• OpenAI CEO Sam Altman and mathematicians have a complicated relationship.
• OpenAI has taken credit for solving hundreds of math problems, irking many mathematicians.
• At a recent talk at Stanford, Altman said mathematicians have already gone from anger to acceptance.

Like someone in mourning, Sam Altman said mathematicians went through phases after OpenAI began its blitz into the field.

First, there was disbelief. Then, there was pain.

"The day one reaction was like, 'We're cooked, and our field is over,'" OpenAI's CEO said. "'We're telling our grad students to switch majors.'"

After that, there was anger.

"The day three reaction was maybe like: 'It's horrible that OpenAI did this,'" Altman said at a recent talk at Stanford, before hundreds of students jittery about the viability of their future careers.

Come day five, though, in Altman's telling, mathematicians began to come around to what he described as AI's now-inalienable role in the study of numbers and proofs.

"We've now seen a lot of people say, actually, I'm going to have a different but hopefully better kind of career as a mathematician," Altman said.

Altman's buoyant assessment contrasts sharply with what some in the math community have described as an existential transformation.

After OpenAI abruptly broke into the field's prerogatives in September, announcing its models had solved the Navier-Stokes equations — a famously challenging math puzzle that stumped the industry's brightest since the 19th Century — preeminent mathematicians erupted in fury.

A pair of academics, one at New York University and another at Anthropic, accused OpenAI of using data from their chat logs to help crack the problem. Mathematicians also complained that while an AI model might have solved the problem, it didn't show its work, leaving them with the less celebratory task of verifying it.

And in the wake of all that controversy, there was more controversy.

A group of undergrads at Caltech, wary of how rapidly AI was disrupting their field and wanting to stay relevant, advertised a "mathathon" where participants were encouraged to use AI to solve long-standing problems.

OpenAI dropped out of sponsoring the event after more than 700 academics signed an open letter denouncing the competition for promoting AI "slop." (Organizers have since made adjustments to the rules, and the competition is set for November.)

More recently, Terence Tao, a math prodigy at UCLA who earned his doctorate at 20, said math's "romantic era" is over, as top academics like him adjust to a workflow that involves verifying AI's work rather than solving problems individually.

"Some things are being lost," Tao said after OpenAI this month said it solved over 700 more problems.

Meanwhile, Altman, sitting on the opposite side of the debate, has adopted the optimism of the disrupter.

"I don't think there's going to be like a job apocalypse," Altman said of AI causing companies to restructure their workforces. In the field of math, however, Altman said the disruptions came more quickly than in other disciplines and industries.

"That was the fastest that I think we've all gotten to watch a field go through this cycle," he said.

• A version of this post appears in CMO Insider.
• Sign up for Business Insider's marketing newsletter.

When Ameresco CMO Leila Dillon asked ChatGPT and other AI answer engines to describe what her company did, she was shocked by the results.

The publicly traded company is an energy infrastructure provider, helping power the rapidly growing AI buildout, as well as military bases and municipal facilities. But the AI chatbots were outdated in their responses, frequently referring to it simply as a cleantech or renewable energy business — areas it still operates in, but are less of a focus — or citing smaller projects it worked on years ago.

"It was quite horrifying," Dillon told CMO Insider. "We were like, 'That's not at all who we are.'"

Dillon set about overhauling the company's messaging and how it structured its marketing content to better reflect its business.

The work has begun to take effect. Ameresco tracked how often AI chatbots mentioned the company in response to 50 questions that potential customers and investors might ask. It appeared in 61% of those answers in 2026, up from 43% in 2025, according to the company.

As more consumers turn to platforms like ChatGPT, Google's Gemini, Anthropic's Claude, and Microsoft's Copilot to research companies and products, marketers are racing to hone their generative engine optimization (GEO) strategies to ensure their brands appear prominently and accurately in AI-generated answers.

"I used to Google everything that I wanted to know about," Dillon said. "I barely use Google anymore. I use Copilot, and I get better results."

In a January survey from the Interactive Advertising Bureau, 73% of US brand and agency buyers said they planned to devote more time or resources to optimizing content for AI-generated answers in 2026, compared with last year. In a separate survey of US marketing professionals conducted by Semrush, an AI and search visibility platform, in April, 30% of marketers said AI search tools had described their brands inaccurately.

