What exactly is Donald Trump doing?

Since taking office, he has reduced his administration’s effectiveness by appointing to essential agencies people who lack the skills and temperaments to do their jobs. His mass firings have emptied the civil service of many of its most capable employees. He has defied laws that he could just as easily have followed (for instance, refusing to notify Congress 30 days before firing inspectors general). He has disregarded the plain language of statutes, court rulings, and the Constitution, setting up confrontations with the courts that he is likely to lose. Few of his orders have gone through a policy-development process that helps ensure they won’t fail or backfire—thus ensuring that many will.

In foreign affairs, he has antagonized Denmark, Canada, and Panama; renamed the Gulf of Mexico the “Gulf of America”; and unveiled a Gaz-a-Lago plan. For good measure, he named himself chair of the Kennedy Center, as if he didn’t have enough to do.

Even those who expected the worst from his reelection (I among them) expected more rationality. Today, it is clear that what has happened since January 20 is not just a change of administration but a change of regime—a change, that is, in our system of government. But a change to what?

[Graeme Wood: Germany’s anti-extremist firewall is collapsing]

There is an answer, and it is not classic authoritarianism—nor is it autocracy, oligarchy, or monarchy. Trump is installing what scholars call patrimonialism. Understanding patrimonialism is essential to defeating it. In particular, it has a fatal weakness that Democrats and Trump’s other opponents should make their primary and relentless line of attack.

Last year, two professors published a book that deserves wide attention. In The Assault on the State: How the Global Attack on Modern Government Endangers Our Future, Stephen E. Hanson, a government professor at the College of William & Mary, and Jeffrey S. Kopstein, a political scientist at UC Irvine, resurface a mostly forgotten term whose lineage dates back to Max Weber, the German sociologist best known for his seminal book The Protestant Ethic and the Spirit of Capitalism.

Weber wondered how the leaders of states derive legitimacy, the claim to rule rightfully. He thought it boiled down to two choices. One is rational legal bureaucracy (or “bureaucratic proceduralism”), a system in which legitimacy is bestowed by institutions following certain rules and norms. That is the American system we all took for granted until January 20. Presidents, federal officials, and military inductees swear an oath to the Constitution, not to a person.

The other source of legitimacy is more ancient, more common, and more intuitive—“the default form of rule in the premodern world,” Hanson and Kopstein write. “The state was little more than the extended ‘household’ of the ruler; it did not exist as a separate entity.” Weber called this system “patrimonialism” because rulers claimed to be the symbolic father of the people—the state’s personification and protector. Exactly that idea was implied in Trump’s own chilling declaration: “He who saves his Country does not violate any Law.”

In his day, Weber thought that patrimonialism was on its way to history’s scrap heap. Its personalized style of rule was too inexpert and capricious to manage the complex economies and military machines that, after Bismarck, became the hallmarks of modern statehood. Unfortunately, he was wrong.

Patrimonialism is less a form of government than a style of governing. It is not defined by institutions or rules; rather, it can infect all forms of government by replacing impersonal, formal lines of authority with personalized, informal ones. Based on individual loyalty and connections, and on rewarding friends and punishing enemies (real or perceived), it can be found not just in states but also among tribes, street gangs, and criminal organizations.

In its governmental guise, patrimonialism is distinguished by running the state as if it were the leader’s personal property or family business. It can be found in many countries, but its main contemporary exponent—at least until January 20, 2025—has been Vladimir Putin. In the first portion of his rule, he ran the Russian state as a personal racket. State bureaucracies and private companies continued to operate, but the real governing principle was Stay on Vladimir Vladimirovich’s good side … or else.

Seeking to make the world safe for gangsterism, Putin used propaganda, subversion, and other forms of influence to spread the model abroad. Over time, the patrimonial model gained ground in states as diverse as Hungary, Poland, Turkey, and India. Gradually (as my colleague Anne Applebaum has documented), those states coordinated in something like a syndicate of crime families—“working out problems,” write Hanson and Kopstein in their book, “divvying up the spoils, sometimes quarreling, but helping each other when needed. Putin in this scheme occupied the position of the capo di tutti capi, the boss of bosses.”

Until now. Move over, President Putin.

To understand the source of Trump’s hold on power, and its main weakness, one needs to understand what patrimonialism is not. It is not the same as classic authoritarianism. And it is not necessarily antidemocratic.

[Read: Trump says the corrupt part out loud]

Patrimonialism’s antithesis is not democracy; it is bureaucracy, or, more precisely, bureaucratic proceduralism. Classic authoritarianism—the sort of system seen in Nazi Germany and the Soviet Union—is often heavily bureaucratized. When authoritarians take power, they consolidate their rule by creating structures such as secret police, propaganda agencies, special military units, and politburos. They legitimate their power with legal codes and constitutions. Orwell understood the bureaucratic aspect of classic authoritarianism; in 1984, Oceania’s ministries of Truth (propaganda), Peace (war), and Love (state security) are the regime’s most characteristic (and terrifying) features.

