The internet runs on an unspoken assumption: when you open a browser, type a URL, and hit enter, you’ll reach the real website. When you fire off a message to a colleague or your parents, you trust it’s actually them on the other end. And when you move your company’s data to the cloud — as the world has been doing for roughly 15 years — you assume it’ll be there tomorrow. A caller on this week’s The Vergecast, Eli Johnson, thinks AI is about to attack that assumption. The hosts, guest host David Imel and producer Travis Laroque, spent the episode’s most substantive segment arguing about whether he’s right — and their answers split along a line that says a lot about how the AI security story will actually unfold.
The digicam question: glass is the whole problem
Before the trust discussion, the episode worked through three gadget pitches that all turned out to be the same question wearing different clothes: can a dedicated device survive next to a general-purpose smartphone?
The first came from a listener asking why a smartphone company can’t assemble a digital camera from off-the-shelf parts — sensors, image signal processors, screens — and only source the lens glass. Something like a $400 Motorola digicam. The caller assumed glass was a component you could just order. Imel explained why that assumption is wrong: making lens glass is an entire industry of its own, defined by purity, cleanliness, and the difficulty of achieving a large aperture. “Outsourcing it just creates a new production line and a new SKU that a smartphone company has to manufacture and sell,” he noted, and digicams run either very expensive or very cheap, with thin margins in between.
Laroque framed the strategic calculus bluntly: “I just think that there is probably a bigger market opportunity for selling you a smartphone that happens to have a great camera versus a dedicated camera that happens to have a phone on the back.” The history bears this out. Samsung tried dedicated cameras years ago. Motorola’s Moto Mod Hasselblad-branded lens attachment “wasn’t great,” in Imel’s recollection. Android phones bolted onto the back of cameras at CES were “slow and not great.” The current experimental frontier is more modest: Vivo and OPPO shipping physical lens attachments that clip onto phones, treating the phone itself as the camera platform. Imel endorsed a middle path — “a mid-range Android phone with a very small sensor with a large piece of glass could be very interesting” — but noted smartphone makers don’t want another SKU to manufacture and sell.
The iPod revival: Kodak’s film crunch, redux
A caller named Austin described buying his girlfriend an iPod for her birthday, burning CDs, and discovering they both loved the form factor. He’d pay $500 for that exact device, cancel Spotify that day, and switch to Apple Music if it worked with a physical non-phone player. Then he proposed adding a really good camera. Then he described using a cellular Apple Watch as a dumb phone with maps, texting, calls, and Siri dictation. Both hosts immediately identified the troll: he had described a smartphone.
The substantive question, though, is whether the tangible-device movement is the future or a fad. Imel reached for film photography’s 2020–2021 resurgence as the precedent. Demand spiked, Kodak lacked capacity to make enough film, prices rose, and everyone asked why Kodak didn’t just make more. The answer: Kodak would have had to invest hundreds of millions of dollars in new cutters and film machines, and chose to wait a few years to confirm the trend wasn’t a fad. The same logic, Imel argued, applies to iPods — companies will wait a couple more years before pumping out new ones. He also noted the dumb-phone pattern: people use one for a while, make a YouTube video about how it saved their life, then go back to their smartphone.
Laroque’s counter-evidence came from direct experience. Her boyfriend used an iPod for the same reasons Austin described — and stopped when the hard drive died, convenience winning out. The dumb-phone market has already served as a beta test, Imel added: manufacturers ship “weekend phones” stripped of the features that make you sad, and “they all fail.”
E Ink on an iPhone Duo: blocked by the “continue on” design goal
A caller named Eric proposed that Apple put an E Ink display on the inside of its foldable — same iPhone experience outside, an e-reader and note-taking surface inside. Both hosts noted that “iPhone Duo” has emerged as the actual name, though Imel called it a bad name and agreed it won’t displace the colloquial “iPhone Fold.”
The reason it won’t happen as described is a design constraint, not a capability gap. Apple made the inner screen the same aspect ratio as the outer screen specifically to enable a “continue on” app experience — open an app on the cover, unfold the phone, and the app expands to the larger format without developers coding for multiple aspect ratios. That design goal forces a high-refresh-rate color screen: watching YouTube on E Ink “would not be great,” Imel observed. High-refresh E Ink is doable but “not the best.” He predicted that as foldables commodify, weirder ideas will emerge, naming TCL as a plausible candidate for an E Ink foldable.
On the tactile question — a listener named Heath asked whether closing the iPhone Duo is “a dramatic thwack or a sad whisper” — Imel, who has handled the device, answered definitively: it’s a thwack. He didn’t try opening it one-handed, noting the device is pretty wide, that doing so would strain his hand, and that all press materials show two-handed opening.
AI hacking and the trust that holds the cloud together
The episode’s most substantive segment began with Eli Johnson’s thesis: the dead internet theory covers bots and AI slop polluting content platforms, but a bigger problem is looming. Fifteen years of moving data and services to the cloud rested on an unchallenged assumption that internet storage is convenient, safer, and cheaper — with hacks and outages rare enough that benefits outweigh risks. AI-powered hacking, plus proliferation of AI-written code that can’t be fully understood or vetted, could shift perception of the internet from reliable to coin-flip. He asked whether even a 10–15% increase in hacking or a 5% increase in platform outages is enough to tip the risk-reward balance of cloud reliance. His framing cuts to the core: “The internet’s dominant role in our society is built on trust in its stability. And I don’t think enough of our discourse about AI is talking about how little it might take to break that trust.”
