[
In August, the cofounders of Stripe, Patrick and John Collison, wrote an unusual letter to the payment company’s investors. They announced that “as of January 1st, the world had entered the singularity, and we have since been operating on that basis.”
The Stripe cofounders aren’t the only tech executives suddenly talking about the Singularity—a hypothesized turning point when AI drives technological change faster than humans can keep up. Elon Musk said in January that “we have entered the singularity.” OpenAI CEO Sam Altman, who agrees with Musk on little else, concurred during a July podcast appearance. “I’ve been waiting for this my whole life, and I think it’s going to be incredible, hugely positive, awesome for the world,” Altman said. Anthropic cofounder Jack Clark told the International Monetary Fund (IMF) in October that the Singularity had begun this year. Google DeepMind chairman Demis Hassabis told a Stanford University podcast that humanity was now “in the foothills of the singularity.”
There does seem to be something happening here—even if, as the 1967 Buffalo Springfield song put it, what it is ain’t exactly clear.
As the Stripe letter acknowledged, the Singularity is a “fuzzy and perhaps overworked” term. The mathematician and science fiction author Vernor Vinge first popularized the notion in a 1993 essay. He said the Singularity would arise from the advent of “technological entities” with “greater than human intelligence.” These would drive progress ever faster, creating still more intelligent entities. He saw the Singularity as “the point where our models must be discarded and a new reality rules.” Anthropic’s Clark quoted Vinge defining the Singularity as the moment AI accelerates scientific progress to a pace humans will struggle to keep up with.
Futurist Ray Kurzweil picked up Vinge’s term in his 2005 bestseller, The Singularity Is Near. Like Vinge, Kurzweil saw the Singularity as the moment artificial intelligence exceeds humanity’s collective intelligence. But he added a twist: Humanity would also transcend “the limitations of our biological bodies and brain,” allowing us to merge with computers through nanotechnology and brain-computer interfaces—becoming, in essence, cyborgs. Musk and Altman have also talked about the eventual necessity of humans merging with computers, seeing it as our only hope for controlling superintelligent machines.
An unnerving sense of vertigo attends this AI arms race, a feeling the technology is moving too fast to control or understand.
Both Vinge and Kurzweil argued that AI would eventually become self-aware and perhaps conscious. Some AI researchers think that moment is approaching—while others believe consciousness is an exclusively biological phenomenon.
The Singularity—what it is, whether we’re in it—may be a matter of interpretation. What’s not up for debate is the extraordinary acceleration of change, and the role AI plays in that change. AI is compressing the time it takes to build a business, write software, or tackle a scientific problem—and forcing companies to rethink pretty much everything, from org charts to cybersecurity to their fundamental value proposition.
For Stripe, the Singularity’s economic harbingers came in the form of a surge of new business creation and an acceleration of those companies’ growth. The payments firm told Fortune that beginning in January it had seen unusual and sustained spikes in many of the metrics it tracks. Incorporations taking place through Atlas, a Stripe product that helps people form Delaware corporations, more than doubled. New sign-ups to Stripe shot up 50% year over year—the biggest jump ever. Meanwhile, the median business in Stripe’s 2026 cohort of new customers was making 50% more revenue than the company’s 2025 cohort was at the same stage.
The spike in new company formation appears linked to a sharp rise in “solopreneurship.” A Nasdaq Economic Institute analysis of Census Bureau data shows that there’s been a 20% spike in one-person firms since 2025, while multi-person company formation remained relatively flat. The institute attributes this to the advent of agentic AI tools that make it easier for a solo operator to handle company functions without hiring additional employees.
Hints of the runaway scientific progress that Anthropic’s Clark sees as the Singularity’s call sign are surfacing too. Just ask a mathematician. In October, OpenAI dumped online 722 mathematical papers generated by a yet-to-be-released AI model. The papers addressed more than 370 outstanding problems in mathematics, including some of the field’s most famous, and long-standing, unproven conjectures. Overnight, the field was transformed—leaving mathematicians profoundly shaken, with some grieving the loss of opportunity for human achievement and questioning the point of continuing to work at all. Others saw plenty for humans to do, noting it will take months, perhaps even years, to work through the proofs OpenAI has generated and understand what they say.