By contrast, patrimonialism is suspicious of bureaucracies; after all, to exactly whom are they loyal? They might acquire powers of their own, and their rules and processes might prove obstructive. People with expertise, experience, and distinguished résumés are likewise suspect because they bring independent standing and authority. So patrimonialism stocks the government with nonentities and hacks, or, when possible, it bypasses bureaucratic procedures altogether. When security officials at USAID tried to protect classified information from Elon Musk’s uncleared DOGE team, they were simply put on leave. Patrimonial governance’s aversion to formalism makes it capricious and even whimsical—such as when the leader announces, out of nowhere, the renaming of international bodies of water or the U.S. occupation of Gaza.

Also unlike classic authoritarianism, patrimonialism can coexist with democracy, at least for a while. As Hanson and Kopstein write, “A leader may be democratically elected but still seek to legitimate his or her rule patrimonially. Increasingly, elected leaders have sought to demolish bureaucratic administrative states (‘deep states,’ they sometimes call them) built up over decades in favor of rule by family and friends.” India’s Narendra Modi, Hungary’s Viktor Orbán, and Trump himself are examples of elected patrimonial leaders—and ones who have achieved substantial popular support and democratic legitimacy. Once in power, patrimonialists love to clothe themselves in the rhetoric of democracy, like Elon Musk justifying his team’s extralegal actions as making the “unelected fourth unconstitutional branch of government” be “responsive to the people.”

Nonetheless, as patrimonialism snips the government’s procedural tendons, it weakens and eventually cripples the state. Over time, as it seeks to embed itself, many leaders attempt the transition to full-blown authoritarianism. “Electoral processes and constitutional norms cannot survive long when patrimonial legitimacy begins to dominate the political arena,” write Hanson and Kopstein.

Even if authoritarianism is averted, the damage that patrimonialism does to state capacity is severe. Governments’ best people leave or are driven out. Agencies’ missions are distorted and their practices corrupted. Procedures and norms are abandoned and forgotten. Civil servants, contractors, grantees, corporations, and the public are corrupted by the habit of currying favor.

To say, then, that Trump lacks the temperament or attention span to be a dictator offers little comfort. He is patrimonialism’s perfect organism. He recognizes no distinction between what is public and private, legal and illegal, formal and informal, national and personal. “He can’t tell the difference between his own personal interest and the national interest, if he even understands what the national interest is,” John Bolton, who served as national security adviser in Trump’s first term, told The Bulwark. As one prominent Republican politician recently told me, understanding Trump is simple: “If you’re his friend, he’s your friend. If you’re not his friend, he’s not your friend.” This official chose to be Trump’s friend. Otherwise, he said, his job would be nearly impossible for the next four years.

Patrimonialism explains what might otherwise be puzzling. Every policy the president cares about is his personal property. Trump dropped the federal prosecution of New York City Mayor Eric Adams because a pliant big-city mayor is a useful thing to have. He broke with 50 years of practice by treating the Justice Department as “his personal law firm.” He treats the enforcement of duly enacted statutes as optional—and, what’s more, claims the authority to indemnify lawbreakers. He halted proceedings against January 6 thugs and rioters because they are on his side. His agencies screen hires for loyalty to him rather than to the Constitution.

In Trump’s world, federal agencies are shut down on his say-so without so much as a nod to Congress. Henchmen with no statutory authority barge into agencies and take them over. A loyalist who had only ever managed two small nonprofits is chosen for the hardest management job in government. Conflicts of interest are tolerated if not outright blessed. Prosecutors and inspectors general are fired for doing their job. Thousands of civil servants are converted to employment at the president’s will. Former officials’ security protection is withdrawn because they are disloyal. The presidency itself is treated as a business opportunity.

Yet when Max Weber saw patrimonialism as obsolete in the era of the modern state, he was not daydreaming. As Hanson and Kopstein note, “Patrimonial regimes couldn’t compete militarily or economically with states led by expert bureaucracies.” They still can’t. Patrimonialism suffers from two inherent and in many cases fatal shortcomings.

The first is incompetence. “The arbitrary whims of the ruler and his personal coterie continually interfere with the regular functioning of state agencies,” write Hanson and Kopstein. Patrimonial regimes are “simply awful at managing any complex problem of modern governance,” they write. “At best they supply poorly functioning institutions, and at worst they actively prey on the economy.” Already, the administration seems bent on debilitating as much of the government as it can. Some examples of incompetence, such as the reported firing of staffers who safeguard nuclear weapons and prevent bird flu, would be laughable if they were not so alarming.

Eventually, incompetence makes itself evident to the voting public without needing too much help from the opposition. But helping the public understand patrimonialism’s other, even greater vulnerability—corruption—requires relentless messaging.

