Is OpenAI’s o3 Model AGI?, Is AI The New Social Media?, Zuck’s Revealing Testimony

Episode Summary

Executive Summary: The episode centers on OpenAI’s O3 release and the claim that it may be a practical form of AGI, while the hosts argue it is better understood as “jagged AGI”: superhuman in some tasks, weak in others. They also frame AI as a new social medium, note memory as a major moat, and discuss Google’s antitrust losses as a possible turning point for big tech and the web.

Main Topics: OpenAI O3 and the AGI debate (Priority: 5/5): The hosts debate whether O3 merits the AGI label, highlighting its multi-step reasoning, tool use, and better performance on many real-world tasks, while acknowledging obvious failures on basic benchmarks and structured tasks. Jagged AGI as a new framework (Priority: 5/5): They endorse Ethan Mollick’s idea that AI may be superhuman in some areas and below-human in others, making AGI feel plausible in day-to-day use even if it does not meet a strict scientific definition. Memory, personalization, and lock-in (Priority: 4/5): OpenAI’s new memory features are presented as a major product shift, making chatbots more useful, more intimate, and potentially harder to leave as they accumulate user context over time. AI as the new social media (Priority: 5/5): The hosts argue that AI is evolving into a social medium: conversational, personalized, and web-connected, replacing the outrage-driven, algorithmic feeds that dominated traditional social platforms. Google’s antitrust setbacks (Priority: 4/5): They discuss Google losing major antitrust cases and consider the possibility of structural remedies or a breakup, noting how the competitive and regulatory environment has shifted. Big tech and the future of the web (Priority: 4/5): The conversation suggests AI may accelerate the decline of the traditional web and search-driven publishing model, while also reshaping how consumers find services, content, and expertise.

Key Arguments: O3 feels like a meaningful step change because it can reason, search, revise, and synthesize in ways older models could not. AGI may be better understood as performance that matches or exceeds most humans on most tasks, not as a binary scientific threshold. The strongest practical evidence for progress is not benchmarks alone, but that ordinary users can now get useful, high-quality results without expert prompting. Memory will deepen user dependence because the model becomes more useful the more it knows about the user. AI is becoming social because people are having back-and-forth, relationship-like interactions with chatbots that also function as media filters and knowledge interfaces. Traditional social media has already shifted away from friends and toward algorithmic content; AI is the next stage of that evolution. Google’s antitrust defeats are more significant now because AI and changing search behavior may reduce the strategic importance of some of its legacy ad businesses. If one lab publicly claims AGI first, it gains major positioning advantages over rivals, which may motivate OpenAI or Google to make that declaration soon.

Data Points: Facebook time spent with friends posts: 17% - Mark Zuckerberg’s testimony about how little of Facebook time is now devoted to friends’ posts. Instagram time spent with friends posts: 7% - Mark Zuckerberg’s testimony about how little of Instagram time is now devoted to friends’ posts. Model name: O3 - OpenAI’s new reasoning model discussed as the centerpiece of the episode. Retailer search error count: Walmart 23 times / Amazon 36 times - An example where O3’s deep research output repeatedly duplicated and confused entities in a list of top U.S. retailers. OpenAI/AGI timing belief: 50-50 - The hosts mention betting markets currently price the chance of an AGI proclamation at roughly even odds. Google I/O date to watch: Week of May 19 - Suggested as a likely moment for Google to respond with major AI announcements. Paid subscribers via ChatGPT: Multiple - Big Technology observed several paid subscribers coming in through ChatGPT referrals. Google antitrust cases lost: 2 - The episode notes Google lost two antitrust cases in the same week, adding to regulatory pressure. Human-like benchmark claim: 99% of humans on 99% of intelligence assessments - A quoted OpenAI trainer’s description of O3’s performance.

Pivotal Quotes: "If you squint, you can see it." — Ranjan Roy / hosts: Used to describe why some people are calling O3 AGI despite its clear flaws. "I think we have to just spend today’s episode answering the question that everyone has on their minds, which is, did Jeff Bezos stage the Blue Origin landing?" — Host: Opening banter before the discussion pivots to serious AI and antitrust news. "The Internet is my best friend." — Host citing Harvard researcher idea: Used to support the argument that AI is becoming a social medium and companion-like interface.

Implications: AI is moving from novelty to infrastructure: a personal, social, and increasingly indispensable layer on top of the web. Expect deeper user lock-in, more competition over memory and product design, and rising pressure on Google, Meta, and publishers as AI reshapes discovery and distribution.