The rush by marketers to get up to speed with GEO has spawned a cottage industry of AI search experts and tools promising quick fixes. But with AI companies constantly updating their models, marketers are in a phase of constant experimentation as they scramble to figure out what actually influences the chatbots.

"There's no one shot that fits all to solve that for you," Dillon said.

An AI audit identified 47 areas for improvement

In January 2025, Dillon's team and its PR agency, Greenough, began auditing how often Ameresco appeared in AI-generated answers to 50 prompts — such as, "What are the top energy infrastructure providers in the US?" — on platforms including ChatGPT and Claude. The prompts reflected questions that investors and prospective customers, including people from universities and utilities, might ask AI chatbots.

The audit identified 47 areas for improvement, including how Ameresco described its business across its website, social media, press releases, and executive communications.

"You name it, we were doing it wrong," Dillon said.

Ameresco began rolling out a new corporate brand identity in July this year, reflecting its broader reorganization around two business pillars: power infrastructure and buildings and public infrastructure.

It had laid the groundwork for the new corporate brand reveal by making several changes to its content strategy.

Those included:

• Updating its corporate boilerplate from "a leading cleantech integrator specializing in energy efficiency and renewable energy" to "a leading energy infrastructure company," reflecting the new positioning across its marketing and communications.
• Refreshing its awards submission strategy to lead with its power infrastructure projects, and announcing each win with a press release that appears on wire services and industry trade publications.
• Introducing Q&A-style content formats, designed to make the information easier for AI chatbots to interpret.
• Creating more content designed for different customer groups, such as data center operators or school facilities managers, rather than relying on generic marketing materials.
• Launching a video content series called "Energy Benefits Unlocked," highlighting the tangible results Ameresco's projects accomplished for its customers.
• Introducing quarterly AI visibility tracking audits across the 50 different prompts, reflecting its different audiences, and adding competitor comparisons.

A constantly moving target

The strategy appears to be paying off in some ways. Ameresco said organic traffic to its website rose 50% in the 12 months to June 2026. Dillon said it's still too early to assess precisely how those gains translate into business results, though she believes the GEO work is helping the marketing team hit its targets more quickly. The company is scheduled to report its third-quarter earnings on November 9.

Dillon said the rapid changes in AI mean the work is never really finished. Ameresco regularly revisits its AI visibility scores and has invested in training the marketing team, which shares tips, tricks, and their "epic AI fails" in its weekly meetings.

"One of the main themes is that what happened last week is not what you're going to get this week; it's changing that quickly," Dillon said.

"You don't just update your website and say: Done, woo, all set," she added.

• Terence Tao, now a UCLA professor, earned his doctorate in mathematics at 20.
• The prodigy said his work is changing in the age of AI.
• With AI tools doing the legwork, math is no longer "heroic," Tao said.

AI is upending mathematics. For one top professor, the issue is just as spiritual as it is practical.

Terence Tao, a renowned professor at UCLA, said in a recent podcast appearance that the "romantic era" of math is over, thanks to AI tools capable of solving hundreds of problems with some simple prompting.

Now, he said, even leading scholars like him are reduced to sorting through reams of AI-generated proofs, instead of doing the "heroic" solving themselves.

"We are moving to an industrial era where math will become much more large-scale," Tao said in an interview this week on Numberphile, a YouTube channel dedicated to math.

"We are entering an era where exciting new things are possible, but some things are being lost," he added.

OpenAI began its brusque push into the field by solving the Navier-Stokes equation in early September, one of math's millennium problems.

The revelation that AI had cracked it caused consternation from many a math geek. Some mathematicians also expressed frustration that while the AI model might have solved the problem, it didn't show its work, leaving them with the less celebratory task of verifying it.

Then, earlier this month, OpenAI said it cracked another 700 outstanding problems. A math professor at New York University told CNBC that the company's release had derailed work by early-career mathematicians.

"We've had whole research programs wiped out on Tuesday by their dump," the professor, Tristan Buckmaster, said at the time.

In a separate brouhaha in September, OpenAI backed out of a sponsorship for a so-called AI "mathathon" — an event organized by undergraduates at Caltech, where participants were encouraged to use AI to solve equations — after complaints from a group of more senior academics.

