2024 In Review, 2025 Predictions — With Casey Newton

Episode Summary

Executive Summary: Alex Kantrowitz and Casey Newton review 2024’s AI and tech progress and predict 2025. They argue AI is genuinely advancing but still struggles with productization; agents, Google’s AI turnaround, TikTok’s fate, antitrust under Trump, Waymo expansion, quantum, and Neuralink are key areas to watch. The tone is optimistic about useful AI, skeptical of hype, and focused on which products will actually matter.

Main Topics: 2024 AI progress vs. product usefulness (Priority: 5/5): The hosts agree AI made real technological gains in 2024, but most consumers still struggle to extract clear ROI. Casey argues the market remains too focused on profitability skepticism and too little on how quickly users and revenue grew. AI agents as the big 2025 feature (Priority: 5/5): Casey predicts the major visible AI story of 2025 will be agents: systems that can navigate email, browsers, and office workflows. He thinks the demos are clunky now but likely to become useful for productivity tasks. Apple Intelligence and Vision Pro as flops (Priority: 4/5): Both Apple AI and Vision Pro are portrayed as underwhelming. Apple Intelligence is mocked as useful mainly for low-value summaries, while Vision Pro is seen as technically impressive but lacking a compelling daily use case. Google’s AI catch-up and product problem (Priority: 5/5): Google’s models are improving, and Gemini tops benchmarks, but product experiences lag OpenAI and Anthropic. Casey argues Google may be “teaching to the test” and needs better product teams to translate model strength into real utility. Platform shifts: TikTok, Meta, Amazon, and antitrust (Priority: 4/5): The episode predicts major policy and platform effects in 2025: TikTok could be banned and then rapidly reinstated; Meta and YouTube could benefit; Amazon may gain from restrictions on Temu and Shein; antitrust enforcement under Trump may be inconsistent and transactional. Emerging technologies: Waymo, quantum, Neuralink, and Snap (Priority: 4/5): Waymo is framed as the most immediately impressive consumer AI experience and likely to expand. Quantum computing is advancing but not ready for broad use, Neuralink is promising but still early, and Snap may be an acquisition target—possibly for Apple.

Key Arguments: AI’s value is becoming clearer in workflow assistance, brainstorming, editing, and document analysis even if it remains unreliable for factual verification. The lack of a GPT-2 to GPT-3 or GPT-3 to GPT-4 level leap in 2024 does not mean the technology is stalling; other scaling methods may continue progress. AI agents may be janky at first but can still unlock useful productivity gains by handling routine tasks like inbox triage and drafting responses. Apple Intelligence feels optimized for elite use cases rather than normal users and therefore misses real-world utility. Vision Pro is not primarily failing because of price; it is failing because there is still not enough to do with it. Google’s model quality is improving, but its consumer products still underperform because the company may be optimizing for benchmarks instead of usefulness. The OpenAI-Microsoft relationship will likely persist because each side needs the other’s strengths: OpenAI needs infrastructure, Microsoft needs frontier models and product credibility. A TikTok ban would likely shift attention and culture toward Meta’s Reels and YouTube Shorts almost immediately. Trump-era antitrust enforcement may look less ideological and more personality-driven, depending on whether he likes or dislikes specific companies. Waymo is the clearest example of AI becoming visibly useful to consumers and could change transportation habits wherever it expands.

Data Points: ChatGPT weekly users: 300 million a week - Casey cites OpenAI’s reported weekly usage growth as evidence of rapid adoption ChatGPT launch-to-scale: Just over 2 years from zero to 300 million weekly users - Used to argue AI adoption has been unusually fast Vision Pro 2024 sales projection: About 420,000 units - Casey uses this to frame Vision Pro as not a runaway hit Vision Pro revenue: Close to $1.5 billion - Estimated first-year revenue from Vision Pro sales Waymo expansion cities: Atlanta, Austin, Miami, and Los Angeles - Examples of 2025 expansion indicating broader rollout Temu and Shein share of Gen Z Black Friday purchases: About 50% - Used to argue Amazon views these competitors as a major threat Apple Intelligence summary behavior: Summarizes 2 or 3 items, often stripping personality from messages - Personal anecdote illustrating weak product value Quantum computing scale: Chips must be kept close to absolute zero - Explains why quantum progress is technically difficult and slow Neuralink patient count: Dozens expected in 2025 - Alex’s prediction for adoption of brain-computer interfaces

Pivotal Quotes: "we are going to see one really notable set of new capabilities in what they call AI agents" — Casey Newton: Casey’s central 2025 AI prediction "Apple intelligence flop of the year" — Alex Kantrowitz: Alex’s verdict on Apple’s AI product during the predictions segment "Waymo is the best consumer technology experience that most people have not had" — Casey Newton: Casey explains why he sees self-driving cars as a transformative consumer experience

Implications: 2025 may be less about breakthrough model hype and more about which AI tools become genuinely useful. Expect agent demos, policy-driven platform shifts, and growing consumer exposure to Waymo, while many flashy products still struggle to prove daily value.