Apple's new CEO just told us exactly how the agent economy will arrive—and why the companies you think will win it are thinking about the problem backward.
The Summary
- Apple execs told Wired they "never think about shipping a technology," only products and experiences—while Wired wants them to build "a killer AI product"
- The piece sparked 222+ points on Hacker News, hitting a nerve about the product vs. platform debate
- Apple's framing reveals the real shift: by decade's end, you won't tap apps—your agent will have the car waiting before you ask
- The strategic question isn't "what's your AI product" but "what survives when AI becomes the interface layer for everything else"
The Signal
John Ternus, Apple's incoming CEO, told Wired's Steven Levy that AI is "an immense kind of inflection point" but framed it as just another technology transition Apple has navigated. The company line: they ship products, features, and experiences, not technologies. Customers shouldn't think about what makes something possible—they should just use it.
Levy pushed back hard. He wants Ternus to deliver an iPhone-scale product for the AI era. The argument: by the end of this decade, people won't open Uber or Lyft to get home. They'll tell their always-on AI agent, or the agent will predict the need and have the car waiting. Levy's vision of AI disrupting the entire iPhone ecosystem is the bear case for every app developer who built on iOS.
"AI threatens to disrupt the entire iPhone ecosystem."
But Ternus and Levy are arguing past each other. Levy wants a discrete AI product—ChatGPT in a new form factor, Claude with better hardware integration, something you can photograph and review. Ternus is describing something harder to see: AI as substrate, not destination. The technology that powers the next product category, not the category itself.
Here's where it gets interesting for Web4:
- Apple's historical pattern: Apple II → Mac → iTunes → iPod → iPhone → iPad. Each built on the previous infrastructure.
- The agent economy pattern: language models → chatbots → workflow automation → autonomous agents → ambient intelligence that acts before you ask.
- The common thread: infrastructure becomes invisible when it's working.
The Hacker News discussion drew 78 comments because this tension is playing out across every sector. Startups are building "AI products" when they should be building products that happen to use AI. Enterprises are buying "AI solutions" when they should be buying solutions that happen to be smarter than the old ones.
Apple's framework suggests the agent economy doesn't arrive with a bang—one killer AI product that changes everything. It arrives when your existing products get so much better at prediction and execution that the AI layer disappears. When you stop thinking about "asking your AI agent for a ride" and just think about getting home, the same way you stopped thinking about "using GPS" and started just navigating.
The Implication
Watch what ships in the next 18 months from Cupertino. If Ternus is right, it won't be called an AI product. It'll be called a better camera, a smarter assistant, a more capable watch. The AI will be in the how, not the what. And if he's wrong, if Levy's "killer AI product" thesis is correct, we'll know because Apple will break pattern and ship something entirely new rather than making something existing dramatically better.
For builders: the companies winning Web4 won't be the ones with "AI" in the name. They'll be the ones whose products work so well you forget there's an agent in the loop. That's the real product challenge—making autonomous intelligence boring enough to trust.