Apple’s new CEO & how AI understands intent
Experts analyze Apple's AI future under new hardware-focused CEO John Ternus, the strategic implications of Anthropic's deep partnership with AWS for custom AI chips, the evolving landscape of customer intent in an era of AI agents, and the ironic security leak of the powerful Claude Mythos model.
Apple’s AI Future Under New Hardware Leadership
With Tim Cook stepping down, Apple’s appointment of John Ternus, a leader with a deep background in hardware (iPad, AirPods, iPhone), signals a continued focus on its core strengths. The panel discusses whether this is a bullish sign for Apple’s AI ambitions or a step away from a software-first approach.
The Hardware Moat and On-Device AI
While some panelists have been “underwhelmed” with Apple’s AI strategy compared to competitors, they agree that Apple possesses a significant “AI moat” in its hardware. The future of AI is seen as increasingly on-device, which presents unique engineering challenges that Apple is well-positioned to solve.
- Compute and Heat: Running powerful LLMs locally generates massive amounts of heat. Apple’s expertise in materials and hardware design, such as using 3D printed titanium as a heat sink in the Apple Watch, gives it a distinct competitive advantage.
- The Dream of Private AI: The vision of distributed, private AI running on-device is still considered technically feasible. The trend shows that smaller, more efficient models are rapidly gaining the capabilities of their larger, cloud-based predecessors. This could lead to an “app store of adapters” or specialized skills that can be downloaded to on-device models.
- Reliability in AI: Ternus’s background in hardware emphasizes reliability, durability, and repairability. Applying this ethos to AI could be a significant benefit, moving the discipline away from a “guess and check” approach to more stable, dependable systems, which is crucial for the Apple brand.
The Future of the App Store
The rise of generative AI and code-generation tools poses a potential threat to the traditional App Store model. As users gain the ability to simply tell their device what they want it to do, the need to download pre-packaged applications may diminish, fundamentally changing a key differentiator for Apple’s ecosystem.
Anthropic’s Deepening Alliance with AWS
Amazon’s announced investment of an additional $5 billion in Anthropic (with a potential total of $25 billion) is a landmark deal, but the dynamics are complex. As part of the agreement, Anthropic will purchase over $100 billion in cloud services from AWS, highlighting a tightly coupled, strategic partnership.
Custom Silicon and Strategic Lock-in
A key aspect of this deal is Anthropic’s use of AWS’s custom silicon, including the Trainium 2 chip, which is an ASIC (Application-Specific Integrated Circuit).
- Performance and Efficiency: Unlike general-purpose GPUs, ASICs are designed for specific workloads. By tightly coupling their chip design with the architecture of Anthropic’s models, AWS can offer significant performance optimizations and cost-effectiveness that would be hard to achieve otherwise.
- Power Dynamics: While Anthropic gains necessary compute capacity, the panel suggests Amazon currently has the upper hand due to Anthropic’s dependence on this infrastructure. For Amazon, it’s a strategic move to showcase its custom chips as a viable alternative to Nvidia for powering frontier AI models. For Anthropic, it signals a strong commitment to enterprise-grade reliability and stability.
- The Rise of AI Stacks: This partnership exemplifies a broader trend of “stacks” or alliances forming in the AI industry, such as Anthropic/Amazon and OpenAI/Microsoft. These ecosystems create tight integrations from hardware and cloud infrastructure up to the model provider, which could pose a challenge to the open ecosystem by prioritizing proprietary optimizations.
Understanding Customer Intent in the Age of AI Agents
A recent IBM Institute for Business Value (IBV) study, released in conjunction with the Adobe Summit, reveals significant shifts in how consumers make purchasing decisions.
Key Findings and Generational Shifts
The study found a 62% increase in the use of AI assistance for purchasing over the last two years. Counterintuitively, the largest adoption growth is among older generations, with an 82% adoption rate for Gen X and 92% for Boomers. The reasoning is twofold:
- Low Tolerance: These users have a low tolerance for “useless, hostile environments online.”
- Disposable Income: They possess the purchasing power to act on their decisions.
They prefer to delegate tasks to an AI agent that can navigate the complexity for them, choosing the “path of least resistance.”
The Agentic Economy and the Future of Advertising
This shift means businesses must now cater to AI agents, not just human eyeballs.
“It’s very much agents talking to agents and that changes the game.”
- From SEO to Agent Optimization: The focus will move from Search Engine Optimization (SEO) to optimizing product information for discovery by agents. This requires a different approach, as agents are not swayed by “fluff.” They require structured metadata and hard numbers.
- The Role of Brand: The panel questions whether traditional, emotion-based brand advertising will be effective on logic-driven agents. Success may depend on how an agent’s metrics are defined, with a focus on task completion rather than subjective appeal.
- A Hybrid Approach: The winning strategy will likely be a combination of maintaining a broad presence across platforms like Reddit and YouTube (which AIs scan for credible mentions) while also providing the structured, agent-readable data needed for direct comparison and analysis.
The Claude Mythos Leak
In an ironic turn of events, Anthropic’s next-generation model, Mythos—which the company announced would be withheld from the public due to its powerful and potentially dangerous cybersecurity capabilities—was reportedly accessed by a small group on Discord. The group allegedly “rattled digital doorknobs until one opened.”
This incident highlights two key points:
- The Mythology of AI: The leak contributes to the “mythology” of the model, echoing ancient stories of tricksters stealing fire from the gods. The name “Mythos” itself invites this kind of narrative.
- The Messiness of Technology: It serves as a stark reminder that despite the structured nature of code, technology development and security are often incredibly messy and fallible processes.