The speakers judged prediction markets broadly successful in calling election outcomes, though the Texas race was a surprise: prices shifted sharply only about two hours before results came in. The host said Kalshi was 95% accurate by the time polls opened. One, two and even three months before the election, its accuracy was 81%, compared with VoteHub’s 77% average. A guest said those figures seemed impressive because many markets had been closer to 60/40 than 80/20, and suggested giving markets somewhat more weight on average when a candidate has a clear lead.

Favorably relaying Joe Weisenthal’s assessment, the host said prediction markets performed well: they offered a confident forecast ahead of the vote and conveyed incoming data promptly on election night. The host also said Nick Carter and Matt Wong considered the markets accurate that night. The host said Polymarket correctly called every result at one awards show—either the Emmys or the Oscars. He compared that with elections: when markets signal a “blue wave” for months, a correct prediction can drain the suspense. A guest countered that a sporting event is still undecided while it is underway, whereas an awards ceremony may simply announce a result already determined. The speakers also noted that people in and around tech who said both polls and prediction markets were wrong had themselves been wrong.

The hosts discussed the effect of political opposition to data centers on the AI industry. They said more than 40 data-center bills were in process and that Greg Abbott had previously enacted a one-month moratorium. The host called compute campuses a major election loser, except in North Dakota, where voters rejected a data-center ban. He suggested that projects already permitted could become more valuable as permitted sites become scarcer, and that every project would need to be reassessed.

The host said Stargate’s and the Abilene, Texas, campus’s prospects depend on whether they already have permits: in his view, much of the project may have been approved, and a moratorium on new capacity might not reach it. At the same time, cloud-compute companies that have built data centers but sold their capacity under long-term contracts could miss out on higher spot prices and struggle to finance new sites.

The speakers suggested that companies seek regions supportive of data centers. Political support alone is not enough: access to energy, pipelines, roads and the ability to transport large materials also matter. The host said natural gas may be abundant, but infrastructure is still needed.

A guest called the rejection of Proposition 40 in California an undernoticed election result. In the guest’s view, the proposal as written could have destroyed the state’s model for building technology companies: many founders, the guest said, are “paper billionaires” whose wealth is tied up in company shares. The guest explained that even if it had passed, the measure would likely have faced litigation rather than requiring an immediate $50 million cashier’s check. Californians nevertheless voted resoundingly against it.

The guest also cited flaws in the proposal’s wording and said some groups that normally favor such taxes had reason to oppose this version because they would not benefit directly. Another host jokingly framed the defeat as thwarting an attempt to hobble competitors.

The hosts discussed Ben Thompson’s argument for Nvidia: if data-center construction is limited, operators will need to get the most tokens per watt, increasing the value of more efficient chips. Available power could become the binding constraint; upgrading older chips at already connected sites could make better use of existing power capacity. As relayed by the hosts, Thompson sees Nvidia as a likely short-term loser if fewer data centers are built, but sees a long-term opportunity in improving computing efficiency per unit of power.

Shopify’s Harley Finkelstein said that after US voters rejected data-center projects, investment could move to Canada. He cited the country’s clean hydroelectric power, open land and cold winters, calling the opportunity massive and generational. The hosts stressed that this was Finkelstein’s hope, not a confirmed relocation of projects. They also noted that new sites would require infrastructure, cooling and adequate power supply.

The hosts relayed Ben Thompson’s view that China faces the inverse situation: it has ample power and the ability to build more facilities, but lacks leading-edge chips and is investing in that area. They questioned how persuasive the China-competition argument would be if the United States slowed data-center construction. In their view, the prospect of losing ground to China is often raised in response to calls to slow AI development for safety reasons. The hosts suggested that a slowdown might have little effect if projects simply move to other states or regions; the actual scale of any slowdown and the response to it remain uncertain.

