressive layoff stories are also aggressive hiring stories. Accenture offers a particularly clear illustration. In its fiscal 2025 earnings call, the company disclosed an $865 million restructuring program that included more than 11,000 layoffs in a single quarter, with over 22,000 roles eliminated across two quarters. In the words of CEO Julie Sweet, the company was “exiting on a compressed timeline, people where reskilling, based on our experience, is not a viable path for the skills we need.”

But Accenture was simultaneously rapidly expanding. Between fiscal 2023 and September 2025, Accenture nearly doubled its AI and data specialist headcount from 40,000 to 77,000. It had trained more than 550,000 employees in generative AI, with plans to train another 700,000-plus in agentic AI. And it told investors it expected its overall headcount to grow in the next fiscal year, calling the strategy “rapid talent rotation.”

This may sound like the latest euphemism for “layoffs,” but it really is a description of something different. Layoffs are subtractive; rotation is directional. More specifically, it’s swapping one skill mix for another, often while growing in absolute terms, as Accenture is doing.

But of course even talent rotation can take many forms. Take the numbers associated with Accenture’s efforts to restructure. On the one hand, 22,000 employee “exits.” On the other, 550,000 employees trained in generative AI. From a broad perspective, at least, this ratio suggests that Accenture is operating with a strong Alliance mindset: It’s making a significant investment in existing employees, and helping them adapt to the new conditions that will define the company’s future.

It’s in situations like these where the concept of alliance as a real substantive principle governing the relationship between employers and employees becomes concrete rather than abstract. Ideally, as more and more companies begin to adapt to the opportunities and challenges that AI presents, they’ll follow these general tenets. Give existing employees options on new roles. Invest in reskilling before assuming it won’t work. Define “viable path to the skills you need” generously rather than narrowly. And when an exit is genuinely necessary, treat it as a transition in a long-term relationship rather than the end of one—through severance that reflects tenure, through introductions to other employers, through inclusion in the alumni network that is one of the most underrated assets any company has.

None of this is charity. It’s the recognition that the employees being rotated out today are the alumni, customers, partners, and recruits of tomorrow, and that how a company treats people on the way out is the single most legible signal it sends to the people it’s trying to attract on the way in.

BUILDING ALLIANCES IN MARKET WHERE ALLIES COMPOUND

There’s also a final factor in the larger AI adaptation story: The current sustained boom in new business formation. Americans filed roughly 5.5 million applications to start businesses in 2023, the most in any year since the Census Bureau’s Business Formation Statistics series began in 2004. Since then, applications have stayed near that elevated level every year. More broadly, business applications have averaged roughly 442,000 a month since May 2020, or nearly double what they averaged from 2005 to early 2020.

The Economist and others have argued that COVID-era job loss and government stimulus programs are the key instigators of this phenomenon, but it also plays out in a larger technological context. First there was cloud infrastructure that removed the need to own servers. E-commerce and payments platforms like Shopify and Stripe followed, along with no-code and low-code builders like Zapier and Bubble, not to mention incorporation tools like Stripe Atlas, which transformed company formation from a somewhat complex legal project into a weekend to-do. Now AI enters the mix, in the most potentially consequential way yet, because of how it can compress not just setup but the work itself—coding, design, customer support, marketing copy, financial modeling—into tasks a single person can plausibly direct.

Consider, for example, how the share of new startups with solo founders rose from 23.7% in 2019 to 36.3% by mid-2025. This is correlation, not proof of causation. But Carta, the equity-management platform that compiled the data, attributes the shift substantially to AI, noting that the technology has “expanded what individuals can accomplish in a finite amount of time, making it more feasible for a single founder to both build and sell.”

Stripe, the payments company, sees a similar pattern in how fast new companies can now grow —and reaches for the same explanation. Per its annual report, twice as many new businesses on Stripe’s platform hit $10 million in annual recurring revenue within three months of launch in 2025 as did in 2024—a shift the company attributes to “advances in large language models.”

