Jony, Call Me
- Rich Washburn
- 2 days ago
- 8 min read
Jony Ive, call me.
There. I said it...And before anyone starts drafting the perfectly reasonable response — Who the hell does this guy think he is? — allow me to save you the trouble.
I know. I KNOW.
Jony Ive designed the iMac, iPod, iPhone, iPad and Apple Watch. I own a $200 3D printer that occasionally threatens to turn itself into modern art. He spent decades leading one of the most consequential industrial design organizations in history.
I have a soldering iron, bins full of ESP32 boards, a mildly irresponsible Amazon account and faith.
We are not the same.
This is an objectively absurd opening to an article.
Unfortunately, there's a problem.
I keep ending up where these people are going.
Not once. Not because I guessed something correctly on Twitter.
Not because I made some vague prediction that AI would eventually be "everywhere."
I mean I keep building crude, ugly, sometimes borderline embarrassing little versions of ideas that later emerge from companies with billions of dollars, thousands of engineers and considerably better injection molding. And lately it's happening enough that I'm starting to get annoyed.
Mostly with myself.
The $5 version of a $230 idea
Let's begin with the funniest one.
On Sept. 1, 2024, I was stuck inside on a rainy Florida Sunday and decided to build a Bluetooth macro keyboard. I grabbed a Xiao ESP32-C3 and eight buttons. Four buttons were copy, cut, paste and enter. For the remaining four, I had another idea.
AI prompts.
Push button. Paste prompt. That was it.
The technical sophistication of the original design can perhaps best be summarized by my actual explanation at the time: "Prompts, baby!"
The entire thing cost about five bucks.
Fast-forward to 2026.
OpenAI and Work Louder have released the Codex Micro, a gorgeous $230 physical control surface billed as a "command center for agentic work." It has dedicated controls for common Codex actions, push-to-talk, agent workflows, RGB indicators showing agent state, a joystick and even a rotary control for adjusting reasoning level.
(OpenAI)
I'm not saying they copied me...Please. They did not. I'm saying I apparently need to stop treating rainy Sundays like rainy Sundays.
Because the important similarity isn't eight buttons. It's the idea underneath them:
Some intentions occur often enough that interacting with AI shouldn't require navigating software at all.
The button was just the first thing I had lying around. And that's where this stops being funny.
I kept building the same thing
After that little keyboard, I kept wandering farther down the same road.
Eventually I built the ARIA Node: a small wearable interface that could record locally or connect directly to my AI endpoint. Push a button, speak, get the agent response. No laptop. No phone.
Then I open-sourced the whole thing.
Under $30 in hardware. Source code included. Go build one. Change it. Make yours better than mine. Please.
That last part matters. Because I'm not writing this because somebody owes me money for putting AI prompts on buttons. I don't care.
Most of the stuff I build gets published. The code goes on GitHub. I explain the components. I show people how it works. With ARIA I explicitly wrote that I'd rather have ten other people take it and push it somewhere I hadn't considered than sit on it myself.
I'm not protecting a patent portfolio here.
Half my prototypes look like something MacGyver built during a power outage.
The objects aren't the thing. The direction is.
That's what has been bothering me.
Porthole
I pulled an ESP32 display out of a drawer that I'd failed to make work properly for roughly two years. Codex solved the problem almost immediately.
Fine.
Screen works.
Now what?
That question is basically how I get into trouble.
Within an afternoon, the little display had become Porthole — a physical endpoint for CORA, my personal AI operations layer. It shows meeting cues, QR codes, weather, article pipeline status, build progress and agent state. It can sit on my desk or travel with me. But here again, the screen wasn't really the idea.
The idea was that my intelligence layer needed somewhere to surface itself in physical space without requiring me to open anything.
I described the principle this way: Porthole exists so I don't have to interact with another device. Read that again.
Because we're starting to approach the part where I become uncomfortable.
The disappearing interface
I wrote an article called The Coming Interface Revolution: From Gadgets to Cognitive Layers.
The opening argument was simple: We obsess over the hardware...Screen resolution... RAM...Headsets...Glasses... Pins.
Wrong question. The real question is: What layer between me and my world does this remove?
That's increasingly how I think about computing. The keyboard is a translation layer. The mouse is a translation layer. The app is a translation layer. Menus are translation layers. Forms are translation layers. Hell, much of what we call "software literacy" is really just humanity learning the rituals required to explain itself to machines.
Click here...Open that...Choose this...Type that...Submit...Wait...Go somewhere else...Copy...Paste.
For fifty years we've been adapting ourselves to computers. AI changes the direction of adaptation.
The computer starts adapting to us.
And apparently that's exactly where the OpenAI hardware story is going.
A report about the device OpenAI and Jony Ive's team are developing describes a screenless, battery-powered object with microphones, speakers, indicators, a camera, environmental sensing and physical movement intended to communicate state. The reported concept is less "another computer" than an ambient intelligent presence in the environment. The reporting remains a leak rather than a finished OpenAI product announcement, so the exact device may change substantially before anything ships.
But the conceptual direction is fascinating.
