What Actually Kills a Device

Here is the specific version of the claim, stated plainly so it can be tested rather than gestured at: the phone gets replaced by a lens and a pair of earbuds. A smart contact lens that puts images directly on the retina, paired with earbuds that both produce sound in the ear canal and pick up what you say — together doing away with the need for a phone, a television, or any separate screen-and-speaker object you have to carry, wake, and hold. The stronger form of the prediction, meant as more than a rhetorical flourish, is that this reaches the same destination as an invasive brain-computer interface without being invasive. This specific version is the one that actually deserves testing.

The honest way to test a claim like this is not to defend it. It is to ask what the historical record shows about how dominant technologies actually end, and then, separately, to look hard at what is verifiably true right now about lenses-on-eyes and machines-that-read-minds — and hold the prediction against both without flinching if it comes up short. That is the method this site exists to apply to Roman fire myths and asteroid mining headlines alike, and a claim about the future is not exempt from it just because it concerns technology still being built.

The case for the prediction

Start with the strongest version of the argument, because it deserves to be made properly before it gets tested. The phone's dominance looks, from one angle, less like a durable standard and more like a transitional shape — an object that happened to be the best available container for a dozen unrelated capabilities at the moment those capabilities matured, not because a slab of glass in a pocket is an especially good permanent home for a camera, a wallet, a map, a key, a screen, a microphone, and an always-on connection to everything else. Interfaces have changed shape before, and quickly, when the underlying capability outgrew its container: the desktop computer did not evolve gradually into the phone, it was overtaken by something built on different assumptions once the pieces — cheap sensors, mobile networks, batteries good enough to matter — arrived together. A lens that projects directly onto the retina and an earpiece that both speaks and listens remove a step the phone has always required: you still have to take it out, wake the screen, find the app, and look at it. An interface with no object to take out at all removes the friction that every phone interaction still pays. That is the honest core of the prediction.

Until recently, that half of the claim — the lens half, specifically — was easy to wave off as a good metaphor with no working hardware behind it. That is no longer true, and it is worth being precise about what changed, because the change is the actual news here. At Mobile World Congress in 2026, a Dubai-based company called XPANCEO showed five working smart-contact-lens prototypes on the floor, off the back of a $250 million Series A that valued the company at $1.35 billion, announced around GITEX Global in late 2025.¹² Trade coverage tracking the company since an earlier MWC appearance had pointed toward 2026; the more recent reporting around the funding round points to early 2027 for an actual product.¹³ What makes this worth taking seriously, rather than filing it next to a decade of AR-glasses vaporware, is the architecture, not the demo. XPANCEO's lens is a display and nothing else. A small companion device worn over the ear does the processing and beams power to the lens wirelessly, because a battery sitting on an eyeball generates heat and bulk you cannot put on a human cornea. That is not a coincidental engineering choice — it is, point for point, the split described above: the lens shows the image, the earpiece does the work. A prediction made about a lens-plus-earbud system and a real prototype built as a lens-plus-earpiece system arriving at the same division of labor independently is a striking convergence, and it deserves to be named as one rather than folded quietly into the rest of the argument.

The company has also made progress on two hard supporting problems. A thin enough power source has been a standing objection to eye-worn electronics for years; XPANCEO's work with the battery maker ITEN on solid-state ceramic microbatteries, in mass production since 2025, is aimed directly at that, and the two companies have jointly announced a proof-of-concept pairing the battery with the lens.¹⁴ Separately, a corneal implant developed with the ophthalmic firm Intra-Ker has been tested in a human donor eye — an early step, but a real one, toward a platform that survives contact with an actual eye rather than a bench mockup. None of this makes the 2027 timetable true. It means the architecture underneath the prediction is no longer speculative the way it was two or three years ago, and that distinction matters for how the rest of this essay should be read.

What the record actually shows: dominant standards get demoted, not deleted

The clearest historical case sits in this site's own back catalog. Base-60 counting was chosen by scribes in Mesopotamia for reasons that are still argued over — divisibility is the leading explanation, since sixty splits cleanly by two, three, four, five, six, ten, twelve, fifteen, twenty, and thirty, more ways than almost any small number that size — and it migrated from cuneiform tablets into Hellenistic astronomy, into Ptolemy's Almagest, into medieval and Islamic timekeeping, and out the other side into every clock face and compass rose in use today.¹ The civilization that picked it has been gone for roughly two thousand years. The choice has not gone anywhere. Another essay on this site covers this at length: the number system should have died with the people who invented it, and instead it became infrastructure nobody thinks to question.²