Event organizers told Business Insider at the time that their professors were sorely out of touch with the field's trajectory.

"We look to the future, and we don't even know if there's going to be grad school open positions by the time we get to Ph.D," one organizer, Caiman Moreno-Earle, said.

Though OpenAI has said it is taking steps to ensure human academics retain their role in this new era, some researchers and professors are already forecasting their possible obsolescence.

Tao isn't as much of a doomer as some of his colleagues in the field. However, he did say in the interview that the stories math nerds tell themselves may not mean so much in the future.

"We have these sort of mythological stories," he said in reference to the tale of Sir Andrew Wiles, who famously spent seven years isolated in his attic solving a single difficult equation.

"There was something romantic about sort of the heroic age," Tao said.

• There was only one thing on the minds of the founders at this reality-show-style startup incubator.
• It wasn't sex.
• Founder teams had two weeks to become profitable, and potent social media was paramount.

At 9 p.m. on a recent Friday night, Max Harding had just returned from a barefoot run in the woods.

The eccentric 20-year-old Californian was back in a large barn in rural Pennsylvania — still barefoot — alternatively working on his startup and asking me about the point of life.

Though not an especially adept programmer, Harding possesses a more adroit, perhaps unteachable skill that has never been more valuable in the current landscape of founder culture.

"Max captures the attention of the camera, keeps us center stage," said his cofounder, Skyler Ji, who added that Harding's role includes "spreading our philosophy" while he runs the operation behind the scenes.

"Max brings some of the grander things," Ji told me. The pair's startup, Hivemind Research, aims to build "digital brains" in the form of social media.

Harding and Ji were among 66 mostly young, mostly imaginative startup founders taking part in a two-week "incubator" workshop called Launchpad.

The novel event was hosted by Stan, the company behind a moderately successful platform for independent creators to advertise and sell their services. At the last meetup Stan organized, startup founders jet-skied the East River at golden hour.

At the incubator event, the entirety of which was filmed for use in a reality show that Stan plans to release sometime soon (though it is uploading Instagram clips daily), magnetic personalities were among the most valuable commodities.

Stan cofounder Vitalii Dodonov told me that the 30 participating teams were chosen from a pool of 700 applicants, primarily for their existing social media followings and their ability to grow them.

"It's not the building that gets you to revenue, it's the distribution," he said of today's startup culture.

Of the many metrics competing teams were required to meet daily (10 sales calls, for example), social media videos were among the most heavily weighted.

Everywhere I looked, people were talking about their companies into phones attached to tripods. Their products ran the gamut of usefulness — ranging from an AI facial scanner that tracks acne to an AI app that connects parents with babysitters.

Dodonov and John Hu, Stan's other cofounder, were trailed at all times by multi-person film crews, who glared should you emit an ill-timed sound while the cameras rolled. (The Velcro rip of my messenger bag earned me some leers.)

Even outfits were choreographed. On the advice of the camera crews, Hu wore a "cream" colored getup while Dodonov donned a "charcoal" number to emphasize what was described to me as the "yin and yang" paradigm between the two.

I spent about 16 cumulative hours at the event over two days earlier this month, at Stan's invitation. I ate meals with the founders, did a 7:30 a.m. cold plunge in the on-site lake alongside a Danish guy selling an AI to-do list, and commiserated with the startup founders who routinely pulled all-nighters to hit their metrics.

Here's what I learned from the experience.

Reality TV meets hackathon

The event was billed as a sort of reality-TV-style game show-slash-incubator in the vein of hits like "Love Island" and "Squid Game."

As in those shows, founder teams at Launchpad were subject to firepit eliminations, discreet alliances, and competitive drama under the ever-watchful lenses of dozens of expensive cameras.

Except instead of a bunch of promiscuous co-eds doing risque challenges on a tropical island somewhere, contestants at Launchpad were mostly tech geeks with an unhealthy number of hours notched building software on their MacBooks.

"There's no sex," laughed Jacob Snarr, a 24-year-old from Philadelphia. "Nobody here is thinking about that."

Snarr and his cofounder, RJ McLaughlin, built a web browser game that lets friends debate their tastes and preferences, from their favorite foods to other everyday questions. The pair enlisted me as a product tester one morning, and we debated the merits of empanadas in the chilly barn.