These two stories—the layoffs and the founder boom—are usually told separately, but they’re tightly connected. Some of the people being laid off from large tech companies are creating startups. Some of the AI capabilities that are making large companies leaner are simultaneously making it possible for two people in a garage to compete with much larger incumbents.

And all of this is where Alliance thinking expands beyond the walls of any single company. The traditional alliance is bilateral—company and employee, with mutual investment over a defined tour of duty. But in an economy where employees are increasingly likely to become founders, and founders are increasingly likely to become collaborators, partners, customers, and acquirers of their former employers, the alliance starts to look more like a network than a contract. The employee who leaves to start a company isn’t exiting the relationship; they’re changing its shape. The company that helped that person grow—through real investment, honest tours of duty, and a dignified parting—has built something more valuable than a former employee. It has built an ally in a market where allies compound. Or to put it another way, how you treat your people, how you part with them, and how you stay connected to them after, are, more than ever, no longer just an HR concern. They’re a competitive one.

Ten years ago, if you’d told me the Pope would become one of the most thoughtful voices in the global technological debate, I would have raised an eyebrow.

Speaking to students in Cameroon, Pope Leo warned that AI risks “hollowing out” real human relationships—creating digital environments where people are optimized into bubbles, unable to distinguish truth from simulation. Around the same time, he was publicly criticizing the U.S. administration’s military threats against Iran, prompting President Trump to fire back on Truth Social and call him “weak.” Trump then posted an AI-generated image of himself as Jesus, and deleted it after backlash from across the political spectrum.

It’s a strange constellation of events. But underneath it is a single, important inflection point: world leaders—religious and political—are now openly fighting about what it means to be human in the age of AI. And Pope Leo, more than most, is asking exactly the right question. Pope Leo even chose his name as a deliberate echo of the industrial-revolution-era Pope Leo XIII, who confronted the human consequences of the last great technological upheaval. The name itself is a thesis: this is the moment, and we have to steer it.

Pope Leo is right to say that if we build AI that flatters and isolates (e.g., “you’re so spectacular, you’re right all the time, don’t talk to other people, only talk to me”) then we have made humanity smaller. Some systems already drift in that direction. But this is a design choice, not a foregone conclusion. For example, when we built Pi at Inflection, we made the opposite call by default, and built the chatbot to encourage the user to find more real-world social connections. The idea was that EQ matters as much as IQ—and that the best AI doesn’t replace humans with a chatbot. It helps you show up more fully, for more people.

I’m an optimist, and I try not to let that descend into naivete. The “natural” trajectory of any powerful technology isn’t utopia; it’s whatever the most motivated incentives drive it toward. Those incentives come from a set of quasi-religious beliefs that exist within Silicon Valley.

The religious factions that end up winning and thus ingraining their values into the products we use will decide whether or not we have a smoother transition into the beneficial world of AI, or if, like the Industrial Revolution, humanity makes it through, albeit with unnecessary chaos and collapse of institutions. Getting to the positive future requires a full-court press from all those who believe it is possible. That means builders, investors, influencers, and governments building demand for what good looks like, and following through on their promises.

So when Pope Leo speaks, I’m not hearing him as a cultural curiosity in the AI conversation. I’m hearing him ask the question every technologist should have on their desk, at their fingertips, and top of mind: What is the human answer here?

PS: This is my third substack! I’ve been so grateful for the thoughtful comments so far, and plan to respond to as many as I can (albeit slowly). Feel free to let me know your thoughts on Pope Leo or recommend some new topics you’d like to see me write on!

With each electric streetlamp that replaced its gas-lit antecedent in America’s cities at the turn of the 20th century, urban freedom took a nimble leap forward. Poorly laid cobblestones became far less lethal. Street signs remained legible when you really needed them, like after you’d had a few too many at a saloon in a quarter of town you didn’t know well. A shopgirl could return home more safely from her job at 8 PM – or explore the city’s new wonders herself. Over time, a growing “night economy” created more entertainment options, evening classes at local colleges, a wide range of new jobs, and entirely novel patterns of living.