That is not an isolated curiosity; it is closer to the normal case. Path-dependence economics has a name for the mechanism, and a canonical case study. Paul David's 1985 paper on the QWERTY keyboard layout argued that a standard adopted early, for reasons that may have had nothing to do with long-run merit, can lock in through the accumulation of complementary investment — training, tooling, habit — long after a technically superior alternative exists, because the cost of coordinating everyone's switch at once exceeds what any individual is willing to pay to move first.³ Brian Arthur's work on increasing returns generalizes the mechanism beyond typewriters: once a technology gets ahead on adoption, each additional adopter makes it more attractive to the next one, and the gap widens rather than closes, regardless of whether the technology in front is the best one on technical grounds alone.⁴ David Edgerton's The Shock of the Old makes the broader argument directly, against a century of futurist writing that keeps predicting the imminent end of old technologies on the strength of new ones' arrival: the twentieth century's actual pattern, technology by technology, is persistence — old techniques, old tools, and old formats staying in heavy use decades after their replacements were declared inevitable, because "inevitable" describes the technical possibility of replacement, not the economic and social cost of actually carrying it out.⁵ None of these sources traffics in a specific number for how long lock-in lasts. What they agree on is the shape: dominant standards do not get deleted by a better option arriving. They get pushed into legacy status while the ecosystem built on top of them keeps them alive far longer than their technical merits alone would justify.

Lock-in scales with how much has been built on top of a standard, not with how good the standard is.

That is the load-bearing sentence in this essay, and it is the one the lens-and-earbud prediction has to answer to. The phone is not a standard with a modest amount built on top of it. It is the standard: the payment rails, the two-factor authentication flows, the identity verification systems, the mapping APIs, the messaging protocols, the accessibility conventions, and the sheer decade-plus of muscle memory in several billion pairs of hands are all built assuming a pocketable screen with a camera and a SIM sits at the center. That is a great deal of complementary investment to write off inside a few years, and the record says complementary investment is exactly the thing that keeps an incumbent standing well past the point its replacement is technically ready.

But technology can genuinely be lost — and the reason why matters

It would be dishonest to stop there, because the record does not only say "standards persist." It also contains a real counter-case, and this site has already made it once: the Antikythera mechanism. A geared bronze device, built somewhere around the second or first century BCE, mechanized the same lunar and eclipse cycles the Babylonian astronomical diaries had spent centuries recording by hand, turning a table you had to read into a machine you could turn a crank on.⁶ Nothing of comparable mechanical sophistication survives from the thousand years that followed it. The capability existed. It worked — the gear-cutting precision and the astronomical accuracy have been confirmed by modern tomographic study of the surviving fragments.⁷ And it did not propagate. Capability, it turns out, is not automatically cumulative.⁸ The mechanism did not die because a better device replaced it; nothing replaced it for a millennium. It died because the narrow tradition of astronomers, patrons, and workshops capable of designing and cutting gears at that precision did not survive the political and economic disruption of the late Hellenistic world.⁹ The Library of Alexandria makes the identical case from a different direction: not a single dramatic fire, but a long decline tracking a withdrawal of royal patronage across centuries.¹⁰ Institutions and technical traditions die of withdrawn support far more often than they die of being outcompeted.¹¹

Smart contact lenses have already run this exact experiment once, and it went the other way. For years, the most prominent name in AR contact lenses was not XPANCEO but a company called Mojo Vision, which spent close to a decade and hundreds of millions of dollars on a lens with an embedded microLED display before announcing, in early 2023, that it was ending the lens program entirely and retreating to licensing the display components it had built along the way.¹⁵ The lens did not fail because the physics stopped working. It failed because the capital ran out before the product did, and the company that had spent nearly a decade building the narrow chain of expertise required to make it — the display engineering, the power management, the ophthalmic partnerships — could not keep that chain funded long enough to reach a shippable product. That is the Antikythera pattern exactly: a real, working capability, dependent on a narrow, patronage-shaped chain of support, and the chain broke. Nobody outcompeted Mojo Vision's lens. Its investors simply stopped paying for the people who could keep building it.

