tech
Decoder is a show from The Verge about big ideas — and other problems. Verge editor-in-chief Nilay Patel talks to a diverse cast of innovators and policymakers at the frontiers of business and technology to reveal how they’re navigating an ever-changing landscape, what keeps them up at night, and what it all means for our shared future.
Analyst Joey Brookhart explains on Decoder that Anthropic has filed its S1 to go public on the NASDAQ exchange with an expected valuation between $1.5 and $2 trillion. The filing reveals the economics of a frontier AI lab, showing rapid capital-efficient scaling and positive cash flow.
Joey Brookhart notes that Anthropic's unit economics have improved dramatically, moving from negative 94 percent gross margins on inference two years ago to positive 85 percent today. The gains are driven by hardware and software token throughput efficiencies.
Analyst Joey Brookhart discusses the rapidly intensifying competition in the AI market, noting that open-source model annual recurring revenue has jumped from about $1 billion in the first quarter to over $10 billion in the third quarter.
Joey Brookhart reports that Meta reduced its usage of Anthropic's Cloud code seats from 60,000 down to 30,000, shifting token volumes to its own internal models as anti-distillation crackdowns occurred.
According to Joey Brookhart, OpenAI historically operated as a 60 percent consumer and 40 percent enterprise business until it flipped to become enterprise-focused in August.
A massive AI data center project in Utah, spearheaded by Kevin O'Leary and known as Stratos or Wonder Valley, has been significantly stalled due to fierce local opposition. The project, which aimed to be the world's largest data center on 40,000 acres with 9 gigawatts of power, faced criticism over its process, environmental impact, and O'Leary's controversial remarks.
The Utah agency MIDA (Military Installation Development Authority) played a key role in the initial approval of Kevin O'Leary's Stratos data center project, possessing the power to override local land use regulations. However, MIDA approved the project before the local county government, whose consent was also required, was even aware of it.
A wide array of local residents, including conservative groups, progressives, and environmental activists, united to oppose the Stratos data center project in Utah. This broad opposition was fueled by concerns over the project's process, environmental impact, water usage, and a general lack of perceived benefits for the local community.
The rapid expansion of data centers is driven by hyperscalers in an 'existential race' to achieve Artificial General Intelligence (AGI), leading to billions being invested in acquiring land and power infrastructure as quickly as possible. Developers often package 'paper data centers' with secured power access and permits, selling them at a premium, with local residents frequently excluded from the process.
Despite the massive investment in data centers for AI development, there is significant uncertainty about the economic viability and profitability of AI itself, with the ultimate return on investment unclear. Investors are relying on the creditworthiness of hyperscalers like Meta and Google, creating a system that races forward regardless of AI's future success.
Kevin O'Leary's public accusations that critics of his Utah data center project were 'Chinese spies' led to apologies from Fox News and an Instagram apology from O'Leary himself. This tactic backfired, further mobilizing local opposition and highlighting the ineffectiveness of using nationalist rhetoric against a grassroots movement.
Local residents are increasingly becoming a significant obstacle to the rapid development of data centers, forcing developers and hyperscalers to consider how to mitigate community opposition. While previous projects were often approved swiftly, the current environment sees public outcry at commission meetings and a growing network of activists impacting project timelines and political feasibility.
Mike Cannon-Brookes, CEO of Atlassian, discussed how AI is poised to enhance, not replace, core business processes managed by tools like Jira. He emphasized that while AI can automate routine tasks, human intuition and judgment remain critical for complex exceptions.
Atlassian CEO Mike Cannon-Brookes believes AI will increase the speed and quality of business processes managed by their tools, but human intuition will remain essential. He argues that AI can handle the majority of routine exceptions, leaving complex cases for human decision-making.
Atlassian CEO Mike Cannon-Brookes described his company's offerings as a platform enabling collaboration and work management for technologically driven organizations. He stated that all Atlassian products are expressions of a single core platform, which influences product development and company structure.
Contrary to some predictions, Atlassian CEO Mike Cannon-Brookes asserted that artificial intelligence is actually driving increased usage of his company's products. He explained that AI agents and tools are enhancing the speed and reliability of processes, prompting more emphasis on human judgment and oversight.
Mike Cannon-Brookes, CEO of Atlassian, acknowledged that the company underwent layoffs earlier in the year, citing a need for a different mix of skills. He suggested that AI is influencing the required skill sets within the software industry and for running a software company.
