Ternus Gets the Ridge

Yesterday, John Ternus became CEO of Apple. That is one hell of a first day to inherit the future.

Over the first two parts of this story, we've been looking at Apple less as a sequence of CEOs and more as a sequence of terrains.

Steve Jobs got the ridge. He arrived at moments when the map was unreliable, climbed high enough to see where several different technological trajectories were converging, and somehow collapsed the mess into objects the rest of us immediately understood. The iPhone was the cleanest example: an industry full of flip phones, sliders, keyboards and increasingly creative hinge arrangements eventually reduced to a slab of glass.

Then Tim Cook got the valley. Once the direction became clear, he industrialized it. He turned the products into infrastructure, the infrastructure into an ecosystem, and the ecosystem into one of the most powerful technological machines ever assembled.

Now Ternus inherits that machine.

Unfortunately, the map has stopped working again. And that may be exactly why Apple chose him.

The Ground Is Rising Again

For most of the Cook era, Apple could see the contours of the landscape pretty clearly. The phone would get better. The camera would get better. The silicon would get better. The watch would become more useful. The services layer would expand. More pieces of the experience could be brought in-house and integrated.

There were enormous technical challenges inside that progression, but the larger organizing model of personal computing was relatively stable.

A device had an operating system. The operating system ran applications. Humans learned the applications. We clicked icons, navigated menus, filled out forms, moved files around and generally became very skilled at translating what we wanted into the language computers required. AI is beginning to mess with that arrangement. Not because ChatGPT got better at answering questions. That is the least interesting version of the story.

The deeper shift is that the computer is beginning to understand enough about language, context, images, environments and intent that it can start doing some of the translating for us.

Instead of learning the ritual required to accomplish something, increasingly we can just describe the outcome.

That sounds like a software improvement until you follow it far enough.

If the machine understands intent, what happens to the application?

If an agent can choose and operate tools, what happens to the interface?

If context can persist, what happens to the idea that every interaction begins cold?

If intelligence follows you between devices, what exactly is the computer?

And if models can increasingly run locally, what belongs in the cloud at all?

We are reopening questions that the technology industry spent thirty years convincing itself had already been answered.

That is a ridge. And personally, this is the part I find irresistible. I have apparently spent enough of my life wandering toward technological edges that I've become a sort of perpetual ridge dweller. The valley is where the money usually is. The ridge is where the interesting questions are. Ternus just got one of the best ridges I can remember.

The New Phone Origami

Go back to the mobile world before the iPhone.

The chaos wasn't stupidity. Quite the opposite. A lot of very smart companies were trying to answer a question that had not yet collapsed into a stable form.

Should the keyboard be exposed? Hidden? Physical? On the side? Should the screen rotate? Should the device fold? Should it have a stylus? A trackball? A wheel? It looked ridiculous in retrospect because we know which answer won.

But at the time, the fragmentation was evidence that the industry was exploring a large possibility space.

AI looks like that now.

Chatbots, coding agents, autonomous browsers, smart glasses, pins, pendants, local models, cloud models, voice assistants, multimodal systems, computer-use agents, ambient devices, robot interfaces, model routers, memory layers, agent swarms.

We don't have phone origami anymore. We have AI origami.

Everybody is folding the thing differently because nobody knows exactly what the thing is yet.

That matters because somewhere inside all this experimentation there is probably another collapse coming. Some combination of models, silicon, sensors, memory, interface and human behavior will eventually resolve into a form that makes much of what we are doing today look hilariously transitional. The next great interface may not look like an interface.

The next great Apple product may not fit neatly into an existing category.

It may not even be one product. That is the target Ternus has to see.

And He Has to See It Years Early

This is where hardware changes the conversation.

Apple cannot wake up in 2029, notice what everyone wants and ship it six months later.

Whatever major computing experiences Apple intends to deliver at the end of this decade are already exerting pressure backward into decisions being made today.