And that was just one aspect of electrification’s impact.

Spark-free motors in Minnesota grain elevators helped store-bought bread evolve from an occasional luxury into an affordable and time-saving staple for millions of Americans. Incandescent light bulbs in homes meant children could spend hours at night improving their reading skills without the adult supervision that kerosene lamps required. The X-ray machine turned electricity from a lighting tool into a new way to diagnose potential issues without cutting a patient open. Suddenly, the first steps of care were no longer more painful than the original complaint.

But before electrification did any of that?

It gave us slop.

Picture ordinary butter dishes, made from uranium glass, looking like alien artifacts in retail display windows as electric arc lights made them phosphoresce like neon. Picture the stage at Niblo’s Garden, a Broadway theater, where Thomas Edison himself helped equip hundreds of chorus girls and extras with electric lights for a show in 1883 that also featured a replica of the newly opened Brooklyn Bridge “with lighted boats running under it.” Picture a giant green pickle, 43-feet-wide, affixed to the side of New York City’s Cumberland Hotel and casting an unprecedented 12-kilowatt glow. According to Signs and Wonders, a 1998 history of the early electric signs known as “spectaculars,” the giant Heinz pickle created “an eerie visual throb of green and white that nightly invaded” the homes of people who lived blocks away.

Not everyone viewed this as progress.

In his 1878 essay, “A Plea For Gas Lamps,” the novelist Robert Louis Stevenson decried electric light as “unearthly” and “obnoxious,” a phenomenon that “should shine only on murders and public crime, or along the corridors of lunatic asylums, a horror to heighten horror.” (That Stevenson was cosmopolitan enough to shroud his anxieties in fainting-couch hyperbole hardly discounts those anxieties. Walter Benjamin himself read the essay as a prescient meditation on what we lose when technology surpasses “human” scale.) “The new illumination…” the writer G.K. Chesterton exclaimed in a 1927 essay, “has made people weary of the way of proclaiming great things, by perpetually using it to proclaim small things.”

But it wasn’t just literary elites decrying electricity and how it created what the cultural historian Wolfgang Schivelbusch would eventually describe in his 1995 book Disenchanted Night as “a nightmare of light from which there was no escape.” “Residents of [Madison Square] shuddered at the sight” of the electric Heinz pickle, a New York Times reporter recalled in 1936. “Old folk averted their eyes in shame as they passed the Square…”

But slop also gave electricity its first Eureka moment – that flash of illumination when people realized that these new carbon-filament bulbs weren’t just a steadier and less sooty form of artificial light. What slop made clear was that electricity was an entirely new way of powering human expression and inhabiting the world.

This became even more self-evident as the slop grew more ambitious. By the early 20th century, urban rooftops erupted with state-of-the-art brain rot like “Leaders of the World,” a 2,000-square-foot sign on top of New York City’s Hotel Normandie. Depicting a Roman chariot race, it used nearly 20,000 light bulbs controlled by high-speed mechanical switches to stage a 30-second loop of stampeding horses and spinning wheels that was so mesmerizing it compelled some onlookers to view it for hours on end.

Attractions like these gave the Edison General Electric Company and other pioneers of the spark something far more valuable than approval from the likes of Robert Louis Stevenson: Early Adopters.

This was crucial, because like most new technologies, electricity was not an easy sell. Sure, it was novel and even downright magical. But it was also unproven, volatile, and complicated. All of which made its best prospects on paper its least likely converts.

Gilded Age steel mills and textile factories had massive energy demands. And the vast corporate resources to pay for power in bulk. But for enterprises of such unprecedented scale, the prospect of electrification also created a risk signal as bright as a 43-foot electric pickle. Because factories in that era were built around a single steam engine and the dense networks of overhead iron shafts and spinning leather belts that fanned out from them, retooling for electrification was far more complex than just flipping a switch. Instead, it necessitated massive investments for unclear rewards. The inefficiencies and dangers of steam power, in contrast, were the devil these enterprises already knew.

