Anthropic cut Opus prices by 67% and claimed benchmark leadership. But confused usage limits, a Microsoft paradox, and user expectations of post-launch degradation reveal competitive pressure reshaping AI economics.
OpenAI's new shopping bot achieves 64% accuracy while the company bleeds $5 billion yearly. As Amazon blocks access and processing takes 5 minutes, the tool reveals a deeper crisis: OpenAI needs revenue desperately but can't build it.
OpenAI paid $6.5 billion for Jony Ive's hardware startup but can't explain what they're building. The former Apple designer now advocates for lickable AI devices while studying the history of pockets.
OpenAI paid $6.5 billion for Jony Ive's hardware startup but can't explain what they're building. The former Apple designer now advocates for lickable AI devices while studying the history of pockets.
Sam Altman wants you to lick his new product. Or maybe take a bite out of it. The OpenAI CEO can't quite remember which metaphor his design partner Jony Ive used, but he's certain the urge to consume their mysterious AI device represents the pinnacle of industrial design.
Last Thursday at Emerson Collective's Demo Day, Altman and Ive spent 45 minutes discussing their hardware collaboration without revealing what it actually does. They have prototypes. They've raised $6.5 billion. They promise it will ship within two years. Yet when pressed for specifics, they offered philosophy seminars about pockets and ceramic insulators.
The performance was vintage Silicon Valley theater. Two titans of technology, seated before Steve Jobs' widow, waxing poetic about research books on "the history of shapes" while avoiding the elementary question of function. Ive, who once made Apple products feel inevitable, now traffics in deliberate obscurity. Altman built ChatGPT into a household name. Transparency was the whole pitch. Now? Mystery sells better.
The vagueness tells the real story though. OpenAI needs hardware. Badly. Every tech giant runs the same AI models on the same glass slabs. What's the differentiation? Get the iPhone's designer. Have him kill the screen.
The Breakdown
• OpenAI acquired Ive's startup IO for $6.5B in equity despite having no shipped product or clear functionality
• The screenless device promises to ship within two years but design process prioritized philosophy over use cases
• Acquisition reveals OpenAI's platform dependency crisis: every ChatGPT query enriches Apple or Google's ecosystem
• Tang Tan's hiring suggests possible custom silicon development despite vague "lickable" design rhetoric
Ive's enthusiasm for the invention of the pocket tells you everything about this project's conceptual foundation. "Did you know someone invented a pocket?" he marveled to the audience, describing his amazement that pants once existed without them. This is the design philosophy guiding a $6.5 billion venture. Not solving contemporary problems but rediscovering prehistoric ones.
The research process, as described, involved months of academic exploration into "the nature of relationships of humans to nature, humans to animals, humans to humans." OpenAI engineers studied the nature of intelligence, the history of tools, values and human rights. They produced "10 books, each this thick" on topics ranging from camera design to relationship evolution. Not one of these exercises, apparently, involved asking what people actually want from an AI device that lacks a screen.
Consider the absurdist progression. First, Ive contacted Altman with what he admits wasn't a product idea but "a tentative thesis" about "the nature of objects and our interface." Months of philosophical wandering followed. Then suddenly, prototypes. How'd they get from abstract theorizing to physical hardware? Nobody knows. It's watching someone spend two years reading about wheels. Then they announce they've built one. Won't show it. Won't explain it either.
The "lick test." Jesus. Great design makes you want to put the object in your mouth. Not use it. Not understand it. Consume it. The metric for a $6.5 billion AI platform is whether it triggers infantile oral fixation.
Altman's critique of modern devices deserves scrutiny. Current technology, he argues, resembles "walking through Times Square" with constant notifications, dopamine hits, and attention hijacking. The iPhone, which he simultaneously calls "the crowning achievement of consumer products," has created an "unsettling" experience of perpetual distraction.
