Apparently I've Been Trying to Remove the Interface
I may have an obsession.
I hadn't really thought of it that way until recently, because in my head these have always been separate things. There are the little electronics projects I build because I like making things. There are the wearables and sensors and absurdly small enclosures that inevitably turn into exercises in battery management, heat, packaging, and wondering why I didn't just choose a larger box. Then there's all the AI work, the writing about Apple, the stuff about apps disappearing, ambient computing, agents, voice interfaces, and this recurring idea that eventually the computer should understand what we're trying to do without making us operate it like a piece of machinery.
For years, those lived in different drawers in my brain.
I'm starting to think they're all the same project.
Apparently, I've spent a surprising amount of time trying to figure out how to remove the interface between a person wanting something and a computer understanding what they mean.
That realization hit me again recently when I came across EITWatch, a research project out of Northwestern University. It's exactly the kind of thing guaranteed to send me down a rabbit hole. The researchers built an electrical impedance tomography system into the back of a smartwatch, using eight electrodes against the wrist to measure tiny changes in the electrical characteristics of tissue as the hand moves. The hardware samples multiple impedance paths dozens of times per second and uses those changing patterns to recognize hand gestures.
Just typing that makes me happy.
Not because I think electrical impedance tomography is necessarily the future of the Apple Watch, or because everyone will be wearing one of these things in a few years. I have no idea. That's not really the interesting part.
What gets me is the direction of travel.
For most of computing history, the machine has sat there waiting for us to explicitly operate it. We learn where the buttons are. We learn menus. We learn applications. We learn shortcuts. We learn the grammar of whatever interface the computer gives us, and then we translate what we want into the sequence of operations required to make the machine do it.
EITWatch moves that relationship a little bit upstream. The hand moves and the computer can potentially understand something about that movement before you tap an icon, press a button, or issue a formal command. The body itself starts becoming part of the input system.
I find that fascinating, probably because I've been playing around the edges of the same idea for longer than I realized.
Around the early Apple Watch era, I started building a tiny wearable project called FireFly. It went through a bunch of iterations, because that's what happens when you actually build things. The original idea mutates. Something doesn't fit. Something consumes too much power. You find a different microcontroller. The enclosure becomes the project. The battery becomes the project. At some point you realize you've spent four hours trying to save one millimeter.
There were versions with RGB lighting, NFC, Bluetooth, different physical configurations and increasingly ridiculous attempts to pack functionality into something genuinely wearable. At one point I went far enough down the ultraviolet rabbit hole to file a provisional patent around a roughly 233-nanometer UV wearable concept.
The provisional eventually expired.
There was no startup montage afterward. No heroic acquisition. No licensing deal. Nobody backed a truck full of money into my driveway.
It was just another strange little thing I'd become interested enough in to build.
Looking back, though, what interests me about FireFly now isn't really the UV or the LEDs or any particular feature. It was the question underneath the project. How much useful capability could I take out of the traditional computer and put directly into the environment around the person?
I didn't phrase it like that at the time. I was mostly trying to make the damn thing fit.
But I think that's what I was doing.
And it's the same reason I keep coming back to this idea that apps, at least as we currently understand them, are eventually going to become much less important.
When I say apps are dead, I don't mean software disappears or that everyone deletes the App Store next Thursday. Applications are going to exist for a very long time. What I think changes is their role as the mandatory interface between human intention and software capability.
Consider how bizarre our current arrangement really is.
I want a ride, so I translate that intention into "find Uber, open Uber, interact with Uber's interface." I want to move a flight, so I find the airline's application, log in, locate the reservation, figure out where they moved the button since the last update, and walk myself through their workflow. I want to send someone money, so once again I convert an intention into a sequence of software-navigation instructions.
We've gotten so accustomed to this that it feels natural.
It isn't.
For decades, human beings have been adapting themselves to the organizational structure of computers.
Artificial intelligence creates the possibility of reversing that.
Instead of me figuring out which software contains the capability I need and then learning how that software expects me to ask for it, the computer can increasingly begin with my intention and work backward.
That's the part of AI that interests me far more than another chatbot.
And this is where Apple keeps wandering back into the story.
I'm hopelessly an Apple person. I have more or less the entire ecosystem. Macs, iPhones, Watches, iPads, AirPods, HomePods. If Apple makes a device, there's an unreasonable probability that one is somewhere in my house.
Except the Vision Pro.
There has to be a line somewhere.
So naturally I've spent an equally unreasonable amount of time wondering what Apple is going to do with artificial intelligence. That's occasionally been a painful hobby because Apple is almost impossible to predict. They're weird as shit. They can spend years looking like they're playing some beautiful ten-year strategic game and then suddenly do something that makes you wonder whether the departments in Cupertino are communicating through carrier pigeons.