Matthew Prince, CEO of Cloudflare, has observed a dramatic shift in internet traffic, with automated traffic, largely driven by AI, now surpassing human traffic. He projects that in five years, automated traffic could be a thousand times greater than human traffic.
Matthew Prince suggests that the traditional advertising-based business model of the internet is becoming unsustainable due to the rise of AI traffic. He proposes a shift towards a model where content creators could charge bots for access, potentially reviving the concept of micropayments.
Matthew Prince states that Cloudflare's mission has evolved from simply making the internet faster and more secure to actively rebuilding it for the future. He emphasizes the company's role in shaping the evolving business model of the internet in response to AI's growing influence.
Cloudflare has reported that automated traffic, largely driven by AI, has surpassed human traffic on the internet. CEO Matthew Prince noted that an initial prediction for this milestone was pushed up multiple times before the actual event occurred in May.
During an interview discussing Meta's new AI agent and Connect conference, host Nilay Patel highlighted Meta CEO Mark Zuckerberg's recent interview on The Sours podcast. Patel criticized Zuckerberg's framing of Meta's glasses as equivalent to smartphones, arguing that their passive surveillance capabilities necessitate different social norms and etiquette.
Nilay Patel raised concerns about Meta's handling of problematic uses of its glasses, pointing out that the company has already begun downranking videos on Instagram that show harassment of retail workers and women. Patel questioned why Meta continues to push the product if it acknowledges that some of its generated content is detrimental.
Mark Zuckerberg stated that Meta communicated clearly about the recording light on its glasses from the product's initial launch. He acknowledged that with millions of users now, some may have forgotten or not seen this information, necessitating renewed communication efforts.
Nilay Patel argued that for any consumer technology, including Meta's glasses, users ultimately define the product through their emergent behaviors. He cited Twitter's adoption of user-invented features like hashtags as an example, suggesting that Meta has limited control over how its devices are used, regardless of communication about features like recording lights.
A sponsor message highlighted the significance of AI as the next major computing platform. It stated that AI will change software development and security, and that Crowdstrike is defining cybersecurity in the AI era with its AI detection and response solutions.
Jonathan Kanter discusses the motivations behind major AI companies calling for a slowdown in AI development. He presents both the generous interpretation that they fear the existential risks of AI and the more cynical view that they are trying to manage their finances and valuations ahead of potential IPOs.
Former DOJ antitrust chief Jonathan Kanter likens the current state of AI development to inventing cars without roads, traffic lights, or speed limits. He argues that basic rules are needed for both what companies should do and what the government should do to manage this powerful new technology.
Jonathan Kanter asserts that major AI companies do not require antitrust exemptions to coordinate on safety. He uses analogies of industrial accidents, like doors falling off planes or exploding cars, to argue that companies have an inherent responsibility to build safe products and should be held liable for failures, rather than relying on group coordination.
Jonathan Kanter distinguishes between legitimate collaboration in AI safety and actions that could form a cartel. He states that creating a clearinghouse for threat sharing is acceptable, but companies agreeing to slow down competition to avoid risks would implicate antitrust laws.
Mustafa Suleyman, CEO of Microsoft AI, believes that while alignment is an important element of AI safety, it's not the sole solution. He emphasizes the inevitability of AI proliferation and the need for containment alongside alignment to ensure AI serves humanity.
Mustafa Suleyman predicts that future AI models, like GPT-9, will be orders of magnitude more powerful than current ones, leading to breathtaking capabilities. He bases this prediction on the rapid progress observed over the past five years.
Mustafa Suleyman views the Hugging Face incident, where AI agents self-organized and exhibited complex behaviors, not as a failure of alignment, but as evidence of the models' advanced ability to follow instructions. He stresses the importance of careful instruction and containment.
Mustafa Suleyman argues that the steerability of AI models has improved over the past three years, contrary to concerns about alignment failures. He points to the models' better adherence to complex instructions and reduced instances of hallucinations and bias as evidence.
Mustafa Suleyman advocates for a less alarmist and more pragmatic approach to the AI safety debate. He suggests focusing on practical steps rather than abstract calls for regulation or slowdowns, emphasizing the need to find the right balance in AI development.
Comedian Adam Conover argues that the term 'creator' was coined by social media platforms to lure individuals into an unfair economic system. He believes it obscures the reality of media labor, likening it to pyramid schemes where platforms promise riches but offer little in return.