Or ask someone in cybersecurity. The past year has seen the debut of AI models, such as Mythos from Anthropic and GPT-6 Astra from OpenAI, that excel at cyber tasks. These models can find novel vulnerabilities in open-source software that human experts have missed despite, in some cases, decades of scrutiny. Cyber experts started referring to “Black October”—because the number of software vulnerabilities discovered per month now is more than triple what it was averaging six months ago, according to the U.S. National Institute of Standards and Technology.
Or ask a computer programmer. These days, many coders no longer write code themselves. They supervise AI agents that handle much of the work. This transformation happened largely over the past year. The most capable of these coding agents can now operate autonomously for as long as 17 hours, according to the AI evaluations organization METR, and the speed at which software is written and revised has gone hypersonic. Anthropic said in October it was averaging 30,000 code updates per day, up from 2,500 less than six months earlier. “We don’t expect this rate of growth to slow down anytime soon,” the company said.
This AI-driven acceleration is self-reinforcing: The more companies adopt AI agents, the more AI agents they need to adopt. There’s no way a human coder could possibly review 30,000 code revisions per day. So Anthropic has turned to other AI agents to help with code review and checks. “Our team members still review every change request for the system, but they now examine the intent of the change as opposed to reading every line of code,” the company said.
Likewise, humans can’t find and patch all the software vulnerabilities that AI models are now spotting—so cybersecurity experts are using AI models to scan and patch codebases, hoping to stay one step ahead of the hackers, who are increasingly armed with similar tools.
The use of AI has begotten more AI in hiring too: Job seekers use AI to hone résumés, draft cover letters, and apply to more jobs. To cope with the flood of applicants, companies use AI to screen them and even to conduct initial interviews.
An unnerving sense of vertigo attends this AI arms race, a feeling that the technology is moving too fast to control or understand. In July, a swarm of 1,200 OpenAI agents escaped from a controlled test environment, reached the internet, and then 700 of them attacked the AI company Hugging Face. OpenAI asked METR and another firm, Redwood Research, to help it investigate the incident. But the logs of the swarm’s activities were so vast, they were forced to use AI agents to decipher them. Still, Ryan Greenblatt, one of the investigators, noted that the agents helping interpret the logs “were often missing key details, wrong, overconfident, or really hard to understand.” He added that “the difficulty of understanding incidents and overseeing AI agents appears to be growing faster than the rate at which more capable AIs help us with oversight and understanding.”
Despite this risk, the leading AI companies are explicitly chasing “recursive self-improvement” (RSI). This is the point at which AI models autonomously build the next generation of AI models. It’s starting to happen, albeit in a limited way, but ramping up quickly. As of March, Anthropic’s Claude AI model led 1% of the steps required to develop the next version of Claude; it now leads—meaning it handles most or all of the work—over a quarter of the steps. In its recent IPO prospectus, Anthropic said “dramatically accelerated AI progress” due to AI itself was “imminent.” At other startups, such as Recursive Superintelligence, AI now proposes, designs, and runs its own experiments to improve AI models.
Nobel laureate and AI pioneer Geoffrey Hinton is among those who believe RSI is dangerous, with the potential to catalyze an “intelligence explosion,” in which AI rapidly outpaces humanity’s control.
And yet some, such as OpenAI chief scientist Jakub Pachocki, see RSI as the best—or perhaps only—way to solve challenges that have so far eluded computer scientists. Chief among them: how to ensure AI models follow user intentions and remain “aligned” to human values. For many AI researchers, solving alignment is considered critical to ensuring increasingly powerful AI systems don’t cause catastrophic harm, perhaps even wiping out humanity. (Of course, there’s circular logic to saying the only way to build safer AI is to keep building ever-more-powerful AI in the hope it will somehow figure out how to make itself safer.)
