Episode Summary
Executive Summary: This episode covers OpenAI’s Dev Day push to make ChatGPT a new internet front door through third-party apps, while also highlighting the privacy, business, and platform-power risks. It then explores OpenAI’s massive AMD chip deal and broader trillion-dollar infrastructure gamble, before turning to Sora’s AI-video slop, copyright backlash, and how the app is being used for playful trolling and potential harassment.
Main Topics: OpenAI Dev Day and the ChatGPT app ecosystem (Priority: 5/5): The hosts report from OpenAI’s developer conference, where the company unveiled apps inside ChatGPT, new models for the API, and Agent Kit. The core idea is that users will increasingly complete tasks without leaving ChatGPT, making it a front door to the web. Privacy risks and platform power in ChatGPT (Priority: 5/5): They compare OpenAI’s app strategy to Facebook’s platform era, emphasizing that ChatGPT conversations may contain far more intimate data than social-network profiles. The discussion warns that sharing memory and conversation data with apps like Zillow or Expedia could create serious privacy and trust issues. OpenAI’s trillion-dollar chip and infrastructure buildup (Priority: 5/5): The episode explains OpenAI’s AMD deal and places it alongside other infrastructure partnerships as part of an enormous bet on compute, data centers, and energy. The hosts stress that OpenAI is trying to lock up future GPU supply and build the physical backbone needed for its ambitions. The economics and circularity of AI financing (Priority: 4/5): They discuss how OpenAI may finance these commitments through fundraising, debt, vendor financing, and future revenue, while noting the growing interdependence among AI companies, chipmakers, and investors. This financial web raises bubble and systemic-risk concerns. Sora 2, AI video slop, and creator backlash (Priority: 4/5): The final segment examines Sora’s popularity, copyright disputes, and the cultural backlash to AI-generated video. Guest Katie Natopoulos describes the app as fun and addictive at first, but also socially risky and potentially useful for bullying or harassment. The limits and future of AI-generated video (Priority: 3/5): The guests debate whether Sora will become mainstream. One view is that it remains mostly a novelty for group chats; the other is that the product will improve enough to gain broader adoption, just as ChatGPT did after early skepticism.
Key Arguments: ChatGPT is evolving from a chatbot into a platform that can host app experiences and potentially become the internet’s new homepage. OpenAI’s app strategy is powerful but risky because users may share extremely intimate conversation history and memories with third-party developers. The OpenAI/Facebook analogy is instructive: platform expansion can create value for developers and the host company, but it also invites privacy crises and trust breakdowns. OpenAI’s chip deals with AMD, NVIDIA, and others show it is not just building software; it is trying to control the compute, hardware, and energy needed for a much larger AI future. The AI infrastructure boom is financially circular and could create systemic risk if growth slows or expectations are not met. Sora is fun and creatively powerful, but its current form is also ripe for trolling, bullying, and copyright conflict. AI-generated video may not go mainstream immediately, but it is likely to keep improving and remain culturally relevant rather than disappearing. OpenAI’s leaders insist they will minimize data sharing and stay careful, but the hosts remain skeptical and believe trust will need to be earned over time.
Data Points: ChatGPT weekly users: more than 800 million - Sam Altman’s Dev Day remarks on ChatGPT’s growth API token throughput: 6 billion tokens per minute - OpenAI’s current API processing volume announced at Dev Day Prior API token throughput: 300 million tokens per minute - 2023 baseline used for comparison Dev Day attendance: about 1,500 people - Estimated crowd at Fort Mason for OpenAI’s developer conference AMD compute commitment: 6 gigawatts - OpenAI’s planned purchase of AMD GPU capacity over several years NVIDIA compute commitment: 10 gigawatts - Referenced earlier OpenAI-NVIDIA deal for GPU supply Total projected infrastructure spend: roughly $1 trillion - Combined scale of OpenAI’s planned chip/infrastructure buildout over the next few years OpenAI stake in AMD: up to 10% - Potential equity OpenAI could obtain through the AMD partnership AMD stock purchase price: a penny per share - Cheap stock rights attached to the deal, contingent on milestones GDP growth contribution: 92% - Jason Furman estimate that data centers and information processing software accounted for most U.S. GDP growth in the first half of the year
Pivotal Quotes: "ChatGPT is the new front door to the internet." — Casey Newton: Explaining OpenAI’s app-in-ChatGPT strategy and its platform ambitions "Ship it or zip it." — Casey Newton: Demanding concrete specs and details instead of vague OpenAI hardware teasers "Sam Altman now has the power to crash the global economy for a decade or take us all to the promised land." — Stacey Razgon (quoted by Casey Newton): Highlighting the scale and systemic importance of OpenAI’s infrastructure spending
Implications: OpenAI is moving from model builder to platform and infrastructure power broker, raising privacy, competition, and systemic-risk questions. For users, the gains may be convenience and creativity; for the industry, the stakes now include compute, energy, regulation, and financial stability.
About Hard Fork
“Hard Fork” is a show about the future that’s already here. Each week, journalists Kevin Roose and Casey Newton explore and make sense of the latest in the rapidly changing world of tech. Unlock full access to New York Times podcasts and explore everything from politics to pop culture. Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. Also, for more podcasts and narrated articles, download The New York Times app at nytimes.com/app.