The hosts relayed a proposal in the context of the Texas governor’s race: companies could build data centers if they forgo subsidies, promise substantial union employment, guarantee that residents’ water and electricity costs will not rise, and add to supply. They said Gina Hinojosa’s website sets out her Texans First priorities for the next legislative session: requiring American union jobs and giving Texans a say in whether and how data centers come to their communities. The hosts added that some communities might view economic gains and well-paying jobs as worth the trade-offs.

Discussing Elon Musk’s idea of data centers in space, a host speculated that initial computing tests could take place in 2027 and some capacity could come online in 2028. He stressed that this was not a story for the next six months. The host connected the plans to SpaceX’s integration with Starlink and said Starship was being built with that system in mind. Separately, he said SpaceX would attempt to catch the returning Starship on its next launch using Mechazilla, the mechanism with the catching “chopsticks.” These were described as anticipated actions and a forecast, not as an existing space-computing capability.

A host joked that data-center moratoria had “won 43 to 1,” naming Harwood, North Dakota, as the sole outlier. He sarcastically urged construction and real-estate professionals to go to Harwood and secure land and power. He also allowed that other receptive cities and states might exist. The remark was presented as a humorous summary of votes, not as a report of a universal construction ban or an official decision to send all projects to Harwood.

A host relayed an investor letter from Elliott Investment Management comparing today’s market—marked by rising debt, risk-taking and hype around new technologies—to the 1920s before the Great Depression. The host said the firm warned that nearly every alarm bell was ringing for a substantial bear market and coined FOGO, “fear of getting obliterated,” as the opposite of FOMO, “fear of missing out.”

Elliott, an $80 billion firm, had flat returns in the quarter ended September 30 and gains of less than 2% over nine months. The host compared that with nearly 13% total return for the S&P 500 over the same period and about 3% for short-term Treasury bills; the firm also lagged a composite hedge-fund index. As relayed by the host, Elliott’s letter pointed to US budget deficits, persistent inflation, private credit and AI hype, as well as market distortions from zero-day options trading and meme stocks.

The firm said it was focused on preserving capital and preparing to act if turmoil created opportunities. The host said Elliott’s unprofitable third-quarter areas included distressed debt and portfolio hedges related to gold, currencies, stocks and interest rates; profitable areas included commodities arbitrage and credit hedges. The firm identified potential opportunities in activist healthcare investing, energy and, eventually, buying from forced sellers in dislocated credit markets. Its recent activist positions included PepsiCo and Synopsys.

The hosts said Prop 38, placed on the ballot by California Institute for Immunology and Immunotherapy co-founder Gary K. Michelson, passed and provides $4.2 billion to a three-person nonprofit. They relayed doubts expressed by Ahmad of Mercury about a measure proposed by that nonprofit and approved by popular vote directing such a large sum to it. One host speculated that the organization would distribute nearly all the funds to research groups; this was an expectation, not a confirmed plan. The discussion also included joking remarks about a possible conversion to a for-profit organization.

Ferrari again raised its full-year guidance, citing stronger-than-expected demand for personalization, lower currency headwinds and increased third-quarter deliveries of newer, more profitable models. The company expects revenue of about €8.56 billion, up from a previous estimate of around €7.6 billion, and adjusted earnings of up to €2.29 billion; a host cited a 30% margin. Ferrari delivered 3,368 cars to customers during the quarter, 33 fewer than a year earlier. The company said its order book extends into 2028 and that two more launches were planned before year-end. The participants were unsure which models the two launches referred to; they speculated about the Luce and a road-legal version of the 296 GTB Challenge.

Deliveries of the Amalfi, 296 Speciale, 12Cilindri Spider and Purosangue increased, while deliveries of the 849 Testarossa and 296 Speciale Aperta continued to ramp up. Deliveries of the 849 Testarossa Spider began, and F80 shipments progressed as planned. Ferrari also phased out the Roma Spider, 296 GTS and SF90 XX family. Shipments rose by 38 vehicles in Europe, the Middle East and Africa, by 5 in the Americas and by 29 in the rest of Asia, but fell by 105 in mainland China, Hong Kong and Taiwan.