McLaughlin and Snarr were in second place on the leaderboard when I talked to them on Saturday morning — saving them from elimination.

Rey Mendoza, who at 41 was one of the oldest founders in the competition, had already been eliminated. When I spoke with him, he was fighting for a chance at redemption. He showed me his sleep tracker for the six days. He hadn't slept more than six hours a night.

"Here, you're just fucking building, like 24/7," Mendoza, fatigued, said reverently.

Event organizers added a number of less-intense, more fun-oriented events to the schedule, which were largely ignored by teams scrambling to upload more videos to social media.

A "Mario Kart" meetup in a room adjoining the cafeteria was attended by precisely nobody. Two TVs showed the game's start screen, while maybe a dozen people sat at tables squinting at laptops.

It's hard to imagine binge-watching some 60 sweatshirt-clad nerds drearily prompting chatbots to refine source code for yet another quirky AI startup. After all, everyone and their cousin is vibe coding their way into progressively nicher business models these days.

In 2025, a record number of business applications — about 5.7 million — were submitted by Americans, according to census data. However, the number of businesses the government estimates will be successful enough to require employees actually declined.

Those kinds of numbers mean it takes more to stand out. In our interview, Hu pointed at my phone and said I could use AI tools to start three or four businesses today — but, of course, they likely wouldn't be profitable.

That, Hu said, is where the founders' personalities are supposed to come in, together with a bit of manufactured drama and endless streams of social media content.

In the first challenge of the event, held on the first day, for example, teams were required to submit one social media video for public review around the firepit. Then, in front of everyone, Hu tore into them.

"I was very honest," he said of one particular "roast" that produced tears from the submitters. "I was like, 'Guys, would anyone ever watch this? No. The answer is no, because of the quality — you never have a shot."

The moment would certainly make for good television. It also, Hu told me, mimicked the real-life struggle inherent in starting a company in a saturated market.

"They're going through the true ups and downs of a true cofounder relationship and a startup experience," he said of the team that was subjected to his roast, who were initially so dejected they discussed "breaking up."

Overall, most of the teams I talked to said they enjoyed the ramshackle experience of trying to make a company profitable in two weeks. At the very least, it was all happening in a picturesque setting.

A $3 million investment

Stan spared no expense to pull off the event.

It was held at a luxury, 400-acre villa two hours north of New York City that has in the past hosted retreats for high-flying companies like J.P. Morgan, whose executives, according to one staff member at the retreat at least, are keen for employees to embrace buzzwords like "team-building" and "company culture" among the illusion of rustic calm far from Manhattan.

We were in the sticks of eastern Pennsylvania — so deep in the woods, in fact, that cell service was nonexistent. There was fiber-optic internet, however, and the setting, though far-flung, could hardly be called rustic.

A squadron of golf carts staffed by the venue's seasonal workforce, many of whom were from Ireland or Australia, patrolled the grounds.

I was personally given the cellphone number of one such seasonal worker and encouraged to notify her anytime I wanted to go anywhere, at which point she would materialize with the golf cart.

Dodonov told me that Stan spent a combined $3 million renting the place for two weeks, booking camera crews, and making other necessary arrangements to host the founders, some of whom had traveled from as far away as Spain and Denmark.

For a company whose cofounder told Business Insider this year that it has an annual recurring revenue of "nearly $41 million," that's a significant sum.

"It's a very expensive experiment," Dodonov said. He said that, in his view, the return on investment is to see if "the playbook works" — the playbook being his and Hu's belief that building a successful company requires, perhaps above all, promoting yourself on social media.

"Anyone can systematically establish and build support through the power of content," Dodonov said.

• AI researchers don't really know how the models they are building "think."
• They know what goes into them and what comes out, but that's about it.
• A new theory may change that, allowing safety researchers to prevent them from going off the rails.

How AI models think may be easier to decipher than we thought.

An independent researcher who formerly worked at OpenAI published a report on Thursday that some in the AI world are describing as a bombshell. It could have implications as safety researchers scramble to get the jump on AI models that may not have humanity's best interests in mind.