Silicon architecture. Memory. Sensors. Packaging. Displays. Radios. Batteries. Thermals. Materials. Manufacturing equipment. Supply agreements.

The arrow has to leave the bow long before the rest of us can see the target. And this is why I keep coming back to Ternus being a hardware engineer. He joined Apple's product-design organization in 2001, eventually led hardware engineering across the company and was involved across iPhone, iPad, Mac, Apple Watch and AirPods. He also played a major role during the Mac's transition to Apple Silicon.

That does not magically turn him into Steve Jobs. But hardware people have one enormous advantage when trying to see several years ahead.

They live closer to the constraints.

Software can mutate almost infinitely. Physics is much less accommodating.

There are only so many watts. Only so much heat can leave a particular enclosure. Memory bandwidth has limits. Battery chemistry moves at a certain pace. Fabrication processes have roadmaps. Sensors have size, cost and power constraints. Displays have manufacturing realities. Radio links have latency and bandwidth.

A sufficiently good hardware engineer may not know exactly what software people will invent three years from now, but he can have a surprisingly informed idea about what the physical world will permit them to run it on.

It ain't a crystal ball. But when everything else is moving this fast, physics may be the closest thing we have.

Apple's Hardware Is Already Telling Us Something

And then Apple did something last week that makes this whole conversation considerably less theoretical. It announced the new Mac Studio. Apple didn't pitch it simply as a faster creative workstation. The company literally called it "the ultimate desktop for on-device AI."

The M5 Ultra version can be configured with 512GB of unified memory and 1.2 terabytes per second of memory bandwidth. Neural Accelerators are integrated directly into the GPU architecture, and Apple says the system can run enormous frontier-class models entirely on the device. That alone is interesting.

Then they added clustering. Multiple Mac Studios can now be connected over Thunderbolt 5 using RDMA, creating a shared memory pool across systems for distributed AI inference. Apple says a four-system cluster can deliver up to three times the inference performance of a single machine. It also introduced Core AI, a framework explicitly built to exploit Apple Silicon's CPU, GPU, Neural Engine and unified memory architecture for local models.

That is a very different Mac story from the one we were telling five years ago. These things are starting to look less like personal computers and more like tiny private AI data centers. And the market seems to have noticed before Apple finished repositioning them.

Earlier this year, Mac minis became unexpectedly difficult to buy as people started using them as always-on local AI boxes, particularly for systems such as OpenClaw. TechCrunch reported sold-out base models and long delays on other configurations, while the Wall Street Journal traced at least part of the demand surge to local AI and agent workloads.

This is one of those weird little signals I pay attention to because users sometimes discover what a machine wants to become before the product category catches up.

A Mac mini was supposed to be a small desktop. Then people started stacking them like server blades. Of course they did.

Unified memory, excellent performance per watt, a mature operating system, small footprint, quiet operation, local inference and no token meter spinning every time an agent thinks. Suddenly the little silver box had another identity.

I Accidentally Ran Into the Same Thing

This stopped being an abstract trend for me when I tried to buy my latest MacBook Pro.

I wanted the completely irresponsible configuration. Max everything. Eight terabytes. The kind of laptop purchase where you briefly wonder whether you should have consulted a financial professional or perhaps a priest.

Couldn't get it.

The delivery window just kept moving. And what I was being told was that some of the demand pressure at the upper end was coming from companies ordering high-memory Apple Silicon systems in quantity for AI work. Not one laptop for an executive. Multiple machines at a time.

Eventually I bought the fastest beast I could actually get my hands on.

Separately, I've now encountered businesses looking seriously at Mac Studios, Mac Mini and high-end MacBook Pros for workloads that traditionally would have sent them toward conventional server or GPU infrastructure.

That is fascinating to me.

Because Apple may have spent years building a consumer and creative-computing architecture that happens to be unusually well positioned for a world where enormous amounts of intelligence need to run close to the user. Maybe that was always part of the thesis. Maybe it wasn't.

Technology history is full of architectures that become important for reasons their creators didn't initially prioritize.