Fair enough. But Altman's solution isn't fixing notification systems or building better attention management tools. It's creating an entirely new device category that somehow knows "everything you've ever thought about, read, said" while remaining so unobtrusive you barely notice it exists. A device that achieves the spiritual experience of "sitting in the most beautiful cabin by a lake."
This is magical thinking dressed as product development. Screens aren't the problem. Business models are. Every ping means money. Instagram doesn't bombard you with updates because phones have displays. It does so because engagement drives ad sales. Unless OpenAI plans to revolutionize digital economics alongside hardware design, their screenless device will simply find new ways to interrupt you. Audio notifications. Haptic patterns. Perhaps it will literally require licking to confirm you're paying attention.
The rhetoric also ignores OpenAI's own contribution to the attention economy. ChatGPT's interface might be minimal, but the product itself is engineered for engagement. The company that created an AI so compelling people spend hours crafting prompts now claims to champion digital minimalism. It's like a casino owner opening a meditation retreat, technically possible but philosophically incoherent.
OpenAI paid $6.5 billion in equity for IO, Ive's hardware startup. Context matters here. Microsoft bought GitHub for $7.5 billion. Amazon got Whole Foods for $13.7 billion. Thousands of employees, millions of users, actual revenue. IO? Jony Ive and some designers.
The valuation makes sense only if you accept that OpenAI faces an existential platform problem. The company's entire business runs through other people's hardware, iPhones, Android devices, Windows PCs. Every ChatGPT query enriches Apple or Google's ecosystem. Every enterprise deployment strengthens Microsoft's grip on corporate computing. OpenAI needs its own hardware not because screens are evil but because platform dependency is fatal.
Yet the device they're building sounds deliberately marginal. No screen suggests limited functionality. "Simple and beautiful and playful" suggests lifestyle accessory, not computing platform. The comparisons to Humane's failed AI Pin are inevitable and probably accurate. That product also promised to liberate users from screen addiction through ambient AI. It shipped, flopped, and now serves as a $700 cautionary tale about solving problems that don't exist.
The research process provides another tell. Successful hardware emerges from specific use cases, not abstract philosophizing. The iPhone solved the problem of carrying multiple devices. The Apple Watch solved fitness tracking and phone-free communication. AirPods solved wireless audio. Each addressed a clear user need with superior execution.
IO did it backwards. Philosophy first. Aesthetics second. Function last, if at all. Like architects spending years contemplating "dwelling" before asking if anyone needs shelter. Then building something beautiful that leaks. The approach might produce art. It won't produce products.
The clearest signal about IO's actual direction comes not from Altman or Ive but from Tang Tan's involvement. OpenAI hired Apple's former VP of Product Design as chief hardware officer, a man who spent 25 years shipping actual products under John Ternus and Ive himself. You don't poach Apple's hardware brass to build meditation stones.
Tan's expertise suggests OpenAI is building something more ambitious than a screenless notification device. The company needs compute at the edge, not just another interface. Consider the constraints: Every ChatGPT query requires massive data center resources. Every image generation burns expensive GPU cycles. Every voice interaction depends on internet connectivity. The economics don't scale.
A device with local AI processing would solve these problems. But that's a semiconductor challenge, not a design philosophy. It requires chips, memory, thermal management, boring engineering problems that resist aesthetic solutions. No wonder Altman and Ive prefer discussing pockets and ceramics.
The timeline provides another clue. "Less than two years" means late 2026 at the earliest. That's an eternity in AI development but standard for custom silicon. Apple spent three years developing the M1 chip. Google took four years to ship Tensor. If OpenAI is building custom AI processors, the schedule makes sense. If they're just removing screens from existing hardware, two years is inexcusable.
Perhaps the most revealing moment came when Laurene Powell Jobs asked what people could expect to see. Ive responded with unusual certainty: "I think even less than that," referring to the two-year timeline. Then immediately backtracked: "No, five years." Then agreed to two years again.