[Read: This is what happens when the DOGE guys take over]

Patrimonialism is corrupt by definition, because its reason for being is to exploit the state for gain—political, personal, and financial. At every turn, it is at war with the rules and institutions that impede rigging, robbing, and gutting the state. We know what to expect from Trump’s second term. As Larry Diamond of Stanford University’s Hoover Institution said in a recent podcast, “I think we are going to see an absolutely staggering orgy of corruption and crony capitalism in the next four years unlike anything we’ve seen since the late 19th century, the Gilded Age.” (Francis Fukuyama, also of Stanford, replied: “It’s going to be a lot worse than the Gilded Age.”)

Paolo Pellegrin / Magnum Photos

They weren’t wrong. “In the first three weeks of his administration,” reported the Associated Press, “President Donald Trump has moved with brazen haste to dismantle the federal government’s public integrity guardrails that he frequently tested during his first term but now seems intent on removing entirely.” The pace was eye-watering. Over the course of just a couple of days in February, for example, the Trump administration:

gutted enforcement of statutes against foreign influence, thus, according to the former White House counsel Bob Bauer, reducing “the legal risks faced by companies like the Trump Organization that interact with government officials to advance favorable conditions for business interests shared with foreign governments, and foreign-connected partners and counterparties”;

suspended enforcement of the Foreign Corrupt Practices Act, further reducing, wrote Bauer, “legal risks and issues posed for the Trump Organization’s engagements with government officials both at home and abroad”;

fired, without cause, the head of the government’s ethics office, a supposedly independent agency overseeing anti-corruption rules and financial disclosures for the executive branch;

fired, also without cause, the inspector general of USAID after the official reported that outlay freezes and staff cuts had left oversight “largely nonoperational.”

By that point, Trump had already eviscerated conflict-of-interest rules, creating, according to Bauer, “ample space for foreign governments, such as Saudi Arabia and the United Arab Emirates, to work directly with the Trump Organization or an affiliate within the framework of existing agreements in ways highly beneficial to its business interests.” He had fired inspectors general in 19 agencies, without cause and probably illegally. One could go on—and Trump will.

Corruption is patrimonialism’s Achilles’ heel because the public understands it and doesn’t like it. It is not an abstraction like “democracy” or “Constitution” or “rule of law.” It conveys that the government is being run for them, not for you. The most dire threat that Putin faced was Alexei Navalny’s “ceaseless crusade” against corruption, which might have brought down the regime had Putin not arranged for Navalny’s death in prison. In Poland, the liberal opposition booted the patrimonialist Law and Justice Party from power in 2023 with an anti-corruption narrative.

In the United States, anyone seeking evidence of the power of anti-corruption need look no further than Republicans’ attacks against Jim Wright and Hillary Clinton. In Clinton’s case, Republicans and Trump bootstrapped a minor procedural violation (the use of a private server for classified emails) into a world-class scandal. Trump and his allies continually lambasted her as the most corrupt candidate ever. Sheer repetition convinced many voters that where there was smoke, there must be fire.

Even more on point is Newt Gingrich’s successful campaign to bring down Democratic House Speaker Jim Wright—a campaign that ended Wright’s career, launched Gingrich’s, and paved the way for the Republicans’ takeover of the U.S. House of Representatives in 1994. In the late 1980s, Wright was a congressional titan and Gingrich an eccentric backbencher, but Gingrich had a plan. “I’ll just keep pounding and pounding on his [Wright’s] ethics,” he said in 1987. “There comes a point where it comes together and the media takes off on it, or it dies.” Gingrich used ethics complaints and relentless public messaging (not necessarily fact-based) to brand Wright and, by implication, the Democrats as corrupt. “In virtually every speech and every interview, he attacked Wright,” John M. Barry wrote in Politico. “He told his audiences to write letters to the editor of their local newspapers, to call in on talk shows, to demand answers from their local members of Congress in public meetings. In his travels, he also sought out local political and investigative reporters or editorial writers, and urged them to look into Wright. And Gingrich routinely repeated, ‘Jim Wright is the most corrupt speaker in the 20th century.’”

[Read: Why Meta is paying $25 million to settle a Trump lawsuit]

Today, Gingrich’s campaign offers the Democrats a playbook. If they want to undermine Trump’s support, this model suggests that they should pursue a relentless, strategic, and thematic campaign branding Trump as America’s most corrupt president. Almost every development could provide fodder for such attacks, which would connect corruption not with generalities like the rule of law but with kitchen-table issues. Higher prices? Crony capitalism! Cuts to popular programs? Payoffs for Trump’s fat-cat clients! Tax cuts? A greedy raid on Social Security!

The best objection to this approach (perhaps the only objection, at this point) is that the corruption charge won’t stick against Trump. After all, the public has been hearing about his corruption for years and has priced it in or just doesn’t care. Besides, the public believes that all politicians are corrupt anyway.

