The Attention Mechanism with Andrew Mayne

On this episode, Justin and Andrew break down Andrew’s new role as host of the official OpenAI podcast — including behind-the-scenes insights, why Sam Altman welcomed tough questions, and what it means for transparency in AI. Then, they dive into Tesla’s Robotaxi rollout in Austin, Elon’s sensor philosophy, and what autonomy could mean for accessibility. Finally, they geek out over ByteDance’s shockingly cheap and fast video model, and why the future of animation might cost just a few cents.Chapters00:00 - Intro00:25 - OpenAI Podcast12:05 - Tesla's Robotaxis27:10 - ByteDance's Video Model35:23 - Wrap-up

A look at a New York Times article about poor mental health outcomes from ChatGPT. Also, the first Veo3 advertisement. Chapters00:44 Ethical Considerations in AI Technology03:33 AI's Influence on Storytelling and Audience Engagement06:32 The Balance Between Human and AI Contributions09:23 The Future of AI in Advertising and Marketing

In the wake of WWDC, Andrew and Justin convene to discuss whether Apple's lack of AI advancements is spelling a troubling sign for the company's future dominance in the tech landscape. Elsewhere, Justin hates OpenAI's new Advanced Voice Mode and its constant giggling, Meta drops some serious cash to continue funding their AI initiative, and a quick dip into the Elon of it all.Chapters00:00 - Intro01:19 - Xcode and ChatGPT 07:08 - Apple Intelligence and its New Research Paper18:51 - OpenAI's Advanced Voice Mode25:34 - Meta's AI Investment33:31 - Elon40:32 - Wrap-up

Justin and Andrew dig into a busy news cycle, with Veo 3 stealing the show at Google I/O. After breaking down the company's latest video generation model and some communication challenges facing Anthropic, they arrive at the biggest AI news of the week: OpenAI's hardware tease. With a Jony Ive partnership locked down, Justin and Andrew discuss the legacy of Steve Jobs and how Sam Altman plays into it, what the gadget could actually be, and whether this could be the biggest tech launch since the iPhone.Chapters00:00:00 - Intro00:00:51 - Google I/O and Veo 300:11:37 - Google AI Ultra Subscriptions00:17:45 - Anthropic Controversy00:22:08 - OpenAI and Jony Ive Partnership00:37:50 - Operator o3 update and wrap-up

Justin and Andrew discuss the challenges Apple faces in the AI landscape in the wake of a blockbuster new report from Bloomberg's Mark Gurman. They explore the decline and failed potential of Siri, the evolving nature of search engines, and the potential future of AI integration in everyday technology. Then the pair turn their attention to the implications of new agentic technologies like Model Context Protocol (MCP), which was featured heavily at Microsoft Build 2025.Chapters00:00 - Intro and Nvidia Stock02:12 - Apple's AI Missteps and Market Position09:47 - Siri's Decline and Apple's AI Vulnerability13:45 - The Future of Search and Apple's Position18:11 - The Impact of AI on Search Engine Dynamics23:41 - MCP: A New Paradigm in AI Interaction30:10 - The Future of Browsers and AI Integration

Justin and Andrew dive into the growing debate over whether we’ve already crossed the AGI threshold — or if we’ll even know when we do. From Tyler Cowen’s “AGI moment” with OpenAI’s o3 model to the jagged frontier of modern evals, they explore why defining intelligence is more subjective than ever.Chapters00:00 - Intro02:20 - The AGI Moment10:22 - Is o3 or Gemini 2.5 AGI?14:50 - The Pseudosingularity20:16 - Humanity's Last Exam

Meta announces not one, not two, but three LLaMA 4 models — and somehow the most impressive one isn’t actually available. Andrew and Justin dive into the weird timing of the release, the confusion around Chatbot Arena scores, and why the open-weight community is raising eyebrows. Plus: what does “open” really mean anymore, and are we on the edge of an eval collapse?Chapters00:00 - Intro00:30 - Llama 406:01 - Open Source vs. Open Weight11:54 - AI Monetization21:37 - A Pending Eval Crisis?31:46 - Wrap-up

Andrew and Justin discuss Manus AI from hype to results.

Agentic browsers are coming. What should we — and Google — expect?Chapters00:00 Intro and Grok's Hardcoded AI Bias06:43 Claude 3.7 Sonnet is Here13:06 Agentic Browsers and the Future of Web Interaction19:11 The Impending Disruption of Google Search26:46 Veo 2's Avengers-sized pricing36:11 Outro