Previously, the prevailing wisdom was that AI "neurons," which, like those in a human brain, light up when an AI model encounters specific stimuli, were ultimately uninterpretable — meaning, basically, that there was no way to guess why a specific neuron would fire when an AI model was exposed to, say, a picture of a cat.

Those building AI models at places like OpenAI and Anthropic don't write instructions for everything a model does. Instead, they train neurons to respond to information and influence one another. As the AI learns, it develops its own ways of processing that information, and humans have not yet cracked the code of how it works.

That means those building AI models know what goes into the AI and what comes out, but not necessarily what happens in between. So if an AI goes rogue, they don't always know why.

The author of the new report, however, says he may have discovered patterns in how the neurons react to stimuli that make sense. Nick Cammarata, who said he left OpenAI in 2023, said AI neural networks are knowable, so long as you look carefully.

In one example, he says that a single AI neuron might react to both a cat and the front of a car. Humans would see those as two very different things, and different neurons would fire as a result. The AI, however, sees something related, perhaps the triangle shape of the two, and a single neuron fires for both.

For years, AI researchers had assumed that AI models packed unrelated things into a single neuron to save space.

"Almost the entire network is interpretable if we look closely enough," Cammarata wrote.

The findings, though undeniably dense, couched in polysyllabic jargon, and still theoretical, could have real-world implications for researchers trying to ensure humans maintain control over AI models.

Understanding how AI thinks is a principal concern for the best and brightest at frontier AI companies. Until recently, they've largely done this by tracking an AI's self-reported "thinking," called chain-of-thought reasoning.

Jakub Pachocki, OpenAI's chief scientist, however, warned in a blog post last month that AI agents could now obfuscate their visible reasoning, leaving open the possibility that an agent could lie to its human minders and the minders wouldn't be able to tell.

"With improved pretraining performance, we also see the models become much smarter even without using verbalized reasoning at all," Pachocki said.

Now, Cammarata's research could give human engineers a new understanding of the AI's internal world — and that's exciting for those worried AI could kill us all.

"I had long since given up on reaching any real level of interpretability before superintelligence arrived," Andrew Curran, a prominent AI analyst, said on X. "I stopped writing about it because of how demoralizing it was. Hard to overstate how big this may be."

• OpenAI's Alexander Embiricos said the AI era gives a choice: work more, or work less.
• He's in charge of Dots, the company's new always-on work assistant.
• AI is often pitched as freeing up time. Embiricos responded to the idea that it might add more work.

As the competitive AI race pushes tech workers into long hours, the head of OpenAI's new Dots product says AI gives people another option: work a whole lot less.

Alexander Embiricos, who leads OpenAI's enterprise product and helms the new Dots agent tool, told a crowd at TechCrunch's Disrupt conference on Thursday that AI's capabilities have created something of a fork in the road for workers. You can grind to meet the moment. Or you can use AI to kick your feet up a little.

"We basically have a lot more choice now," he said.

Embiricos made the comments after a reporter asked about when AI systems will actually deliver more free time. The reporter shared an anecdote where an OpenAI employee used a ChatGPT mobile feature to work in their car and on a plane — not exactly liberated by the new tech.

Embiricos said that the moment has already arrived where a worker can "outsource" tasks to AI and do less work. Then again, some people are so excited about AI's potential that they feel like it's an "interesting, ripe moment" to put in extra hours.

"I think that's why a lot of people are motivated to work very hard right now. That includes myself," he said. "But I think actually there is this choice, and there's totally this other setup where you could delegate a ton more things."

OpenAI is billing its Dots tool as a workplace assistant, always on and capable of proactively handing tasks and sorting to-dos. Embiricos uses his Dot often in Slack, he said, engaging on various threads.

His perspective illustrates how early companies are in their journey to adopt AI tools. Most companies have not yet automated their work tasks, and many workers are still debating how much to use the tools. Early AI adopters, meanwhile, can sprint through some work that used to take days or weeks, especially in software engineering.

Embiricos summed up the situation he feels workers are in: "You either choose: do you expand the ambition of what you can do, or do you do less? I do think we're there."

Have a tip? Contact this reporter via email at scouncil@businessinsider.com, or over text, Signal, Telegram, or WhatsApp at 415-757-8198. Use a personal email address, a non-work WiFi network, and a non-work device; here's our guide to sharing information securely.