Either way, Apple now clearly sees it. The language around the new Mac Studio could hardly be more explicit. AI inference, giant models, shared memory, clustering, local privacy and avoiding cloud token costs are now part of Apple's own product pitch.

The hardware is starting to lean toward the ridge before we know what is on the other side.

But Seeing the Constraints Is Only Half the Job

This is where we have to resist the easy story.

Hardware engineer takes over Apple. Hardware becomes important in AI. Therefore Apple wins. No.

If it were that easy, history would be full of Steve Jobs's.

Knowing what can be built is not the same thing as knowing what should be built.

This is the part I explored recently in Jony, Call Me, mostly through the mildly embarrassing lens of my own workbench. I keep building crude little versions of ideas because I tend to see technologies relationally. I don't see a microphone and think microphone. I wonder why the intelligence layer needs a screen. I don't see an ESP32 and think microcontroller. I see a physical nerve ending. I don't see MCP and think protocol. I see connective tissue.

Then I build the stupid version and see if the idea survives contact with physics.

I've called the underlying cognitive habit unreasonable resolution: seeing several moving systems and getting an unusually strong sense of the shape they're trying to become.

Jobs had that at an extraordinary level.

The components were often somebody else's invention. His talent was recognizing the object hiding between them. That is the unanswered question about Ternus. He may be able to see the hardware horizon beautifully. Can he see the object?

The Target Is Moving This Time

And his version of the problem may be harder than Jobs's.

The technologies leading toward the iPhone moved quickly, but they moved on relatively understandable curves. Displays, storage, processors, mobile networking and batteries improved. There were surprises, obviously, but you could model quite a lot of the terrain.

AI is behaving more like weather. A model architecture changes and suddenly an assumption about required compute moves.

Inference efficiency jumps.

A capability that looked five years away shows up in an API on Tuesday.

Agents get better at using computers and whole interface assumptions start wobbling.

Local models improve and the cloud and edge balance shifts.

New memory architectures alter what can fit on a device.

Someone publishes a paper, somebody else open-sources an implementation, and by Friday there are six startups and a guy on YouTube running it on a Mac mini in his laundry room.

This is not a stationary target. Ternus has to pull the bow back now and aim somewhere around 2029 while thousands of people are actively moving the target in midair. That's brutal. But he does not have to predict every variable.

That's the deeper systems lesson.

When the surface becomes chaotic, move down the stack until you find slower-moving truths.

Humans will still care about latency.

They will care about privacy when an intelligence knows their lives intimately.

Energy will still matter. Memory will matter. Heat will matter. Context will matter. Embodiment will matter. Attention will matter. Trust will matter.

The exact model names will change. Those constraints won't disappear. You navigate the fog using the things that move slowly.

What Is the Next Slab of Glass?

This brings us back to the original iPhone moment. The genius of that product wasn't that Apple won the competition to make the greatest flip phone ever devised. It realized the flip phone was the wrong abstraction.

That may be the question sitting in front of Apple again.

• What is the wrong abstraction today?

• Is it the app?

• Is it the phone?

• Is it the idea that one device contains the experience?

• Is it the cloud account?

• Is it the assistant?

Maybe the mistake is assuming the AI computer needs a screen at all. Maybe the next Apple computing environment is distributed across a phone, watch, AirPods, glasses, Mac and some persistent personal model that treats the hardware as interchangeable surfaces.

Maybe the Mac becomes the private cognitive server in your house or office while lighter devices become its sensory and interface layer.

Maybe Apple Silicon becomes the substrate for a personal intelligence architecture in the same way the iPhone became the substrate for mobile computing.

Or maybe all of those ideas are phone origami. That's the point.

We're still early enough that smart people can be very wrong.

Somewhere in this mess is the thing that eventually looks obvious.

Ternus's job is to find it before it does.

And Then Make Apple Bet

Seeing the future is romantic in retrospect. Committing to it beforehand is terrifying. Because eventually unreasonable resolution has to become an unreasonable purchase order.