That is the discipline this essay owes the prediction: XPANCEO's five prototypes are not evidence that a phone-replacing lens is inevitable on any particular clock. They are evidence that the physics and the engineering are more tractable than they looked three years ago, sitting next to a very recent, very direct demonstration that the same category of device can be killed by a funding decision with nothing to do with whether it worked. A trade-show prototype and a shipping consumer product that actually replaces the phone are separated by manufacturing at consumer volume, by regulatory clearance for what would be a Class III device worn on the eye, and by cost — none of which move on a five-year clock just because a demo worked on a conference floor. Put the two halves of the record together and the pattern is more precise than either "standards last forever" or "anything can vanish": persistence is the default when a standard has broad, distributed support built on top of it, because that support is expensive and slow to replace all at once. Loss happens when a capability depends on a narrow, fragile chain of support — a specific guild, a specific patron, a specific round of venture capital — and that chain breaks. The phone, whatever else is true of it, is the first kind of thing. The lens, right now, is still closer to the second.

The clearest modern case: a working device, discontinued

Everything so far has been either safely ancient or still speculative. One case skips both hedges: a device that worked, in living people, within the last decade, and stopped not because anything better replaced it but because the company paying to keep it running decided it no longer could. It is the clearest evidence this essay has for its own central claim.

Second Sight Medical Products built the Argus II, a retinal prosthesis: a camera mounted on a pair of glasses sends a signal to a grid of electrodes implanted at the retina, restoring a limited form of sight to people blinded by retinal degeneration. Over 350 people worldwide received one.¹⁶ Say plainly what that restoration amounted to, because overselling it would cheapen the case: signal passed through on the order of sixty electrodes, against roughly a million fibers in an intact optic nerve.¹⁸ It let users make out a doorway, a curb, the edge of high-contrast movement — not read a page or recognize a face. A real restoration of function, and a narrow one. Both are true, and the argument below needs both to stay true.

In 2019, Second Sight discontinued the retinal implant line, telling patients in writing that the technology was being phased out for financial reasons. By 2020 it had stopped technical support altogether — for the Argus, the Argus II, and its cortical implant, Orion, alike — no repairs, no upgrades, no replacements.¹⁷ People with a functioning device already in their bodies were left facing the prospect of watching restored sight go dark, unit by unit, with nowhere to take it when it failed.¹⁶ The company nearly went out of business that year and spent 2022 pursuing a merger to survive. In 2023 the technology was taken over by a successor, Cortigent, which continues developing it — a genuine qualification; this is not a story with no survivors.¹⁸

Nothing in that sequence points to obsolescence. No rival prosthesis arrived and outperformed it. It stopped because the balance sheet failed, in a company with no realistic second customer if the first one walked away — the Antikythera mechanism and the Library of Alexandria exactly, minus every hedge those two need: no missing fragments, no scholarly argument over "decline," no two thousand years of distance to make it abstract. This happened within the last decade, to people still alive to notice their sight going dark, and the only cause on record is that someone stopped paying for the work. Technologies die of withdrawn support, not competition, and this is that claim with a pulse.

The design also sharpens a distinction this essay has so far only assumed. A lens on the eye, like XPANCEO's, is a display: it puts an image where a working retina can still receive it, which is why it can sit on the eye's surface and leave a companion earpiece to do the processing. Argus II solves a different problem — a retina or optic nerve that no longer works — and its answer was not to put the camera on the eye but to take it off entirely, mounting it on glasses and injecting the signal downstream, past the part of the visual system that had already failed. A consumer lens and a medical bypass are two different products for two different populations, one with a working retina and one without, and conflating them is what makes a phone-replacing lens sound like science fiction instead of the narrower claim it actually is.

The better question: not "what's next" but "what would have to give way"

This reframing is the actual payload of testing the prediction against the record, and it is worth stating plainly: the wrong question to ask about the smartphone is "what replaces it, and when." That is a futurist's question, and futurists have a poor track record precisely because the "what's next" framing skips past the part of the story that actually determines the timeline — how much has to unwind first. The right question, on the evidence of base-60 and QWERTY and the Antikythera mechanism alike, is "what specifically would have to collapse, or unwind, for this standard to actually go away."

Run that question on the phone honestly, and the first thing it surfaces is that "the phone" is not one thing to begin with. It is a bundle: camera, payment instrument, identity token, physical key, map, screen, messaging client, authentication device, and always-on network endpoint, all currently living inside one object because that was, for about fifteen years, the most efficient container available for all of them at once. Bundles like this do not fail atomically. They unbundle component by component, on independent timelines set by whatever institution controls each piece — a bank's authentication requirements move on the bank's schedule, not a device maker's; a country's identity-document infrastructure moves on a government's schedule; a building's door lock moves on whatever cycle commercial hardware gets replaced. A lens and an earpiece can plausibly take the interaction layer away from the phone faster than they can take the wallet, the identity document, or the door key away from it, because those three things are not primarily interaction problems — they are trust and liability problems belonging to institutions with their own, much slower, replacement cycles.