This isn't playful misdirection. It's genuine uncertainty. They have prototypes but no ship date. They have philosophy but no product definition. They have $6.5 billion in equity but no business model. The entire project feels less like product development than performance art, a elaborate demonstration that even AI companies need mystery to maintain their valuations.
The comparison to Florence in the Renaissance, which Powell Jobs raised, is unintentionally apt. Florence produced extraordinary art because wealthy patrons funded artists without demanding immediate returns. The Medici didn't ask Michelangelo for quarterly metrics. They subsidized genius and hoped for masterpieces.
OpenAI is betting $6.5 billion on the same model. They're funding Ive not to solve problems but to create desire. Not to build products but to manifest inevitability. The device, whatever it becomes, matters less than the narrative. Silicon Valley's last true magician is working on something mysterious. That story alone justifies the valuation.
The IO acquisition reveals OpenAI's platform dependency as its greatest strategic vulnerability. Without proprietary hardware, the company remains a tenant in Apple and Google's ecosystems, bleeding margin to landlords who control distribution.
Investors betting on AI hardware should recognize the pattern: Every generational computing shift produces failed intermediate devices before the winning form factor emerges. OpenAI's mystery box will likely join the Newton, Google Glass, and Humane Pin in the museum of premature hardware.
The two-year timeline suggests OpenAI knows ChatGPT's competitive moat is evaporating. When every tech giant offers comparable language models, brand and distribution determine winners. Ive provides brand. Mystery provides time. Neither provides sustainable differentiation.
Q: When did OpenAI and Jony Ive actually start working together?
A: First discussions were reported in 2022. Ive confirmed the collaboration in November 2023. OpenAI officially acquired Ive's startup IO for $6.5 billion in May 2024. They now have working prototypes and expect to ship by late 2026 or early 2027.
Q: Who is Tang Tan and why does his hiring matter?
A: Tang Tan is Apple's former VP of Product Design with 25 years experience shipping actual products. OpenAI hired him as chief hardware officer. His expertise in semiconductor development and product manufacturing suggests OpenAI is building custom AI chips, not just removing screens from existing hardware.
Q: What's the Humane AI Pin and why did it fail?
A: The Humane AI Pin was a $700 screenless, wearable AI device that promised to replace smartphones through voice commands and laser projection. It launched in 2024 but failed due to poor battery life, limited functionality, and slow response times. It's now considered a cautionary tale about solving non-existent problems.
Q: What does "platform dependency" actually cost OpenAI?
A: Every ChatGPT query on iPhone enriches Apple's ecosystem. OpenAI pays for compute while Apple and Google control distribution and user relationships. This dependency limits OpenAI's pricing power, data collection, and direct user access. Building proprietary hardware would let OpenAI capture full value from its AI services.
Q: What exactly did they research before building prototypes?
A: They spent months studying abstract concepts including human-animal relationships, the nature of intelligence, and the history of tools. They produced 10 thick books covering topics from camera design to relationship evolution. Notably absent: any research into what users actually want from a screenless AI device.
Robert Brown
Maria Garcia
Marcus Schuler
Anthropic cut Opus prices by 67% and claimed benchmark leadership. But confused usage limits, a Microsoft paradox, and user expectations of post-launch degradation reveal competitive pressure reshaping AI economics.
OpenAI's new shopping bot achieves 64% accuracy while the company bleeds $5 billion yearly. As Amazon blocks access and processing takes 5 minutes, the tool reveals a deeper crisis: OpenAI needs revenue desperately but can't build it.
California granted Waymo permission to operate across 47,493 square miles—but the company runs just 2,500 vehicles in five cities. The gap between regulatory approval and actual deployment reveals unit economics Waymo won't disclose and local opposition the state ignores.
Google kept AI Mode ad-free for a year to compete with ChatGPT. Last week, ads appeared. The retreat reveals more than monetization strategy—it shows Google's $175B advertising business makes competing structurally impossible.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