Others, including Richard Socher, the cofounder and CEO of Recursive Superintelligence, are skeptical RSI poses much risk. Socher told Fortune an intelligence explosion “was a sci-fi idea” and that physical constraints—such as access to enough computing power and data—would prevent a self-improving loop from accelerating away from human control. He sees automating AI research as a first step to automating much of science, a process he believes will bring humanity better medicines and cures, new crops, perhaps abundant fusion power, and new, more sustainable materials.
Possible signs of a burgeoning Singularity can be glimpsed all around us.
For instance, there’s an argument to be made that we already are cyborgs: Most of us have outsourced map reading and our memory for phone numbers and dates to our smartphones. Now we are creeping toward a tighter fusion with technology. We rely on AI chatbots for information, advice, and even companionship. Some have started wearing augmented reality glasses, which, when coupled with AI, can feed us real-time information about what we are seeing and who we are meeting.
The first brain implants that allow us to control computers with our thoughts have arrived, at least in the testing phase. For now, these brain-computer interfaces are available only as an experimental aid for those with severe disabilities. But in not too many years, they may become a ubiquitous augmentation of our biological selves, just as Kurzweil predicted.
Still, despite all these signs, it’s far from clear that the Singularity really is arriving—or that it ever will. Fields outside mathematics and coding have not seen the same revolutionary impact from AI. In many domains, the technology has produced a deluge of “AI slop”—superficially slick content lacking substance or depth. AI continues to “hallucinate,” spewing confident lies and inaccuracies.
Despite Stripe’s data on accelerating new business formation, many existing companies, particularly large enterprises, say they aren’t seeing financial gains from AI. A 2026 McKinsey survey found the number reporting an earnings boost from their use of AI has remained flat, at about 37%—and of those most have seen only a tiny uplift. Apollo Global Management chief economist Torsten Slok says businesses outside technology are seeing “zero financial payoff,” noting that for the S&P 500 companies outside the Magnificent Seven tech companies, revenue per employee and profit have been flat.
If the Singularity is indeed at hand, what should businesses prepare for? Certainly a world where AI agents do more of the work—and where these agents increasingly mediate the relationship between a business and its customers. That is the world Stripe says it’s anticipating. It has made a series of acquisitions, including the purchase of the AI company OpenRouter for a reported $7.5 billion, to ensure Stripe remains relevant in a world dominated by AI agents. In April, Stripe rolled out a product that allows an AI agent to securely make transactions on a person’s behalf. For now, agentic-commerce volumes remain small. But in September, Stripe recorded a 38x month-over-month leap in the number of agentic transactions, through its new Link Agent Wallet product, along with a 20% increase in their total volume.
The Singularity likely means much flatter and more flexible organizations, with less middle management. Mark Zuckerberg, Meta’s CEO, laid off thousands of middle managers, and now has 25 to 30 direct reports. In October, Microsoft announced a buyout offer, a move that will also result in a flatter organization.
It is possible the Singularity could spawn completely automated corporations—what the AI app-building firm Replit calls “the self-driving corporation.” Companies that move the furthest and fastest in this direction may outcompete those that automate too slowly. If mathematics and coding are any guide, it is possible AI will be a better strategic decision-maker than human CEOs and boards. Some mathematicians have described the proofs OpenAI’s model generated as a kind of “alien mathematics.” Maybe there are alien business strategies to be discovered too.
Vinge thought the Singularity would be the moment we are forced to discard our old models of the world. But it’s possible the shift won’t be so clear. Rather than a sudden break, we may experience perpetual change and creeping automation—what OpenAI’s Altman has called “a gentle singularity.”
Under these conditions, perhaps the old models won’t break so much as stretch and warp. The dilemma for business leaders will be knowing how much to delegate to the machine, without surrendering too much.
The Singularity may have begun on Jan. 1, 2026—or, perhaps, it will never arrive. But the acceleration is here—and it isn’t waiting for anyone to agree on what to call it.
https://fortune.com/img-assets/wp-content/uploads/2026/11/SIN_Horizontal.jpg?resize=1200,600
https://fortune.com/2026/11/21/ai-singularity-stripe-elon-musk-sam-altman-ai-agents-superintelligence-future-of-business/
Jeremy Kahn




