Apple’s Anthropic Flirtation, Can Meta Build Superintelligence?, AI Browser Wars — With M.G. Siegler

Episode Summary

Executive Summary: The episode centers on Apple’s rumored move to outsource parts of Siri’s brain to Anthropic or OpenAI, with both hosts arguing Apple must modernize quickly or risk losing relevance in AI-native phone experiences. They also debate why Anthropic may be favored despite weaker consumer voice tools, Meta’s high-risk push to assemble a superintelligence team, and the rise of AI-first browsers like Dia as a practical new interface layer for everyday AI use.

Main Topics: Apple reconsidering Siri’s architecture (Priority: 5/5): The hosts argue Apple’s Siri is far behind current AI assistants and that replacing or augmenting Siri with an LLM is long overdue. They discuss the reporting that Apple is testing Anthropic and OpenAI models as the backend of Siri and whether that would be a public launch or a quiet rollout. Why Apple might choose Anthropic over OpenAI (Priority: 5/5): They explore why Apple may prefer Claude despite ChatGPT’s stronger voice product, including Anthropic’s safety-first branding, a cultural fit with Apple, and Apple’s desire not to become overly dependent on OpenAI or any single vendor. Economics and negotiation leverage in AI partnerships (Priority: 4/5): The discussion focuses on Anthropic allegedly seeking a multibillion-dollar annual fee and how Apple might resist that model. They consider alternatives like investment-based compensation and whether rumors about Anthropic could be used to pressure Perplexity in search or browser negotiations. Meta’s superintelligence talent war (Priority: 5/5): The hosts analyze Meta’s attempt to build a new superintelligence team by hiring top researchers from OpenAI, Anthropic, and DeepMind. They debate whether assembling a star-studded team can overcome Meta’s historical focus and execution problems. The limits of scale and the shift toward people and algorithms (Priority: 4/5): They discuss how raw compute alone may no longer be enough in AI. The conversation frames the current race as increasingly about talent, better algorithms, and organizational focus rather than just bigger data centers. AI browsers as the next interface layer (Priority: 4/5): MG describes using Dia, an AI-first browser from The Browser Company, and explains how native chat integration across tabs changes how he searches and analyzes web content. They contrast this with Google’s weaker Gemini integration in Chrome and note that AI browsers may become a major workflow upgrade.

Key Arguments: Apple cannot keep relying on Siri’s legacy architecture when consumer AI assistants are rapidly becoming voice-first, web-connected, and central to mobile usage. Claude may fit Apple better culturally because Anthropic emphasizes safety and control, even if OpenAI currently has the better consumer product. Apple likely wants to avoid permanent dependence on any one AI provider, so it may use multiple partners or keep a hybrid model. A big concern for Apple is hallucinations and privacy; it may therefore prefer an AI provider that can be tightly controlled or layered behind guardrails. Anthropic could struggle operationally if Apple drove massive consumer demand, because Siri-scale distribution would create enormous server and cost burdens. Meta’s superintelligence push is impressive on paper but may fail because it is a newly assembled “super team” without the shared mission or cohesion of OpenAI. Scale alone is not enough anymore; AI leadership increasingly depends on people, algorithms, and product execution. AI browsers are compelling because they let users query all open tabs and web pages natively, reducing friction versus copying content into a separate chatbot. Browsers may become a key battleground in AI because they sit directly between users and web knowledge, much like operating systems do on mobile. Elon Musk’s AI and political moves show how difficult it is to control the systems and institutions you create, whether it is Grok or a third party.

Data Points: Siri age: 15+ years - MG says Apple has been working on Siri for more than 15 years and still has not solved basic assistant tasks well. Potential Apple-LLM rollout timing: 2026 - The reporting discussed suggests Apple would not implement the Siri change until next year. Largest Apple acquisition: $3 billion - Used as a benchmark against Anthropic’s reported annual fee request. Anthropic reported fee request: multi-billion dollar annual fee - MG references reporting that Anthropic wants a large and escalating annual payment from Apple. Tesla stock move: down 7% intraday; about 22%-23% on the year - Mentioned while discussing Elon Musk’s political party announcement and Tesla’s market reaction. Tesla share drop after political party news: 8% - CNBC-reported market reaction to Musk launching the America Party. Training/data center example: Colossus - Referenced as Elon Musk’s large AI data center effort to scale Grok quickly. Meta AI hiring time frame: day one - MG emphasizes that Meta’s superintelligence team has just started, so success is unknown.

Pivotal Quotes: "They've got to replace Siri with something else, whether it be ChatGPT or Claude or Google Gemini or one of the other products." — MG Siegler: On why Apple needs to overhaul Siri rather than keep patching the existing assistant. "I think it's long overdue." — MG Siegler: His core reaction to Apple seriously considering an LLM-powered Siri. "It’s almost the inverse of what we were talking about before. I would just start to slowly roll out the actual Siri version of Siri, the Apple-made version of Siri again with certain types of queries that they feel confident that they can answer now." — MG Siegler: His proposed transition strategy if Apple outsources Siri’s brain temporarily.

Implications: Apple may need to accept outside AI help to stay competitive, Meta’s AI ambitions remain uncertain despite top-tier hires, and AI-native browsers could become a mainstream workflow tool as assistants move from apps into the browser and OS layer.