The speakers suggested that a local market downturn or stronger competition from Chinese automakers might have affected demand in China; Chinese cars may rival supercars on specifications such as 0-to-60 mph acceleration. They also speculated that buyers in China might have less reason to wait for the forthcoming Ferrari Luce if Chinese EVs, including BYD, offer comparable performance. These were proposed explanations, not established causes. In a discussion of the 849 Testarossa’s styling, one participant offered the personal view that it looks much better in customers’ individual specifications—especially all black—than in the version used for its presentation.

The European Commission plans to lower its assessment of money-laundering and terrorist-financing risk in the art market and antiques trade from high to moderate, compared with its 2022 assessment. According to the discussion’s account of January 2026 meeting minutes, the change is in the forthcoming supranational risk assessment.

The Commission’s Financial Markets Directorate recommended the downgrade, citing limited EU evidence linking high-profile museum thefts to money laundering or terrorist financing. The Commission nevertheless acknowledges persistent vulnerabilities related to opacity and anonymity in the sector. As described in the discussion, a moderate rating still calls for mitigation measures, greater awareness and stronger oversight. The speakers said one art-market expert considered even the moderate rating too high.

Pangram is expanding its AI-generated-content detection to images and video. Max Sparrow of Pangram said the team had been working on the capability for months and that it is now available. The discussion also noted that detection is complicated by the variety of models and differences between diffusion-based and autoregressive architectures.

Discussing AI agents, Apoorva Mehta said Instacart already saves customers hours each week by sparing them the trip to a store, the search for parking and checkout lines. The hosts suggested that reducing a few screen taps may be a smaller improvement than the original time savings from grocery delivery.

Mehta did not dismiss agents’ value: he said they will be useful and that he is particularly interested in agents that can understand a family’s dietary needs and preferences, then proactively help people eat more healthily. He stressed, however, that this use case has not yet been fully realized and that the field is at an early stage. He called Instacart an “original agentic company” in the sense that a human shopper already carries out a task on a customer’s behalf.

In Mehta’s view, companies that have solved the difficult operational problem beneath their services can improve their own interfaces with natural language and other features. A host separately opined that improving individual apps may be more promising than a general-purpose agent told to do the shopping.

Apoorva Mehta said his new company is building long-running agents that continuously seek information and improve their results over time while using less cost and time. He compared an agent’s accumulation of context to onboarding an employee: as the employee gains context and completes more projects, performance improves. In his account, hardware can be one way to provide an agent with more context.

Mehta expects such agents to appear in consumer products too, but thinks their commercial value will become clearer sooner in enterprise settings: companies are willing to pay for faster and better information when it gives them a competitive advantage. He said it is not yet clear how to make consumer AI services sustainably profitable and predicted that this problem would be solved over the next few years. For now, he considers enterprise applications more compelling.

Discussing Abundance, the guest said machine learning and AI have mostly been used to identify statistical patterns, while reasoning systems could potentially analyze fundamental questions—for example, a company’s competitive moats and the durability of its growth. He said such systems and agentic capital allocation remain exceptionally hard to build; obvious opportunities get priced in quickly, and their alpha decays fast.

Asked about hardware and a race for compute, he said Abundance had thought about hardware but expected to build advantages in other ways too. The company allocates capital in public markets and in other areas where, in his view, returns may be larger, more durable, and more impactful. He did not specify those other areas.

The guest also described his goals for the company: after leaving Instacart, he wanted to build a company with the potential to become a trillion-dollar business and, in five to six years, hopefully a multi-trillion-dollar one, while benefiting humanity. These were ambitions and an assessment of potential, not achieved results. He said AI is redrawing business boundaries: companies may focus not only on a product such as SaaS but also on outcomes and new kinds of holding companies.

Alex Spiro, a lawyer and former prosecutor who advises founders and other clients on legal and business matters, said he believes AI will eliminate an intermediate layer of work in many areas, including his day-to-day work. In his view, that could make it faster to establish facts and synthesize information, helping parties decide sooner whether to settle a dispute or take it to trial.