The Apple Intelligence rollout was a good example. Apple announced a vision that was enormously compelling: personal context, a much more capable Siri, awareness of what was happening onscreen, the ability to operate across applications, intelligence on the device when possible and private cloud infrastructure when necessary.
Then the execution got messy. Timelines slipped. The Siri everyone thought was arriving didn't arrive. Some of Apple's marketing got sufficiently ahead of the actual product that lawsuits followed. For a while, I think it's fair to say Apple lost its grip on the AI story.
That's why I've never been particularly interested in writing some retrospective about how I "predicted" Apple's AI strategy. Predicting Apple is a great way to eventually make yourself look stupid.
What I have been doing is hoping.
I've been hoping Apple would build a particular kind of computer because Apple is unusually well positioned to build it.
I don't particularly want Apple to give me another AI application. I don't even care very much whether Apple itself builds the smartest model in the world.
There are already companies spending astonishing amounts of money doing that.
What I want Apple to do is make all of those things disappear.
I want to be able to tell my computer, "Move my flight to tomorrow afternoon, but make sure I can still make dinner," and have that be the interaction.
Not the beginning of seventeen interactions.
The system should know, within whatever permissions I've given it, which flight I'm talking about. It should know what's on my calendar. It should know where dinner is. It should know how long it will take me to get there. It should know which application or service can modify the reservation. It should know when it needs to ask me a question and when it doesn't.
Maybe an airline app performs part of that task. Maybe Calendar contributes another piece. Maybe Maps does some work. Maybe Messages contains the dinner plans. Maybe a small model on the device handles part of the request while a larger model somewhere else handles another part.
I genuinely don't care.
That's the point.
I don't want to manage the machinery.
I want the machinery to understand the outcome.
Apple's work around App Intents is interesting to me for exactly that reason. An application stops being only a destination with buttons and screens and starts becoming a collection of capabilities that can be made available to the larger system. The airline application still exists, but perhaps I don't have to visit it. The useful thing it knows how to do becomes callable by something operating at a higher level.
At that point, the architecture of personal computing begins to change in a subtle but profound way.
The old arrangement is that the human chooses the software and then operates it.
The emerging arrangement is that the human expresses intent and an intelligence layer figures out which software, tools, services and models are required to fulfill it.
That's the computer I've been waiting for.
And weirdly, it brings me all the way back to those little boards on my workbench.
When I built FireFly, I was trying to move useful functionality closer to the person. When researchers put electrical impedance sensing on the back of a watch, they're moving input closer to the body. When vision systems understand what we're looking at, they're reducing how much context we have to explain. When voice models become good enough to understand messy natural speech, they're removing another layer of syntax. When applications expose their capabilities to agents, they're removing another layer of navigation.
These don't feel like separate technologies to me anymore.
They all remove distance.
Distance between the human and the sensor.
Distance between intention and instruction.
Distance between instruction and capability.
Distance between capability and action.
And once you begin looking at the problem that way, the obsession with which company currently has the highest-scoring model starts feeling slightly less important.
The models matter enormously, of course. I live in this stuff. I'm not suggesting otherwise. But model capabilities are advancing so quickly that today's miraculous advantage has an uncomfortable tendency to become tomorrow's commodity API.
The more interesting question may be who controls the layer above them.
Who has the context?
Who has identity?
Who understands permissions?
Who has access to the applications?
Who has the sensors?
Who owns the interface through which the human expresses intent?
Apple has an unusual answer to those questions because Apple has spent decades assembling a computer around the person.
There's a phone in your pocket, a watch touching your skin, earbuds in your ears, a computer on your desk, perhaps a tablet nearby, all attached to an operating system and identity framework that already mediates an enormous part of your digital life.
That's potentially a remarkable foundation for an intelligence layer.
And if Apple gets this right, I'm not sure the user should even know which model is answering a particular request.
Maybe it's Apple's...Maybe it's Gemini...Maybe it's running locally...Maybe it goes through Private Cloud Compute... Maybe five years from now some specialized model none of us have heard of today handles one narrow portion of the task extraordinarily well.
Why should I have to know?
I don't choose the network protocol when I open a webpage. I don't choose the codec when I play a song. Most people don't know or care which CPU instruction made their photo appear on the screen.
Apple's entire history is built around taking complicated systems and placing a nicer abstraction over them.
Artificial intelligence should eventually receive the same treatment.
That's where this gets more interesting than an Apple product strategy discussion, because an operating system that routes intelligence is doing something much more consequential than deciding which processor executes some code.
It potentially decides which intelligence receives the request, what information accompanies that request, what applications that intelligence can touch, what personal context it can access, what actions it's authorized to perform, which services are available in a particular jurisdiction, and which answer eventually reaches the person.
I recently came across the phrase "custody over intelligence," and I've been chewing on it ever since.