But driving a strategic, coordinated message against Trump’s corruption is exactly what the opposition has not done. Instead, it has reacted to whatever is in the day’s news. By responding to daily fire drills and running in circles, it has failed to drive any message at all.

Also, it is not quite true that the public already knows Trump is corrupt and doesn’t care. Rather, because he seems so unfiltered, he benefits from a perception that he is authentic in a way that other politicians are not, and because he infuriates elites, he enjoys a reputation for being on the side of the common person. Breaking those perceptions can determine whether his approval rating is above 50 percent or below 40 percent, and politically speaking, that is all the difference in the world.

Do the Democrats need a positive message of their own? Sure, they should do that work. But right now, when they are out of power and Trump is the capo di tutti capi, the history of patrimonial rule suggests that their most effective approach will be hammering home the message that he is corrupt. One thing is certain: He will give them plenty to work with.

In the isolation of a Washington, D.C., office building, with a small team of acolytes, Elon Musk is dismantling the civil service and fulfilling an old dream. Deep within the folds of the Western brain resides a yearning for a savior: a master engineer who imposes reason and efficiency on the messiness of modern life, who can deploy his acumen to usher in a golden age of abundance and harmony. This is a fantasy of submission, where the genius takes charge.

Given American conservatives’ recent rhetoric, their surrender to Musk’s vision of utopia is discordant, to say the least. Ever since the pandemic, the MAGA movement has decried the tyranny of a cabal of self-certain experts, who wield their technical knowledge unaccountably. But even as the right purports to loathe technocracy, it has empowered an engineer to radically remake the American state in the name of efficiency.

Trumpists might be surprised to know that they are fulfilling a dream first conceived by a 19th-century French crank, Henri de Saint-Simon. A utopian polymath who fought in the American Revolution and claimed to be a descendant of Charlemagne, he imagined a society in which engineers and industrial managers usurped the aristocracy at the top of the pecking order. The ruling cadre of engineers, he theorized, wouldn’t just solve social and economic problems, but serve as high priests, guiding society to efficiency, progress, and harmony. Technocracy and spirituality were intertwined in his doctrine, which he called the “New Christianity.”

[Read: Elon Musk is president]

In the last years of his life, Saint-Simon struggled to find a publisher for his books. His despair led him to shoot himself seven times in the head, a failed suicide attempt. Only after his death, in 1825, did he win cultlike devotion; his wider influence became unmistakable. Scholars dubbed him the “father of socialism,” and his veneration of the engineer ricocheted through the history of the left, especially in its faith in centralized planning. “Master technology,” Stalin famously implored his followers. “It is time that the Bolsheviks become experts.” (Eventually, Stalin murdered and imprisoned those who followed this command.)

The worship of the engineer is not confined to any single strain of ideology. It’s a modern impulse, and even ardent critics of the state have fallen victim to it. In Atlas Shrugged, every high-school libertarian’s favorite novel, Ayn Rand’s heroic protagonist, John Galt, is an engineer whose solitary capacity for invention and heterodox thinking make him a sort of über-mensch. And there are hints of this same heroic self-conception in the right-wing swatches of present-day Silicon Valley. Engineers are prophets of a new order because they promise inventions that will usher in the purest expressions of freedom: realms (cryptocurrency, space colonies) that are beyond the reach of the state.

One pivotal figure in American political history briefly embodied the noblest aspirations for technocracy—President Herbert Hoover, nicknamed the Great Engineer. After training at Stanford, he made a fortune in the mining business. Hoover believed ardently in scientific management: Any procedure could be simplified through studying the data. By monitoring workers, the engineer could cull waste from the productive process. Born a Quaker, Hoover delivered lyrical descriptions of his life’s work, which aren’t so far from Saint-Simon’s faith. Where other occupations were “parasitic,” in Hoover’s view, the engineer was the handmaiden of a humane social order because he “elevates the standards of living and adds to the comforts of life.”

[Tom Nichols: Trump and Musk are destroying the basics of a healthy] democracy

At his best, Hoover’s technocratic skills were something to behold. He was a genius at orchestrating responses to catastrophes; his coordination of food and supply shipments in Europe during World War I became the basis for his political mystique. Progressives were so enamored of his work that they desperately hoped he would run for president as a Democrat, so that they could preside over a new era of rational, well-organized government. Franklin D. Roosevelt, a fan before he became a foe, tried and failed to draft Hoover to run as his party’s standard-bearer in 1920.

Elected as a Republican in 1928, Hoover was in the White House when the nation’s economy collapsed. History regards him with disdain, less for his policies than for his distinct lack of warmth and his disregard for human suffering. He treated food distribution as an engineering problem, yet he never managed to describe victims with compassion. According to his biographer Joan Hoff Wilson, “They all became statistics—by the same impersonal scientific engineering approach and temperament that was to shock and dismay his fellow Americans during the Great Depression and erode his political credibility with them.”