Spiro predicted that some drawn-out litigation in the middle could disappear: some cases would settle quickly, while others would go to trial. He said there might be more trials if analysis makes it easier to identify which cases should be pursued to the end. This was a prediction, not a report that the change has already occurred.

Spiro said he has not changed his approach to trying cases or presenting them to juries in 20 years: human understanding does not change because of AI. In his view, technology may change how many people gather information and how quickly it reaches a lawyer, but not how the lawyer addresses jurors. Noting that he is not an AI expert, he said extracting information from documents seems less susceptible to error than working with case law, where kinks may remain. As an example, he described a police report in which a system could say Detective Lamb appeared three times in a case—January 3, 8 and 12. He believes this kind of factual analysis can speed answers to routine investigative questions. When AI generates case law, his team double-checks it for hallucinations.

Asked whether clients might become overconfident after using AI, the lawyer said he has long dealt with clients who believe they know the law better than he does. In his view, people should not choose their own legal arguments or medical procedures: personal involvement can undermine sound judgment, making professional expertise and emotional independence important. He said he has offered clients who wanted to represent themselves the chance to deliver their own opening argument while he watched; he said that usually changed their minds.

Discussing how founders approach lawsuits, the lawyer said some entrepreneurs want to sue everyone, while others would shut down a company after receiving a demand letter. A counselor’s role, he said, is to help them find a balance between those extremes. Companies working on the technological frontier may face challenges from competitors, regulators who do not understand their work, or short sellers. His advice to founders: do not panic, and be prepared to defend your position when conflict arises.

The lawyer said he decided to attend law school at 21 almost on a whim and has not found a business activity that matches the excitement of legal work. But after a long legal career, he said a new chapter is beginning: his business career is developing on its own trajectory and will take up more of his time over time. He added that he is interested in activist investing and in creating value by bringing companies together once they are mature enough—not at the product-development and engineering stage.

The lawyer said he and Tom Brady are partners in sports and health-and-wellness businesses and see each other weekly in Florida. He singled out Escape, a robotic massage company. According to him, its machines are available at Equinox, gyms and hotel chains around the country; customers can book online and receive a massage from a robot rather than a human therapist. He described the method as drawing on Brady’s approach and his pliability coaches, and said the company is expanding. He predicted that more people would be able to get massages using the Tom Brady protocol in the coming months and years, and that the company could affect health care.

Discussing the idea of investing both in frontier technology and in assets less exposed to AI, the lawyer did not call it his overarching investment thesis but described it as a reasonable hedge. His own view is that beachfront property and unique sports teams could continue to appreciate as wealth is created in the AI boom and a major wealth transfer unfolds over the next 10 years. In his view, the pool of people able to afford such assets may grow while supply remains limited. He qualified that outlook: it holds as long as valuations do not become too extended.

Traversal, which is building an AI system to diagnose outages and help fix them, raised $120 million at a $1 billion valuation. Co-founder and CEO Anish disclosed the funding in an interview.

Traversal began in January 2024. Anish, a researcher and Columbia University professor, said the company draws on causal machine learning: its aim is to identify cause-and-effect relationships in data when investigating the root cause of an incident.

Anish said Traversal focused on large customers from the outset, including American Express and Pepsi, because the problem is especially pronounced in large-scale infrastructure with many observability tools and thousands of engineers. Traversal connects to code and telemetry systems such as Datadog and Splunk, and investigates abnormalities. Pricing is usage-based: customers pay according to the number of investigations run.

Anish said the product is becoming more than a tool for responding to incidents. Companies want to use the platform while code is being written to assess whether it could break in production, and when alerts are created. Coding agents are becoming some of its biggest users, he said: they query Traversal to find out whether a deployment could cause an outage or security exposure.

Anish also tied companies’ heightened anxiety to the fact that reliability failures and security incidents can produce similar symptoms, such as users being unable to log in or a slow website. In his view, coding agents reduce the context available within companies, while security patches themselves introduce changes that can lead to further incidents.