The claim that a lens-and-earbud stack "reaches the same destination as an invasive brain interface without being invasive" deserves to be taken seriously and pushed on directly, because it is where this prediction is strongest and also where it is most incomplete. The honest comparison is the invasive approach associated with Neuralink and similar implant efforts, and the instinct behind the claim is correct in one direction: the eye and the ear are already the highest-bandwidth channels for getting information into a human being, so a lens that paints images on the retina and an earpiece that puts sound in the ear canal really can deliver everything a screen and a speaker deliver, without cutting into anyone's skull. That is not a small claim, and it is not wrong. But the case for a brain interface was never mainly about that direction. It was about the return path — reading a person's intent without requiring a hand, a voice, or a working motor nerve to carry it, which matters most for people for whom eyes, ears, or hands do not work at all. A lens-and-earbud stack replaces the screen. It does not replace the interface in full, because it still needs some way for a person to tell it what they want, and that half of the problem is exactly as unsolved by a retinal display as it was by a rectangle of glass in a pocket. Voice is the obvious candidate for that return path.

The right question is not what replaces the phone. It is what would have to give way first, and how much of it there is.

That is where even the sharpened, better-evidenced version of the prediction runs into the same wall base-60 has stood behind for two millennia. Even if the interaction layer is not just plausible but already arriving in prototype form, an interaction layer taking over does not require the underlying bundle of payment rails, identity infrastructure, and physical-world integrations to unwind on the same schedule. Those pieces are exactly the kind of broadly distributed, expensively complementary infrastructure that Edgerton's record says persists past the point its replacement is technically ready, not the kind of narrow, patronage-dependent tradition that both the Antikythera mechanism and Mojo Vision's lens show can actually disappear.

The verdict, stated plainly

So here is where the evidence actually lands, stated without softening. The architecture behind the prediction is more plausible than it sounded even a few years ago. A lens as a pure display, an earpiece doing the processing and the power delivery, a solid-state microbattery aimed at the thinness problem, an implant tested in a real eye — that is a specific, checkable picture of how a phone-replacing device could actually be built, and XPANCEO's prototypes are the first hardware in this story that can be pointed at rather than a slide deck. The brain-interface comparison holds up on the input side and clarifies, rather than weakens, the claim once it is pushed on: a lens and an earbud can genuinely do what a screen and a speaker do, non-invasively, which is a real and underappreciated point in the prediction's favor.

What still does not survive contact with the record is the timetable, and now it is possible to say precisely what would have to happen for that to change. Manufacturing would have to scale from five hand-built prototypes at a trade show to a device made at the volume a phone-replacing product requires. A device that sits on the eye would have to clear the regulatory bar for a Class III medical device, which moves on a schedule set by regulators, not by funding rounds. Something would have to solve the return-path problem this essay just raised — an input channel that is not a phone, whether that is voice or something else, matched to a hardware platform that can actually run it. And the rest of the bundle — payment, identity, the camera, the door key — would have to unbundle from the phone on its own institutional timelines, which the base-60 case, the QWERTY case, and Mojo Vision's own collapse all suggest is the slow part, not the display technology. Nothing in that list is impossible. Nothing in it moves in a few years just because a prototype worked at MWC. And the same lesson cuts at the prediction from a second direction: whether a lens ships, and keeps working, depends on whether someone keeps paying for it a decade after launch — a question about institutions, not optics, and Argus II is the clearest illustration of exactly that.

The direction of travel this claim points at is more real with working prototypes than without them, because a working prototype is a different kind of evidence than a hunch. But a prediction does not get to keep its most dramatic clause just because the mild version of it is being built, and it does not get to keep it just because a demo exists either. What outlasts a shipwreck, a burned patronage system, or a cancelled product line is rarely the object that was technically best in its moment. It is the object, or the standard, that had the least left to unwind when the moment for replacing it finally arrived. The phone has a great deal left to unwind — payments, identity, the whole bundle. The lens has a great deal left to build, and a recent, direct demonstration in Mojo Vision that "great deal left to build" is exactly the condition under which a working idea can still die of nothing more than withdrawn patronage. On the record, both halves of that sentence are true at once, and the honest verdict treats them as true at once rather than picking the one that makes the better headline.