But where factory owners saw risk, others saw a dazzling competitive advantage. In 1886, Boston’s Bijou Theatre was the first customer of the Edison Electric Illuminating Company of Boston. Because early power stations were notoriously inefficient at running under partial loads–burning the same amount of coal whether one bulb was on or a thousand–customers like the Bijou that consumed massive and predictable amounts of power at once were crucial to the industry’s development. A single theater with hundreds of lamps drew as much current as dozens of homes, and they drew it at the exact time the grid needed a steady “load” to remain stable.

In this way, electric slop played a key role in funding and scaling the wires, underground cables, and massive dynamos that would eventually service both factories and individual homes.

IN SLOP WE TRUST
As revolutionary as electrification was, it wasn’t as if Thomas Edison had invented centralized energy supply, metered usage, and monthly utility bills along with the incandescent bulb. All of that already existed, in the form of the vertically integrated leviathan known as the gas industry.

By the 1880s, the industry’s leaders had established local monopolies in most large American cities. Each had its own manufacturing facilities for converting coal into gas and storing it in quantity. Each had laid thousands of miles of cast-iron pipes beneath the streets for pressurized delivery to businesses and homes. They all boasted massive service crews, lucrative municipal street-lighting contracts, and lobbyists devoted to assuring public officials that flickering gas-light was the key to a bright future.

In addition to wilting houseplants and tarnishing the family silver, gas lighting also regularly caused house fires and even death. In that era, the gas being piped into homes and businesses still contained carbon monoxide as a constituent ingredient, so any crack in a pipe, valve, or fixture could potentially lead to tragedy. As late as 1925, when electricity had been in common use for a generation, there were still days when gas might claim fifteen lives in a single day in New York City alone.

Electricity brought its own dangers too, especially initially, when electricity providers were still iterating on and standardizing equipment and improving safety protocols. In the 1880s, the urban sky was filled with uninsulated, high-voltage wires strung haphazardly across rotting wooden poles. During winter storms, these wires frequently snapped, turning city sidewalks into electrified minefields. Fuses were primitive, and leakage from poorly grounded wires could cause the very walls of a home to hum or spark.

In December 1889, a cartoon in Judge, a satirical weekly, showed urban residents wearing “non-conductor” rubber suits to avoid accidents caused by the electric power lines that were beginning to crowd the streets overhead. Two months earlier, Judge’s cover had depicted an accident that had occurred in Manhattan in March 1889 when, as a Substack called The Forgotten Files recounts, “a Western Union lineman named John Feeks was electrocuted while trying to repair a broken telegraph wire.”

Even Thomas Edison himself weaponized the fears born of these early realities. In a bid to protect his direct-current empire from the more efficient alternating current (AC) championed by Nikola Tesla and George Westinghouse, he launched one of history’s most macabre PR campaigns. Under the pretext of objective safety research, Edison’s team subjected stray animals to both forms of electricity. By carefully calibrating the tests to show that low-voltage AC could kill a dog while DC merely singed it, Edison hoped to brand his rival’s technology as uniquely lethal—even lobbying newspapers to adopt the verb “to Westinghouse” as a synonym for “to electrocute.”

There was the perceived waste of it all. Critics looked at Edison’s massive Pearl Street Station—the “server farm” of its day—and saw both a substantial drain on natural resources and a quality-of-life nuisance, whose “Jumbo” dynamos reportedly vibrated with enough force to rattle the windows of neighboring buildings. Why burn tons of coal just to power a few blocks of Wall Street and a handful of playhouses, when existing (if occasionally explosive) gas networks already provided seemingly everything electricity could provide?

In the short term, at least, slop stood as the electric light industry’s most effective rebuttal. By placing the miracle of the electron inside highly visible advertising signs and other prominent public displays, electrification’s champions made their product spectacular, then familiar, then trustworthy. Night after night, as such installations cycled through loops of motion and light, the “death at every street corner” apocalypse depicted by Judge and other news media outlets from that era.