There's something slightly unsettling and enormously important about that idea.
The operating system of the future may not merely manage the machine. It may mediate the relationship between the human and machine intelligence itself.
That's a hell of a responsibility.
But before I wander too far into the geopolitical implications of Siri, I keep coming back to the much simpler thing that makes me geek out about all of this.
I want the computer to disappear.
Not literally. I like computers way too much for that.
I mean I want the act of using one to disappear.
I don't want to think about which application contains the thing I need. I don't want to think about which model should answer me. I don't want to manually move context from one system to another. I don't want to learn another menu because somebody redesigned the interface.
I want to look at something, speak, gesture, move, or simply express an intention, and have the machine understand enough about the situation to help.
When it doesn't know, it should ask.
When it does know, it should act.
The rest should increasingly become plumbing.
Maybe that's why projects like EITWatch hit me the way they do. I see eight little electrodes on the back of a smartwatch and immediately start imagining where the idea goes twenty steps later. It's the same reason I spent too much time cramming LEDs and batteries and radios into tiny wearable enclosures years ago. It's the same reason I keep talking about apps turning into capabilities and voice becoming an intent aperture.
Apparently I've been working on the same problem all along without particularly realizing it.
Sometimes I approached it with a soldering iron.
Sometimes with software.
Now we're approaching it with artificial intelligence.
And for the first time in a while, Apple looks like it may be assembling enough of the pieces to make that particular computer real.
I don't know if they will.
They're Apple.
They could unveil the future next year or spend eighteen months moving a button.
But I know what I'm hoping for.
A computer where the models disappear underneath the intelligence, the applications disappear underneath their capabilities, the sensors quietly understand the environment, and the interface stops demanding so much attention from the person using it.
A computer that spends less time asking us how to operate it and more time understanding what we're trying to accomplish.
Apparently I've been trying to build little pieces of that thing for years.
I'd really like somebody to finish it.
Sources & Further Reading
Apple Intelligence and Siri AI — Apple Developer
Apple's developer documentation describing how Apple Intelligence combines language models with app actions, content, App Intents, App Entities, schemas, Spotlight, and Siri AI.
App Intents — Apple Developer
Apple's framework for exposing an application's actions and data to Apple Intelligence and system experiences such as Siri, Spotlight, Shortcuts, and widgets.
App Intents: Custom Actions — Apple Developer
Technical overview of how an application exposes individual capabilities to the operating system rather than requiring every action to originate inside the app interface.
EITWatch Hardware — Xuanyou Liu, Novel Alam and Karan Ahuja / Northwestern University
Open-source hardware accompanying the EITWatch research project. The prototype places eight electrodes against the wrist and uses electrical impedance tomography to recognize hand gestures from changes in tissue.
Apple Sued by Shareholders for Allegedly Overstating AI Progress — Reuters, June 20, 2025
Reporting on the shareholder lawsuit alleging that Apple overstated its progress toward delivering the advanced Siri capabilities demonstrated with Apple Intelligence.
Apple Agrees to $250 Million Settlement Over Advertised Siri AI Features — Associated Press, 2026
Coverage of Apple's agreement to settle a consumer class-action case involving Siri and Apple Intelligence capabilities advertised for supported iPhones but not available as originally expected.
Earlier Stops Down My Own Rabbit Hole
Apple's Vision for the Future of Artificial Intelligence for the Rest of Us
My early reaction to Apple Intelligence and the idea that Apple's real opportunity was not another standalone chatbot, but intelligence embedded throughout the operating system and personal computing environment.
Apple Got Sued for Failing at AI. Then They Showed Up With This.
My look at FastVLM and why efficient multimodal models made me think more seriously about intelligence moving into sensors, devices, wearables, and the physical environment.
I Said Apps Were Dead. Apple Just Proved It by Building What Comes Next.
The argument that applications increasingly become callable capabilities underneath an intelligence layer rather than destinations humans must manually navigate.
The Sleeping Giant Wakes: What WWDC Will Actually Tell Us About Apple's AI Play
My attempt to think through Apple's potential advantage not as ownership of the best model, but as ownership of the hardware, operating system, context, permissions, application ecosystem, and routing layer surrounding those models.
richwashburn.com/post/the-sleeping-giant-wakes-what-wwdc-will-actually-tell-us-about-apple-s-ai-play
Jony, Call Me
The broader rabbit hole: what happens when intelligence becomes part of the environment and the computer increasingly adapts itself to the human rather than requiring the human to adapt to the interface.
FireFly 3.0: Refresh — The Apple Product That Wasn't
A return to one of my older wearable hardware projects and the continuing exercise in putting useful capability into very small physical devices.
FireFly Project
The broader FireFly project page and some of the hardware experimentation behind the wearable platform.
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.
Comments