Sharp Tech with Ben Thompson

Ben and Andrew begin with reactions to Anthropic’s Mythos announcement and Project Glasswing, including thoughts on the security risks, the business benefits of keeping this model private, lessons on the “Boy Who Cried Wolf,” and renewed focus on Anthropic’s relationship with the U.S. government. From there: Anthropic’s new deal with Broadcom and Google, a year of stunning Anthropic success that began in 2024, the threats that Anthropic poses to Microsoft, and where AI can and can’t help with taxes. At the end: How the New York Times is adapting to the future, understanding Sam Altman’s history at OpenAI, and a question on the implications of de-globalization.

Ben and Andrew begin with reactions to Anthropic’s Mythos announcement and Project Glasswing, including thoughts on the security risks, the business benefits of keeping this model private, lessons on the “Boy Who Cried Wolf,” and renewed focus on Anthropic’s relationship with the U.S. government. From there: Anthropic’s new deal with Broadcom and Google, a year of stunning Anthropic success that began in 2024, the threats that Anthropic poses to Microsoft, and where AI can and can’t help with taxes. At the end: How the New York Times is adapting to the future, understanding Sam Altman’s history at OpenAI, and a question on the implications of de-globalization.

Ben and Andrew begin with Q&A on Apple after 50 years, including thoughts on Steve Jobs weaknesses, putting iTunes on Windows, the best Apple ads, Chinese manufacturing counterfactuals, and tech company Mount Rushmore. From there: Thoughts on Apple’s AI bet and the downside risk, the signs that Cupertino sees AI as a disruptive technology, and extended thoughts on the Axios hack and why why AI will make security issue worse in the short-term, but may be the solution in the long run. At the end: Delta chooses Amazon Leo over Starlink, two questions on Anthropic and OpenAI IPOs, why headhunters won’t be disrupted by AI, streaming sports abroad, and new fronts emerge in the AirPods battle.

Ben and Andrew interrupt Stratechery’s spring vacation with a mailbag. First, they discuss the end of Sora, the difference between Sora and Instagram, and where the OpenAI/Microsoft parallels break down. Then: A great take on advertising, ChatGPT engagement farming, Formula 1’s new era, the NFL’s world takeover, and how NBC solved tape delay at the Olympics. At the end: A question about Vision Pro and wives, whether elementary schoolers should have smart phones, Elon’s continued adventures with xAI, a Netflix dating show, LLM-aided dogfooding etymology, and Ben’s (admittedly boring) Taipei routine.

Ben and Andrew begin with the news that OpenAI is shifting away from “side quests” and allocating resources to the enterprise space, including Dropbox history to explain OpenAI’s present, lessons in the enterprise space generally (and what you learn in business school), and OpenAI taking cues from 1980s Microsoft. From there: Talking through Ben’s article on Monday, including the implications of agents and questions about integration as durable differentiation for Anthropic and OpenAI. At the end: Nvidia’s new messaging on inference chips and Groq integration, and a word about winters (and whiners) in Wisconsin.