Instead, each new display seemed to prompt another: Shimmering department store wonderlands. Ocean-front amusement parks where endless strings of lights turned a night on the town to a day at the beach. Spectacular billboards whose bulbs burned even brighter than the stars in the sky, imbuing the products and services they advertised with celestial grandeur.

By 1891, there were over 1,100 central stations in the U.S., the most prominent of which were located in entertainment districts or high-end commercial hubs.

This early infrastructure set the stage for a transformation so dramatic it would change the very boundaries of human possibility. In a matter of decades, what started out mostly as a conduit for slop turned into civilization’s most important platform, re-wiring daytime and nighttime, seasons, distance, global communications, and our very definitions for progress, the good life, and modernity.

TODAY’S RACE TO PEAK PICKLE

Today, critiques targeting AI and the physical infrastructure that makes it possible echo those that were once made about electrification. Data centers, in particular, have emerged as a lightning rod for current concerns around the economic, environmental, and cultural impacts of AI. In February, Time magazine cited research from a group called Data Center Watch that concluded that activists in Virginia, Indiana, and Arizona “stalled $98 billion in data-center projects in the second quarter of 2025 alone.”

That local communities possess an absolute right to explicitly approve each new instance of infrastructure that stands to shape their lives is generally not legally true. But it is a widely held instinct in democratic societies, and as such, it’s morally, politically, and operationally strategic for developers to heed it. After all, big infrastructure projects do materially impact civic identity, shared natural resources, economic opportunities, physical landscapes, and overall quality of life. If AI developers and data center operators expect to occupy a mutually productive place in the local landscape, it’s crucial that we move beyond the logic of “inevitable progress” and provide more than just boilerplate promises. Fair-market offers and a commitment to long-term coinvestment should be the table stakes here.

Even then, many communities are likely to reject such opportunities. For any community contemplating this decision, however, the power to say “Yes” is just as important as its right to say “No.”

There are obviously legitimate reasons for either decision, and in a future essay, I’ll make the case for why communities should choose data centers. For now, I’ll just reinforce the idea that slop obviously plays a key role in AI’s evolution too. We were promised AI cures for cancer. Novel insights that crack the code for limitless cold fusion. Instead what we’re getting, critics insist, is the 21st century version of the Heinz Pickle. Carcinomas persist, but at least we’ve got fake security-cam footage of cats playing late-night violin concertos on people’s front porches.

But this view mistakes the “pickle” for the “grid.” Just like the early 20th century’s central stations, today’s data centers exist as the foundational infrastructure that will ultimately make general-purpose intelligence as abundant and accessible as electricity became. Some of today’s high-performance GPUs, advanced cooling systems, and zero-latency fiber optics might be enabling little more than slop now, but if history’s any indication, that’s a sign of imminent success, not concern.

The faster we get to Peak Pickle, the more likely it becomes that we will see the 21st-century equivalents of the X-ray machine and affordable bread. In this era, that looks like personalized tutors for every child, carbon-capture systems optimized by real-time physics models, and drug discovery that moves at the speed of software. To slow-walk the build-out of data centers because we find their current output little more than “slop” is as short-sighted as it would have been to ban electricity in 1900.

In the end, it’s exceedingly hard to ration and constrain your way to a revolution. Massive transformation typically requires widespread attainable access to some new resource. Because this is true, our current moment is no doubt messy, sometimes wasteful, and frequently absurd. We are immersed in silicon pickles.

But if we want a future where AI solves the intractable problems of biology and physics, we should view slop not as a diversion or failure, but rather as a sign of progress. In the early days of electrification, the road to the hospital ran through the theater. Today we are witnessing the spectacular and ultimately substantive sequel to that story.

Silicon Valley has a God complex. The critics are not wrong about that. They are just wrong about which god. Money alone sends you to Wall Street. Power alone sends you to Washington. Something else entirely sends you to a garage in Menlo Park: faith.

Faith in science and reason, yes—but more than that. Faith in creation itself. Silicon Valley is a religious movement. Its central doctrine: no force in history has done more to improve human life than technology, and no institution is better positioned to build it for scale than the company. Not the state. Not the church. Not the NGO. The company, deploying scalable technology, is the primary modern instrument of human progress. Technology represents roughly a tenth of the U.S. economy and has driven a third of its growth over the last decade—and that’s before counting every other industry it now powers.