New York-based writer and Vanity Fair special correspondent Derek Blasberg discussed generative AI’s effect on photography and Richard Avedon’s legacy in an interview. In Blasberg’s view, AI works for joke images and visual experiments, but he prefers real photography for meaningful, emotionally valuable pictures—whether a family photograph or a business image. He described a Gucci campaign with a GTA-like aesthetic as strange and said the reaction to it was negative. The interviewers noted that e-commerce photography may be an area where AI is already significantly changing the practice. A guest said commercial e-commerce photography is being profoundly disrupted, particularly in Los Angeles: an entire industry photographs new clothing collections in different styles and on different models, and in the guest’s view that business is facing a “tsunami” of change.

The panel also discussed using AI and OpenAI to scale garments and determine sizing: designers can feed measurements in to make the work faster, cheaper and less labor-intensive. One speaker stressed, however, that the designer remains the source of inspiration.

Discussing Avedon, Blasberg said he photographed Merchant Marine sailors against a white background, and that early work influenced the recognizable style of his portraits, including John F. Kennedy’s official portrait and photographs of Marilyn Monroe. After World War II, Harper’s Bazaar editor Carmel Snow commissioned Avedon to capture postwar Paris. His photograph “Dovima and the Elephants” sold at auction for more than $1 million, and a copy later resold privately for more than $2 million.

Blasberg said Avedon’s estate restricts posthumous printing of his photographs: the prints shown at an exhibition were made and signed by the photographer during his lifetime, which creates limited supply. He contrasted this with images whose negatives and rights are sold to Getty Images, giving as an example a Slim Aarons photograph available there for about $15.

Blasberg said a Richard Avedon documentary directed by Ron Howard will arrive on Netflix on November 11. Avedon died in 2004 at age 81 while working on The New Yorker’s “Democracy” project; shortly before his death, he photographed the young senator Barack Obama.

A participant suggested that exhibitions people must attend in person will become more valuable in a society increasingly reliant on digital technology. He linked the growth of live sports to the idea that, after COVID-19 and amid the rise of AI, in-person experiences have become more meaningful and enriching. As an example, he cited the $12.5 billion sale of the Los Angeles Lakers. In his view, people attend events not only for the experience itself but also to create content and take photos that show they were there.

A TBPN guest described the postwar years—and especially the 1960s “youthquake” culture—as a golden age of fashion photography. He said Harper’s Bazaar’s publisher asked photographer Richard Avedon to reshoot images of Asian model China Machado with a white model. Avedon, whose contract was nearing renewal, threatened not to renew it unless the photographs ran. According to the guest, Machado became the first non-white model in a nationally syndicated fashion magazine not because the publication had suddenly turned progressive, but because Avedon was prepared to leave. The guest also noted Avedon’s influence on Calvin Klein’s advertising style, including early CK One ads.

Participants said the weakened economics of publishing, particularly fashion magazines, make it harder to secure photographers with exclusive contracts. A guest contrasted the current situation with practice as recently as the 1990s, when publishing houses could tie photographers to million-dollar deals. He cited Annie Leibovitz, who, he said, shot only for Vanity Fair and Vogue. In another recent case discussed on the program, one photographer reportedly accepted commissions from about four competing publications; similar looks, including the same Chanel dress, drew criticism. The guest suggested that the photographer’s lack of an agent may have contributed to the situation.

A participant said that tech media and new media businesses often fail to understand the difference between a good photograph and a great one. He expressed a wish to create a magazine and suggested there are people in the technology industry with interesting stories who need high-quality portraits. The panel jokingly discussed the idea as a possible future project but did not announce a launch.

Participants suggested that people may eventually feel nostalgic for early AI art from 2023, with its six-fingered hands and conspicuous errors. One speaker said such visuals are already being used as kitsch and irony, citing Katy Perry’s appearance at the Met Gala with an image of a six-fingered hand. In his view, early AI imagery was memorable because it was easy to recognize and did not fool viewers. The panel also observed that nostalgia cycles seem to be speeding up: styles once took decades to return, whereas people are now nostalgic for relatively recent AI artifacts.