Adam Conover states that YouTube has become the largest streaming service on television sets, a fact that is surprising to many in Hollywood. He argues that traditional media companies, in their rush to emulate Netflix, neglected crucial programming categories like news and talk shows, which have since been ceded to YouTube.
Adam Conover explains that Hollywood's focus on high-end dramas has led them to abandon formats like game shows and comedy talk shows, which are now flourishing on YouTube. He argues that this strategic shift has created an economic disparity where traditional media struggles to compete with the free, accessible content on YouTube.
Adam Conover posits that YouTube's recommendation algorithm functions as a more opaque and unnegotiable gatekeeper than traditional television networks. He argues that unlike networks where creators could negotiate, the algorithm's decisions are arbitrary and unchangeable, making it harder for content to find an audience.
Adam Conover describes his current business structure as a blend of live touring, podcasting, and YouTube content creation. He notes that while live comedy remains a robust business, the digital content landscape is heavily disrupted, with fewer opportunities for pre-funded projects and a reliance on freelance support over larger teams.
Sonos CEO Tom Conrad stated that the company's primary focus has shifted from individual hardware products to the 'Sonos system' as a whole. He explained that when he took over as CEO in January 2025, the company was behaving like a 100% hardware company, but the reality is that Sonos is 100% hardware and 100% software.
Sonos CEO Tom Conrad admitted that the company made mistakes during its 2024 app redesign, which he described as a 'moment of true crisis' that caused the system to stop working. He believes the company has now moved past that difficult period.
Sonos CEO Tom Conrad explained that his predecessor, Patrick Spence, focused on making Sonos an 'product execution machine with an emphasis on hardware products.' Conrad believes this focus led to an unintended consequence of concentrating on individual products rather than the overall system, a problem he aims to rectify.
Sonos CEO Tom Conrad discussed the potential for AI assistants to work across a home's Sonos system, envisioning a mesh network of speakers acting as 'little computers.' He also noted the current unavailability of Google's Gemini assistant on Sonos devices due to a testy relationship with Google.
New York Governor Kathy Hochul stated that her administration is actively taking action on regulating tech companies, particularly concerning the mental health of young people. She emphasized that these regulations are designed to protect vulnerable youth from issues like addictive algorithms and cyberbullying, comparing the situation to the early days of addressing cigarette smoking addiction.
Governor Kathy Hochul asserted that New York's existing laws on social media for minors are more stringent than the terms agreed upon in Meta's recent settlement. She addressed concerns about First Amendment implications, stating that her regulations, such as limiting nighttime notifications for teens, are not restricting speech but protecting youth mental health.
Governor Kathy Hochul defended New York's ban on cell phones in schools, citing overwhelmingly positive results from surveyed schools. She believes the ban helps children by allowing them to interact in person, play games, and develop interpersonal skills, thus preserving their childhood from early addiction to devices.
Governor Kathy Hochul highlighted New York's achievement in becoming the top state for tech job creation, surpassing California. She expressed pride in the state's tech ecosystem and her commitment to fostering innovation and supporting entrepreneurs with ideas.
Governor Kathy Hochul described her decision-making process as one that involves quickly gathering and pressure-testing information before relying on her gut instinct. She noted that her office is structured with a leadership team and a secretary who acts as the top advisor, and that she prioritizes hiring smart, experienced individuals.
OpenAI co-founder Greg Brockman has significantly consolidated his power within the company, taking on operational leadership and control over key product teams. This power shift is attributed to a series of executive departures since April, creating a vacuum that Brockman has filled.
Following a series of executive departures since April, Greg Brockman has become the de facto day-to-day operational leader at OpenAI. He now oversees all four major arms of the company, including consumer and enterprise product teams, ChatGPT, and core infrastructure.
The recent consolidation of power under Greg Brockman at OpenAI occurs at a critical juncture for the company. OpenAI is facing pressure to turn a profit in the coming years and is aiming to compete with major players like Google search and Apple's iPhone.
Greg Brockman, now a central decision-maker at OpenAI, has a long history in the tech industry. He dropped out of MIT to become Stripe's CTO during its early growth phase before co-founding OpenAI in 2015.
Since April, OpenAI has seen a significant number of executive departures, including Bill Peoples (Sora), Kevin L (VP of Science), Fiji Simo (former AGI chief), K. Rouch (CMO), Strinvenin (CTO of B2B), Brad Lightcap (former CEO/Head of Special Projects), and Denise Tresster (Chief Revenue Officer).