Factories have to be tooled. Silicon has to tape out. Components have to be bought. Design teams have to stop exploring ten possibilities and choose one. That is where vision becomes expensive. And Ternus has inherited perhaps the most formidable machine in the industry for converting a decision into physical reality.

Cook built that for him. Apple even reorganized the hardware leadership around the handoff. Johny Srouji, the longtime architect of Apple's silicon strategy, became Chief Hardware Officer, taking responsibility for both Hardware Engineering and Hardware Technologies as Ternus moved upstairs.

Look at that structure. Ternus at CEO. Srouji controlling the combined hardware and silicon organizations. A new Mac architecture openly being positioned around local AI. Cook remaining as executive chairman, still available for the institutional, geopolitical and policy machinery surrounding a company of Apple's scale.

You don't have to believe Apple has solved AI to find the arrangement interesting.

I certainly don't think they have.

But it looks like a company arranging itself around the possibility that the next computing war will be decided much lower in the stack than the chatbot leaderboard suggests.

This Is Why Ternus Gets the Best Moment

Jobs had the hardest imaginative leap. Cook had the hardest scaling job.

But Ternus may have gotten the most interesting moment. He gets a mature civilization and an unexplored frontier at the same time.

Behind him are billions of endpoints, custom silicon, manufacturing capability, developers, services, identity, payments, sensors, retail, customer trust and one of the largest pools of deployable capital in corporate history.

In front of him, the categories are melting. The machine is extraordinary. The map is shit. That's a ridge. In the original Ridge piece, I wrote that the ridge doesn't force you forward. It reveals what's possible. The valley is familiar because we've built it. The ridge is exposure. You climb because the horizon is worth seeing. Apple has reached that point again.

Only this time it isn't some half-dead computer company trying to save itself.

It is one of the most powerful organizations on Earth. That makes the stakes much larger. Because when Apple successfully collapses a computing paradigm, it doesn't merely sell a device. Civilization reorganizes around the answer.

The iPhone changed how we photograph our lives, navigate cities, consume media, date, bank, shop, work, travel, communicate and spend our attention. Thousands of companies and entire industries grew from assumptions encoded into that little slab of glass.

Whatever comes after the smartphone could have consequences at least that profound because this time the machine isn't merely becoming mobile.

It is becoming cognitive. It can increasingly understand. Remember. Interpret. Act. And eventually participate in the world alongside us.

Whoever collapses that into a coherent human experience isn't just designing another gadget. They're helping decide what the relationship between humans and machine intelligence feels like.

Three CEOs, One Arc

And that's where the three stories finally come together.

Jobs arrived at a chaotic edge and somehow saw a coherent future hiding inside it.

Cook took that future and spent fifteen years turning it into infrastructure so reliable, pervasive and familiar that much of the world stopped noticing it was infrastructure at all.

Ternus inherits the result exactly when computing becomes unsettled again.

He doesn't need to be Steve Jobs. Trying to cosplay Jobs would probably be the fastest way to screw this up. But he does need something Jobs had.

He needs the ability to stand above the components, understand the terrain underneath them, ignore most of the noise and recognize what all of this wants to become.

The hardware background may help him see farther. The machine Cook built gives him enormous leverage once he chooses a direction.

What none of us know yet is whether Ternus has that last strange ingredient: the kind of mind that can look at a future full of AI origami and see the slab of glass hiding inside it. That's the test.

Steve Jobs got the ridge and saw something.

Tim Cook built an entire civilization in the valley below.

Yesterday, John Ternus walked to the top carrying everything they left him. He has the silicon. He has the factories. He has the ecosystem. He has the capital. He has the bow. Somewhere out in the fog is the target. And now we find out whether he can see it.

---

Rich Washburn is a technologist, strategist, and Founder & Chief AI Architect of ARIA AI Labs, working at the intersection of AI, infrastructure, communications, and capital. He also serves as Managing Partner and Chief AI Officer at Eliakim Capital.