The hosts argue that a key test for agentic AI on iOS is its ability to perform OS-level tasks with a single prompt, like automatically organizing a messy desktop of apps into logical folders. This trivial-seeming task demonstrates deep OS integration and is a practical benchmark for "Apple Intelligence."
Related Insights
As AI agents become the primary 'users' of software, design priorities must change. Optimization will move away from visual hierarchy for human eyes and toward structured, machine-legible systems that agents can reliably interpret and operate, making function more important than form.
To discover high-value AI use cases, reframe the problem. Instead of thinking about features, ask, "If my user had a human assistant for this workflow, what tasks would they delegate?" This simple question uncovers powerful opportunities where agents can perform valuable jobs, shifting focus from technology to user value.
While complex tasks are the long-term goal, agentic AI like Claude Cowork finds immediate value in simple, one-shot commands like "clean up my desktop." This provides a tangible, low-stakes demonstration of its capabilities for a broad, non-technical user base.
The evolution of AI assistants is a continuum, much like autonomous driving levels. The critical shift from a 'co-pilot' to a true 'agent' occurs when the human can walk away and trust the system to perform multi-step tasks without direct supervision. The agent transitions from a helpful suggester to an autonomous actor.
The test intentionally used a simple, conversational prompt one might give a colleague ("our blog is not good...make it better"). The models' varying success reveals that a key differentiator is the ability to interpret high-level intent and independently research best practices, rather than requiring meticulously detailed instructions.
User workflows rarely exist in a single application; they span tools like Slack, calendars, and documents. A truly helpful AI must operate across these tools, creating a unified "desired path" that reflects how people actually work, rather than being confined by app boundaries.
Instead of an exclusive AI partner, Apple could offer a choice of AI agents (OpenAI, Anthropic, etc.) on setup, similar to the EU's browser choice screen. This would create a competitive marketplace for AI assistants on billions of devices, driving significant investment and innovation across the industry.
A new software paradigm, "agent-native architecture," treats AI as a core component, not an add-on. This progresses in levels: the agent can do any UI action, trigger any backend code, and finally, perform any developer task like writing and deploying new code, enabling user-driven app customization.
While AI models excel at gathering and synthesizing information ('knowing'), they are not yet reliable at executing actions in the real world ('doing'). True agentic systems require bridging this gap by adding crucial layers of validation and human intervention to ensure tasks are performed correctly and safely.
A conflict is brewing on consumer devices where OS-level AI (e.g., Apple Intelligence) directly competes with application-level AI (e.g., Gemini in Gmail). This forces users into a confusing choice for the same task, like rewriting text. The friction between these layers will necessitate a new paradigm for how AI features are integrated and presented to the end-user.