The problem with applying scientific management to the government is its hollow heart, as the former auto executive Robert McNamara later showed to horrifying effect. As the secretary of defense, he presided over the escalation of the Vietnam War in the 1960s, deploying a data-driven approach that rendered casualties in the vernacular of statistics. (McNamara didn’t train as an engineer, but he self-consciously employed the mindset.) In his enthusiasm for optimization and efficiency, he paid no heed to the terrible human toll of his immaculate systems.

[Read: Trump advisers stopped Musk from hiring a noncitizen at DOGE]

In a far more benign way, Jimmy Carter, the only other engineer to become president, struggled to form human connections with the public. As the New York Times columnist Tom Wicker put it, he used an “engineer’s approach of devising ‘comprehensive’ programs on this subject or that, but repeatedly failed to mobilize public opinion in their support.” Carter’s brain was ill-equipped to process the irrationality of politics.

Despite this history of failure, Americans haven’t shaken the hope that some benevolent, hyperrational leader, immune to the temptations of political power, will step in to redesign the nation, to solve the problems that politicians can’t. That hope is unbreakable, because American culture invests engineers with the aura of wizardry. This is true for Elon Musk. For years, the media glorified him as a magician who harnessed the power of the sun, who revived the American space program, who rescued the electric car. Given that hagiographic press, some of it deserved, he could easily believe in his own ability to fix the American government—and think that a large chunk of the nation would believe that, too.

But in his short stay in Washington, Musk has already evinced the same moral shortcoming that afflicted Hoover and McNamara, the same inability to calculate the costs of cruelty. He has casually paused global aid programs that alleviate suffering; he has moved to destroy bureaucrats’ careers without concern for the rippling personal consequences. He has done this with an arrogance suffused with the spiritual self-certainity of Saint-Simon’s priestly caste of engineers. To a brain as rational as Musk’s, democracy is waste and inefficiency. The best system is the one bursting forth from his mind.

In an ember storm, every opening in a house is a portal to hell. A vent without a screen, a crack in the siding, a missing roof tile—each is an opportunity for a spark to smolder. A gutter full of dry leaves is a cradle for an inferno. Think of a rosebush against a bedroom window: fire food. The roses burn first, melting the vinyl seal around the window. The glass pane falls. A shoal of embers enter the house like a school of glowing fish. Then the house is lost.

As the Palisades Fire, just 8 percent contained this morning, and the Eaton Fire, still uncontained, devour Los Angeles neighborhoods, one thing is clear: Urban fire in the U.S. is coming back. For generations, American cities would burn in era-defining conflagrations: the Great Chicago Fire in 1871, the San Francisco fires of 1906. Then came fire-prevention building codes, which made large city burns a memory of a more naive time. Generations of western firefighters turned, instead, toward wildland burns, the big forest devastations. An urban conflagration was the worst-case scenario, the one they hoped they’d never see. And for a long time, they mostly didn’t.

Now more people live at the flammable edges of wildlands, making places that are primed to burn into de facto suburbs. That, combined with the water whiplash that climate change has visited on parts of California—extraordinarily wet years followed by extraordinarily dry ones—means the region is at risk for urban fire once again. And our ability to fight the most extreme fire conditions has reached its limit. The Palisades Fire alone has already destroyed more than 5,300 structures and the Eaton Fire more than 4,000, making both among the most destructive fires in California’s history. When the worst factors align, the fires are beyond what firefighting can meaningfully battle. With climate change, this type of fire will only grow more frequent.

The start of the Palisades and Eaton Fires was a case of terrible timing: A drought had turned abundant vegetation into crisp fire fuel, and the winter rains were absent. A strong bout of Santa Ana winds made what was already probable fire weather into all but a guarantee. Something—it remains to be seen what—ignited these blazes, and once they started, there was nothing anyone could do to stop them. The winds, speeding up to 100 miles an hour at times, sent showers of embers far across the landscape to ignite spot fires. The high winds meant that traditional firefighting was, at least in the beginning, all but impossible, David Acuna, a battalion chief for Cal Fire, told me: He saw videos of firefighters pointing their hoses toward flames, and the wind blowing the water in the other direction. And for a while, fire planes couldn’t fly. Even if they had, it wouldn’t have mattered, Acuna said. The fire retardant or water they would have dropped would have blown away, like the hose water. “It’s just physics,” he said.

California, and Southern California in particular, has some of the most well-equipped firefighting forces in the world, which have had to think more about fire than perhaps any other in the United States. On his YouTube livestream discussing the fires, the climate scientist Daniel Swain compared the combined fleet of vehicles, aircraft, and personnel to the army of a small nation. If these firefighters couldn’t quickly get this fire contained, likely no one could. This week’s series of fires is testing the upper limits of the profession’s capacity to fight wind-driven fires under dry conditions, Swain said, and rather than call these firefighters incompetent, it’s better to wonder how “all of this has unfolded despite that.”  