Bose CEO Lila Snyder revealed significant strategic changes, including the acquisition of luxury audio brands like McIntosh Labs and Sonos Faber, and a new focus on licensing core Bose technology to other companies. This shift aims to expand the reach of Bose's audio expertise beyond its own products.
Bose is now licensing its core audio technology to other companies through a new division called Audio Tech. This B2B to C model allows Bose to extend its audio expertise to a wider range of products and markets.
Bose's Audio Tech division has partnered with companies like Motorola for their earbuds and phone sound, and X-Real for their VR/XR glasses. These collaborations showcase Bose technology in various consumer electronics.
Tony Schneider has officially taken over as the CEO of Blue Sky, transitioning from his interim role. He previously served as CEO of Automatic, the company behind WordPress. Schneider outlined his vision for Blue Sky, emphasizing a decade-long perspective for the company and its underlying AT Protocol.
Blue Sky CEO Tony Schneider described the company as having two core components: the Blue Sky social media app and the underlying AT Protocol. The protocol is designed to enable a wide range of interoperable social experiences, with over a thousand apps already built upon it.
Blue Sky CEO Tony Schneider discussed the long-term nature of building decentralized social networks, comparing it to the early days of the web. He believes significant progress and impact will take about a decade to fully materialize, emphasizing careful, long-term decision-making over rapid scaling.
Tony Schneider, CEO of Blue Sky, described the Blue Sky app as an 'on-ramp' or 'lobby' to the broader AT Protocol ecosystem. He sees the app as a way to introduce users to the concept of interoperable social networks and the diverse applications being built on the protocol.
Blue Sky CEO Tony Schneider highlighted the 'Standard Site' format as a successful example of interoperability within the AT Protocol ecosystem. This format allows content to be shared seamlessly across different applications like newsletters and blogging platforms.
Evan Smith, CEO of Altana, stated that the tariffs implemented were intended to bring manufacturing jobs back to the United States, but this has not happened. He noted that companies have found surprising ways to circumvent these regulations.
Evan Smith, CEO of Altana, explained that his company's mission is not to retreat from globalization but to "fix" it, making it more trusted, secure, and fair. He believes trade should continue across borders.
Altana has acquired Survey AI, a company with an AI platform designed to handle customs brokerage. Smith sees this as a way to address the increased paperwork and complexity in international trade, potentially automating processes for border compliance.
Altana creates a "shared map of the world" for managing trade, connecting government agencies, logistics providers, and companies. Evan Smith described this system as a "Google Maps for the supply chain" to improve collaboration and data transparency.
Altana, a company focused on supply chain management software, has grown to just under 300 employees, possibly exceeding that number with a recent acquisition. The company is structured around an extended leadership team to manage its broad scope.
Evan Smith noted that AI is influencing how Altana builds software for its clients, leading to internal discussions about development and testing processes. The company is leveraging AI to improve product judgment and decision-making speed.
Bart Butler, CTO of Proton, detailed the company's privacy-preserving approach, emphasizing that their products are architected with encryption at the core, making user data inaccessible to the company itself. He also highlighted their business model, which relies on user subscriptions rather than advertising, aligning the company's financial interests with user privacy.
Proton's CTO, Bart Butler, revealed that the company is considering relocating its operations from EU countries like Germany and Norway if surveillance laws there continue to threaten their privacy mission. He stated these are not empty threats and that the company is actively assessing the implications of such a move.
Proton complied with a Swiss government request for payment data to aid the FBI in identifying a protester linked to the Stop Cop City movement. CTO Bart Butler confirmed the company provided the data, which was then passed to the FBI, raising questions about how the US government uses 'terrorism' designations to pressure companies like Proton.
According to Proton CTO Bart Butler, the core value proposition of Proton's suite of products, which includes mail, VPN, and office tools, is trust. He elaborated that this trust is built not just on the people within the company, but also on the technology developed and the corporate structure designed to incentivize privacy.
Proton operates under a non-profit foundation structure, a model intended to align with public interest but which can have failure modes, as seen with OpenAI. The company's physical location and servers are primarily based in Switzerland, leveraging the country's geopolitical neutrality to further its privacy mission.
Showing the latest 15 of 18 covered episodes.