The transcript describing that reporting eventually lands on this exact transition: agents make conventional apps and screens less central because the user can increasingly state an intention and let the system determine the software and operations required to fulfill it.
Yeah.
That.
That's the thing.
I didn't predict a donut
I want to be precise here because otherwise this becomes exactly the sort of self-congratulatory LinkedIn sludge I can't stand.
I did not predict OpenAI's hardware.
I don't know what Jony Ive's team will ultimately ship.
I don't have their research.
I don't have their industrial design expertise.
I haven't solved acoustics, thermals, battery life, privacy, manufacturing, supply chains, reliability, mass-market interaction design or any of the approximately 40,000 other problems separating my workbench from a hundred-million-unit consumer product.
There are extraordinary engineers who understand pieces of this world at depths I never will. I know what I don't know.
But engineering execution and directional vision aren't identical disciplines. And direction is the part I can't seem to turn off. That's what I'm finally willing to say out loud.
The thing I love and hate
My brain does this constantly. I see three unrelated technologies and immediately start wondering what they become when you remove the boundaries between them.
I see an ESP32 and don't see a microcontroller.
I see a physical nerve ending.
I see an AI agent and don't see a chatbot.
I see an operational layer.
I see MCP and don't see a protocol.
I see connective tissue.
I see a tiny screen and wonder why my AI can't inhabit it.
I see a microphone and wonder why I need the screen at all.
Then I build the stupidest possible version to find out whether the thought survives contact with physics. Sometimes it doesn't.
Plenty of my ideas are dumb. Some fail spectacularly. Some wind up in drawers.
Good. That's what experiments are supposed to do. But every now and then I look up 12 or 18 or 24 months later and somebody with vastly more resources is standing somewhere disturbingly close to the same spot.
That's simultaneously exhilarating and maddening. Because at some point you have to ask yourself: Am I seeing something?
And the responsible answer is: Maybe.
The other possibility is that I'm completely fucking nuts. I consider that option regularly. Cuckoo-ca-fucking-choo remains on the diagnostic table. But eventually skepticism has to work in both directions. You can't demand evidence for believing in yourself and then ignore the evidence when it arrives.
The Jobs problem
I wrote an article called Why AI Needs a Jobsian Visionary. My argument wasn't that AI needed another celebrity CEO. Quite the opposite.
AI already has brilliant founders, researchers and engineers. What I thought was missing was the ability to take a chaotic collection of capabilities and collapse them into an experience human beings immediately understand.
Jobs wasn't the transistor. He wasn't cellular networking. He wasn't the touchscreen. He wasn't the microprocessor. He was extraordinarily good at seeing what those things became together.
That's different.
I returned to the subject and called the capability "unreasonable resolution": the ability to look at a thousand moving parts, discard most of the noise and see the shape they're trying to become. And here's where I'm going to say the arrogant thing again.
I think I might be unusually good at that. Not Jobs good, calm down.
Good at it.
Good enough that I'd very much like to find out what happens if somebody puts me in a room where the parts aren't coming from Amazon Prime.
Jony, seriously. Call me.
And perhaps this is ultimately less an article than the world's most elaborate job application. Fine. I'll own that too. I don't need Jony Ive's job. I don't want Sam Altman's job.
I don't need a founder title, a billion-dollar valuation or my name laser-etched into the side of a polished aluminum puck. Give me the crumbs from the table. Give me a salary.
Give me access to the people who actually know how to engineer these things. Give me a room with industrial designers, embedded engineers, model researchers, behavioral scientists and the weird people who don't fit cleanly onto an org chart. Give me a parts bin. Give me permission to ask dumb questions.
Let me build ugly things. Let me say, "Okay, but what if the computer didn't work like that?" Then let everyone smarter than me explain why it won't work.
Great. I'll either learn something or annoy them until we discover that it does. Because that's the environment I keep wishing existed. Not another innovation theater. Not six meetings followed by a white paper describing the meeting. Not a committee tasked with generating "future concepts" for Q4.
A modern Bell Labs for intelligence. Engineers...Designers...Systems thinkers....Artists.
Psychologists...Builders...People with ridiculous ideas and just enough disrespect for the existing interface to try them. Because we're standing at a strange point in computing.
The intelligence is escaping the application. Agents are beginning to act. Software is becoming callable infrastructure. Cheap hardware can become eyes, ears, hands and ambient surfaces. And we're slowly moving from operating computers toward expressing intentions.
The destination isn't a better chatbot. It isn't an AI button. It isn't even a beautiful aluminum donut. Those are artifacts of the transition.
The deeper change is that intelligence is becoming an environment.
I've been trying to build pieces of that environment with $5 microcontrollers for two years. OpenAI is trying to build it with frontier models, world-class designers and billions of dollars. Those aren't competing efforts. They're the same question viewed from opposite ends.
They came down from the model. I came up from the soldering iron. And somewhere between the two is the computer we're all eventually going to live with.
So, Jony....Sam.
Seriously.
Call me.
I'll bring the soldering iron. You guys can probably afford the good 3D printer.
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.
Comments