The reality is that in conditions like these, once a few houses caught fire in the Pacific Palisades, even the best firefighting could likely do little to keep the blaze from spreading, Michael Wara, a former member of California’s wildfire commission who now directs a climate-and-energy-policy program at Stanford, told me. “Firefighting is not going to be effective in the context we saw a few days ago,” when winds were highest, he said. “You could put a fire truck in every driveway and it would not matter.” He recounted that he was once offered a job at UCLA, but when the university took him to look at potential places to live in the Pacific Palisades, he immediately saw hazards. “It had terrible evacuation routes, but also the street layout was aligned with the Santa Ana winds so that the houses would burn down like dominoes,” he said. “The houses themselves were built very, very close together, so that the radiant heat from one house would ignite the house next door.”

In California, the shift toward ungovernable fires in populated places has been under way for several years. For the former Cal Fire chief deputy director Christopher Anthony, who retired in 2023, the turning point was 2017, when wildfires in populated places in Northern California’s wine country killed 44 people and burned nearly a quarter million acres. The firefighting profession, he told me, started to recognize then that fortifying communities before these more ferocious blazes start would be the only meaningful way to change their outcome. The Camp Fire, which decimated the town of Paradise in 2018, “was the moment that we realized that this wasn’t, you know, an anomaly,” he said. The new fire regime was here. This new kind of fire, once begun, would “very quickly overwhelm the operational capabilities of all of the fire agencies to be able to effectively respond,” he said.

As Wara put it, in fires like these, houses survive, or don’t, on their own. Sealed against ember incursion with screened vents, built using fire-resistant materials, separated from anything flammable—fencing, firewood, but especially vegetation—by at least five feet, a house has a chance. In 2020, California passed a law (yet to be enforced) requiring such borders around houses where fire hazard is highest. It’s a hard sell, having five feet of stone and concrete lining the perimeter of one’s house, instead of California’s many floral delights. Making that the norm would require a serious social shift. But it could meaningfully cut losses, Kate Dargan, a former California state fire marshal, told me.

Still, eliminating the risk of this type of wind-driven fire is now impossible. Dargan started out in wildland firefighting in the 1970s, but now she and other firefighters see the work they did, of putting out all possible blazes, as “somewhat misguided.” Fire is a natural and necessary part of California’s ecosystem, and suppressing it entirely only stokes bigger blazes later. She wants to see the state further embrace preventative fires, to restore it to its natural cycles. But the fires in Southern California this week are a different story, unlikely to have been prevented by prescribed burns alone. When the humidity drops low and the land is in the middle of a drought and the winds are blowing at 100 miles an hour, “we’re not going to prevent losses completely,” Dargan said. “And with climate change, those conditions are likely to occur more frequently.” Avoiding all loss would mean leaving L.A. altogether.

Rebuilding means choosing a different kind of future. Dargan hopes that the Pacific Palisades rebuilds with fire safety in mind; if it does, it will have a better chance of not going through this kind of experience again. Some may still want to grow a rosebush outside their window. After this is over, the bargaining with nature will begin. “Every community gets to pick how safe they want to be,” Dargan said.

The human cell is a miserable thing to study. Tens of trillions of them exist in the body, forming an enormous and intricate network that governs every disease and metabolic process. Each cell in that circuit is itself the product of an equally dense and complex interplay among genes, proteins, and other bits of profoundly small biological machinery.

Our understanding of this world is hazy and constantly in flux. As recently as a few years ago, scientists thought there were only a few hundred distinct cell types, but new technologies have revealed thousands (and that’s just the start). Experimenting in this microscopic realm can be a kind of guesswork; even success is frequently confounding. Ozempic-style drugs were thought to act on the gut, for example, but might turn out to be brain drugs, and Viagra was initially developed to treat cardiovascular disease.

Speeding up cellular research could yield tremendous things for humanity—new medicines and vaccines, cancer treatments, even just a deeper understanding of the elemental processes that shape our lives. And it’s beginning to happen. Scientists are now designing computer programs that may unlock the ability to simulate human cells, giving researchers the ability to predict the effect of a drug, mutation, virus, or any other change in the body, and in turn making physical experiments more targeted and likelier to succeed. Inspired by large language models such as ChatGPT, the hope is that generative AI can “decode the language of biology and then speak the language of biology,” Eric Xing, a computer scientist at Carnegie Mellon University and the president of Mohamed bin Zayed University of Artificial Intelligence, in the United Arab Emirates, told me.

Much as a chatbot can discern style and perhaps even meaning from huge volumes of written language, which it then uses to construct humanlike prose, AI could in theory be trained on huge quantities of biological data to extract key information about cells or even entire organisms. This would allow researchers to create virtual models of the many, many cells within the body—and act upon them. “It’s the holy grail of biology,” Emma Lundberg, a cell biologist at Stanford, told me. “People have been dreaming about it for years and years and years.”