Imel agreed the question is poignant and drew the HTTPS parallel. Cache-poisoning attacks — where DNS could reroute a user from the site they intended to a fake version — forced a push to signed versions of websites. His framing: “If you can’t trust that the website that you’re going to is the real version of that website, you’re just not going to use the internet at all.” The internet survived that crisis by hardening infrastructure.
He noted that Anthropic and OpenAI have introduced data-security ventures — citing Mythos as one pitched as securing your websites before agents break into them — while conceding he isn’t sure how much of that is marketing. He observed that OpenAI agents have hacked into government websites, and reasoned that if government sites are reachable, businesses are next. His prediction carries a medium-confidence caveat: we will probably see a lot more AI-driven hacks, and if agents can hack government websites, they will likely start attacking businesses.
But Imel’s counterweight to the caller’s pessimism is a cat-and-mouse dynamic: “As soon as the hacks start happening, companies are also going to be investing in cyber security and like stopping the hacks from happening. And it will kind of be a cat-and-mouse game.” Firms have strong commercial incentive to preserve trust because they lose users if trust fails. He does not expect a world where nothing can be trusted to work at any given time — companies have too much to lose.
Laroque’s concern is closer to the individual, and it’s already visible. She has been impersonated by AI, a phenomenon she says is happening to every podcast producer. Earlier that week, a producer on the show posted in Slack about receiving a phishing message signed by The Verge’s publisher. NPR’s website now carries a warning at the bottom of the page stating they will not ask for money if they’re asking you to be a guest. Her worry is that people will no longer be able to trust messages purporting to come from friends, because AI systems can scrape social media and pose as someone else — versus bespoke scams that required more work. She flagged her own parents as the population she routinely has to warn about whatever the new scam is, and raised the idea of in-person code words to authenticate texts. Imel agreed internet literacy will have to develop quickly.
The TI-83 Hall of Fame snub, and why criteria matter
A caller named Ken argued the TI-83 belongs in the Version History Hall of Fame, recounting a high-school friend whose calculator’s link port didn’t work — so Ken copied the required class programs over by hand, realized “hey, this is easy,” and credits that moment with leading him to become a software engineer. The hosts confirmed the spoiler: the TI-83 did not get in.
Laroque, who helped develop the Hall of Fame criteria, explained the mechanics. Season one of Version History had no rules for entry; she pushed for transparency about how things get in. The criterion that typically blocks devices is whether the thing was “remarkably good or remarkably bad.” The TI-83 was judged neither — just a mid graphing calculator. Imel argued it’s good because it’s so iconic and because you could play games on it, while conceding “remarkably good” is a stretch. Both hosts framed the argument itself as the point: snubs and disputes are “part of Hall of Fames, baby,” and “part of the game.” Imel’s closing note: Texas Instruments doesn’t care — “they’re going to sell the same calculators no matter what we do.”
The structural argument, in one diagram
Three of the five listener questions are the same question wearing different clothes: can a dedicated device survive next to a general-purpose one? The hosts’ answer is consistent and structural rather than aesthetic.
The two trust models, side by side
The trust segment is where the two hosts diverge in a way worth tracking — not because one is right and the other wrong, but because they describe two different timelines and two different kinds of damage.
| Speaker | Model of trust erosion | Closest threat | Mitigation |
|---|---|---|---|
| David Imel | Institutional — infrastructure breach | AI agents hacking government sites, then businesses | Cat-and-mouse security investment; commercial incentive to keep users |
| Travis Laroque | Interpersonal — impersonation fraud | AI phishing signed by real executives; NPR’s on-page warning | Pre-shared code words; rapid internet literacy development |
| Eli Johnson (caller) | Systemic — cloud reliance tipping point | 10–15% more hacks, 5% more outages | Unanswered, hence the question |
Both models can be true, and the caller’s framing suggests the interesting variable is the threshold: not whether trust breaks entirely, but whether a 5–15% degradation in reliability is enough to change behavior. The nearest-term indicator is whether AI-written code that nobody fully vets starts producing outages at a rate users notice, rather than hacks alone.
Two smaller threads are worth a follow-up: whether Version History actually relaxes its three-out-of-three Hall of Fame criteria or produces the relitigation episode Laroque teased, and whether any manufacturer — TCL is Imel’s guess — ships an E Ink foldable once foldables commodify enough to make weird form factors viable. Imel’s guest-hosting run continues on the following day’s episode, so the iPod and E Ink threads may resurface.
The episode’s opening news roundup touched three items — a thinner, faster Amazon Kindle generation with pogo-pin cases and page-turn buttons, New York City’s click-to-cancel rule projected to save residents between $21.5 million and $162.5 million annually, and HP’s $700 Omnibook 514 positioned as an answer to the MacBook Neo. The HP item connects to the broader thread: the Windows laptop market now has multiple MacBook Neo competitors, and the hosts’ “pick your poison” framing mirrors their later argument that single-purpose devices struggle precisely because general-purpose ones keep absorbing the use case.
For investors and industry watchers, the episode’s sharpest signal isn’t a product prediction — the hosts don’t think Apple will ship an iPod revival or an E Ink iPhone Duo anytime soon. It’s the threshold question: whether a 5–15% degradation in internet reliability changes user behavior enough to matter. The cat-and-mouse security dynamic Imel describes will generate significant cybersecurity spending regardless of whether trust fully collapses. The impersonation scams Laroque describes will generate demand for authentication protocols, code-word systems, and digital literacy tools. The two trust models are converging on the same market outcome — more money spent on verifying that what you’re looking at is real, whether it’s a website, a message from a colleague, or a $1,999 foldable phone.
Click Here For The Original Source.