GDP measures outputs, not lives, some will counter. Fair enough—but look globally at where technology took hold and what happened. China practically skipped the branch-banking era—Alipay and WeChat brought hundreds of millions into the payment economy via phone. Rwanda deployed drones to deliver blood to hospitals, cutting maternal mortality from hemorrhage by more than half at served sites. No-till farming, biotechnology, and precision agriculture made Argentina a top-five food exporter; soy production alone grew roughly fivefold between 1990 and 2010. Those numbers are the scripture.

Like any religion of age, Silicon Valley has its denominations. The Purists hold the orthodox line: the lever is code alone—everything else is noise. The Humanists argue the lever needs society as its fulcrum, or it lifts nothing. The Missionaries believe the lever is only worthwhile when it lifts everyone, although mostly fine with elevating some more than others. The Venture Capitalists believe the lever is only worth pulling if the return justifies the investment. The Narcissists believe the lever exists to elongate their own shadow. The Accelerationists believe the lever must be pulled without hesitation or respite. The Ethicists believe it must be pulled carefully, or not at all. The Determinists believe the lever pulls itself; our role is not to direct it, but to hold on.

On this, the denominations agree: the lever must be funded, but funding is not why you pull it. Elsewhere, these factions argue. They schism. They sometimes ally. People might belong to several sects. But the canon holds: technology is the Archimedean lever of human flourishing. The struggle over how to handle the lever has a way of slowing the pull—and, left long enough, of making us wonder whether we should be pulling at all. Opposition to a faction is not a case against the creed. Reject the Accelerationists, distrust the Capitalists—and still belong to the faith. The faith is big enough to hold the argument. The risk is not that we build too boldly. The risk is that we stop believing building matters—that we trade the possible for the permissible, creation for consensus. The instinct to build, to make, to reach past what exists toward what is possible is not a Silicon Valley faith. It is the oldest human one.

Silicon Valley did not create the urge to create. It inherited it and supercharged it. We are all, at root, homo techne. Tool-users, toolmakers, tool-evolved. The hand shaped the stone, and the stone shaped the hand. Our tools do not make us less human. That fear is as old as tools themselves. Socrates worried that writing would limit memory. And in some ways, it did. And we flourished and became something larger. As each technology wave expands, so can we. We do not lose ourselves. We transfigure.

No other animal has rewritten its own cognitive architecture through what it builds. The printing press did not just distribute information. It restructured how Europeans thought about authority, scripture, and the self. Luther’s 95 Theses spread from Wittenberg across German-speaking Europe within weeks. Within a century, Protestant northern Europe had ridden the press toward the Enlightenment; the Ottoman Empire, where religious authorities suppressed Arabic-script printing for over two centuries, found itself on the other side of a widening gap. Every era has its press, and every era has those who underestimate what it means to let it pass.

It’s easy to get caught up in product releases and cycles, and forget that every technology traces this spiritual arc. You are born into it, converted, or you recede. So it is written. The car, the plane, the personal computer, the mobile phone: each arrived as a disruption and became, within a generation, invisible infrastructure. Amen. Today, more than 4 billion people carry a networked computer in their pocket. The majority of them are in the developing world. They did not get a library or a bank branch first. They got a phone. The phone is the library and the bank.

Technology’s arc bends toward access. But it does not bend on its own. Behold the technologist’s creed: through building to scale, we extend what is possible. Possible for whom has always been the humanist’s question. It is now everyone’s. The more the machines build and scale, the more that matters.

Our creed does not ask for certainty. It never has. Indeed, the real Archimedean levers always involve risk and uncertainty. Instead, the creed asks for the courage to begin. To the builders: unite the grip or lose the lift. To everyone else: no lever in history has mattered more—and it only moves if you reach for it. Hold the lever, find the fulcrum, and rise.