These grandiose claims—about so ambiguous and controversial a technology as generative AI, no less—may sound awfully similar to self-serving prophesies from tech executives: OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis, and Anthropic’s Dario Amodei have all declared that their AI products will soon revolutionize medicine.

If generative AI does make good on such visions, however, the result may look something like the virtual cell that Xing, Lundberg, and others have been working toward. (Last month, they published a perspective in Cell on the subject. Xing has taken the idea a step further, co-authoring several papers about the possibility that such virtual cells could be combined into an “AI-driven digital organism”—a simulation of an entire being.) Even in these early days—scientists told me that this approach, if it proves workable, may take 10 or 100 years to fully realize—it’s a demonstration that the technology’s ultimate good may come not from chatbots, but from something much more ambitious.

Efforts to create a virtual cell did not begin with the arrival of large language models. The first modern attempts, back in the 1990s, involved writing equations and code to describe every molecule and interaction. This approach yielded some success, and the first whole-cell model, of a bacteria species, was eventually published in 2012. But it hasn’t worked for human cells, which are more complicated—scientists lack a deep enough understanding to imagine or write all of the necessary equations, Lundberg said.

The issue is not that there isn’t any relevant information. Over the past 20 years, new technologies have produced a trove of genetic-sequence and microscope data related to human cells. The problem is that the corpus is so large and complex that no human could possibly make total sense of it. But generative AI, which works by extracting patterns from huge amounts of data with minimal human instructions, just might. “We’re at this tipping point” for AI in biology, Eran Segal, a computational biologist at the Weizmann Institute of Science and a collaborator of Xing’s, told me. “All the stars aligned, and we have all the different components: the data, the compute, the modeling.”

Scientists have already begun using generative AI in a growing number of disciplines. For instance, by analyzing years of meteorological records or quantum-physics measurements, an AI model might reliably predict the approach of major storms or how subatomic particles behave, even if scientists can’t say why the predictions are accurate. The ability to explain is being replaced by the ability to predict, human discovery supplanted by algorithmic faith. This may seem counterintuitive (if scientists can’t explain something, do they really understand it?) and even terrifying (what if a black-box algorithm trusted to predict floods misses one?). But so far, the approach has yielded significant results.

[Read: Science is becoming less human]

“The big turning point in the space was six years ago,” Ziv Bar-Joseph, a computational biologist at Carnegie Mellon University and the head of research and development and computational sciences at Sanofi, told me. In 2018—before the generative-AI boom—Google DeepMind released AlphaFold, an AI algorithm that functionally “solved” a long-standing problem in molecular biology: how to discern the three-dimensional structure of a protein from the list of amino acids it is made of. Doing so for a single protein used to take a human years of experimenting, but in 2022, just four years after its initial release, AlphaFold predicted the structure of 200 million of them, nearly every protein known to science. The program is already advancing drug discovery and fundamental biological research, which won its creators a Nobel Prize this past fall.

The program’s success inspired researchers to design so-called foundation models for other building blocks of biology, such as DNA and RNA. Inspired by how chatbots predict the next word in a sentence, many of these foundation models are trained to predict what comes next in a biological sequence, such as the next set of As, Ts, Gs, and Cs that make up a strand of DNA, or the next amino acid in a protein. Generative AI’s value extends beyond straightforward prediction, however. As they analyze text, chatbots develop abstract mathematical maps of language based on the relationships between words. They assign words and sentences coordinates on those maps, known as “embeddings”: In one famous example, the distance between the embeddings of queen and king is the same as that between woman and man, suggesting that the program developed some internal notion of gender roles and royalty. Basic, if flawed, capacities for mathematics, logical reasoning, and persuasion seem to emerge from this word prediction.

Many AI researchers believe that the basic understanding reflected in these embeddings is what allows chatbots to effectively predict words in a sentence. This same idea could be of use in biological foundation models as well. For instance, to accurately predict a sequence of nucleotides or amino acids, an algorithm might need to develop internal, statistical approximations of how those nucleotides or amino acids interact with one another, and even how they function in a cell or an organism.

Although these biological embeddings—essentially a long list of numbers—are on their own meaningless to people, the numbers can be fed into other, simpler algorithms that extract latent “meaning” from them. The embeddings from a model designed to understand the structure of DNA, for instance, could be fed into another program that predicts DNA function, cell type, or the effect of genetic mutations. Instead of having a separate program for every DNA- or protein-related task, a foundation model can address many at once, and several such programs have been published over the past two years.

Take scGPT, for example. This program was designed to predict bits of RNA in a cell, but it has succeeded in predicting cell type, the effects of genetic alterations, and more. “It turns out by just predicting next gene tokens, scGPT is able to really understand the basic concept of what is a cell,” Bo Wang, one of the programs’ creators and a biologist at the University of Toronto, told me. The latest version of AlphaFold, published last year, has exhibited far more general capabilities—it can predict the structure of biological molecules other than proteins as well as how they interact. Ideally, the technology will make experiments more efficient and targeted by systematically exploring hypotheses, allowing scientists to physically test only the most promising or curiosity-inducing. Wang, a co-author on the Cell perspective, hopes to build even more general foundation models for cellular biology.

The language of biology, if such a thing exists, is far more complicated than any human tongue. All the components and layers of a cell affect one another, and scientists hope that composing various foundation models creates something greater than the sum of their parts—like combining an engine, a hull, landing gear, and other parts into an airplane. “Eventually it’s going to all come together into one big model,” Stephen Quake, the head of science at the Chan Zuckerberg Initiative (CZI) and a lead author of the virtual-cell perspective, told me. (CZI—a philanthropic organization focused on scientific advancement that was co-founded by Priscilla Chan and her husband, Mark Zuckerberg—has been central in many of these recent efforts; in March, it held a workshop focused on AI in cellular biology that led to the publication of the perspective in Cell, and last month, the group announced a new set of resources dedicated to virtual-cell research, which includes several AI models focused on cell biology.)

In other words, the idea is that algorithms designed for DNA, RNA, gene expression, protein interactions, cellular organization, and so on might constitute a virtual cell if put together in the right way. “How we get there is a little unclear right now, but I’m confident it will,” Quake said. But not everyone shares his enthusiasm.

Across contexts, generative AI has a persistent problem: Researchers and enthusiasts see a lot of potential that may not always work out in practice. The LLM-inspired approach of predicting genes, amino acids, or other such biological elements in a sequence, as if human cells and bodies were sentences and libraries, is in its “very early days,” Quake said. Xing likened his and similar virtual-cell research to having a “GPT-1” moment, referencing an early proof-of-concept program that eventually led to ChatGPT.

Although using deep-learning algorithms to analyze huge amounts of data is promising, the quest for more and more universal solutions struck some researchers I spoke with as well-intentioned but unrealistic. The foundation-model approach in Xing’s AI-driven digital organisms, for instance, suggests “a little too much faith in the AI methods,” Steven Salzberg, a biomedical engineer at Johns Hopkins University, told me. He’s skeptical that such generalist programs will be more useful than bespoke AI models such as AlphaFold, which are tailored to concrete, well-defined biological problems such as protein folding. Predicting genes in a sequence didn’t strike Salzberg as an obviously useful biological goal. In other words, perhaps there is no unifying language of biology—in which case no embedding can capture every relevant bit of biological information.

[Read: We’re entering uncharted territory for math]

More important than AlphaFold’s approach, perhaps, was that it reliably and resoundingly beat other, state-of-the-art protein-folding algorithms. But for now, “the jury is still out on these cell-based models,” Bar-Joseph, the CMU biologist, said. Researchers have to prove how well their simulations work. “Experiment is the ultimate arbiter of truth,” Quake told me—if a foundation model predicts the shape of a protein, the degree of a gene’s expression, or the effects of a mutation, but actual experiments produce confounding results, the model needs reworking.

Even with working foundation models, the jump from individual programs to combining them into full-fledged cells is a big one. Scientists haven’t figured out all of the necessary models, let alone how to assemble them. “I haven’t seen a good application where all these different models come together,” Bar-Joseph said, though he is optimistic. And although there are a lot of data for researchers to begin with, they will need to collect far more moving forward. “The key challenge is still data,” Wang said. For example, many of today’s premier cellular data sets don’t capture change over time, which is a part of every biological process, and might not be applicable to specific scientific problems, such as predicting the effects of a new drug on a rare disease. Right now, the field isn’t entirely sure which data to collect next. “We have sequence data; we have image data,” Lundberg said. “But do we really know which data to generate to reach the virtual cell? I don’t really think we do.”

In the near term, the way forward might not be foundation models that “understand” DNA or cells in the abstract, but instead programs tailored to specific queries. Just as there isn’t one human language, there may not be a unified language of biology, either. “More than a universal system, the first step will be in developing a large number of AI systems that solve specific problems,” Andrea Califano, a computational biologist at Columbia and the president of the Chan Zuckerberg Biohub New York, and another co-author of the Cell perspective, told me. Even if such a language of biology exists, aiming for something so universal could also be so difficult as to waste resources when simpler, targeted programs would more immediately advance research and improve patients’ lives.

Scientists are trying anyway. Every level of ambition in the quest to bring the AI revolution to cell biology—whether modeling of entire organisms, single cells, or single processes within a cell—emerges from the same hope: to let virtual simulations, rather than physical experiments, lead the way. Experiments may always be the arbiters of truth, but computer programs will determine which experiments to carry out, and inform how to set them up. At some point, humans may no longer be making discoveries so much as verifying the work of algorithms—constructing biological laboratories to confirm the prophecies of silicon.