Episode Summary
Executive Summary: The episode centers on two big AI-era product stories: Humane’s much-hyped AI Pin launch and OpenAI’s Dev Day platform push. The hosts argue Humane is over-serious, under-tested, and essentially a screenless phone with unclear usefulness, while OpenAI is rapidly becoming the “App Store of AI” through custom GPTs, despite serious copyright, hallucination, and platform-control risks. A lighter lightning round touches Steam Deck OLED upgrades, YouTube’s TikTok-like personalization, smart-home lockouts, and Rivian/Cybertruck hardware quirks.
Main Topics: Humane AI Pin launch and skepticism: The hosts dissect Humane’s launch strategy, the product’s leaked and announced features, and the company’s unusually controlled rollout. They argue the Pin is effectively a screenless phone with a high price, uncertain value, and too much self-serious branding. OpenAI Dev Day and the GPT platform: Alex Heath recaps OpenAI’s developer conference, where the company unveiled custom GPTs, a forthcoming GPT Store, lower costs, larger context windows, and a broader push to become the AI platform layer. AI, hallucinations, and trust: The conversation returns repeatedly to the unreliability of LLMs for factual retrieval, especially when those outputs are used as the primary interface for real-world actions like shopping, directions, or contact management. Copyright, creator compensation, and legal risk: The hosts explore how OpenAI’s proposed revenue-sharing and copyright-shield policies may collide with fair use law, creator rights, and the economics of training data and custom GPTs. Wearable product success framework: Neili proposes a humorous but useful model: wearables succeed when usefulness outweighs fiddliness and face-adjacency considerations. The AI Pin scores poorly because it is fiddly, not on the face, and only vaguely useful. Tech product culture and Apple’s influence: The discussion critiques Apple’s influence on product marketing and launch perfectionism, which other companies imitate even when they don’t have Apple’s resources or coherence. Lightning round hardware and platform updates: The episode closes with quick takes on the Steam Deck OLED refresh, YouTube’s increasingly TikTok-like UX, MyQ garage door lockout drama, and a Rivian R1S test drive.
Key Arguments: Humane is presenting a finished, world-changing narrative around a product that still feels unfinished and hard to evaluate in practice. The AI Pin is best understood as a screenless phone that depends on cloud AI, making it more like an alternative interface than a true replacement for a smartphone. Humane’s launch was overly controlled: many reporters and reviewers were excluded, photography was restricted, and key features were marked as “coming soon.” OpenAI’s Dev Day showed that the company wants to be the AI App Store: a platform where users and developers create custom GPTs and discover them via a store. The custom GPT model is powerful but creates major copyright and compensation problems because uploaded knowledge can include paid or proprietary material. Hallucinations remain a central barrier: the more people rely on LLMs for actions and facts, the more dangerous inaccuracies become. Wearables only win when the value is high enough to justify the fiddliness; face-mounted devices get a usability advantage that non-face devices lack. Apple’s design and marketing culture has taught the industry to overclaim certainty, but that playbook does not work for every hardware company. The future of AI products may involve bots acting on behalf of users and talking to other bots, but that also increases legal, technical, and trust risks. Better UX can still win in the long run, even if the legal and ethical framework remains unsettled.
Data Points: Humane AI Pin price: $700 - Announced retail price for the device Humane service fee: $24/month - Monthly subscription/service charge required for the Pin Humane weight: About the same as a tennis ball - Hosts use this to illustrate how bulky it feels on clothing OpenAI developer attendance: About 1,000 developers - Approximate crowd at OpenAI’s first Dev Day OpenAI developers: Over 2 million - Altman’s reported figure for developers using the platform ChatGPT weekly users: 100 million weekly users - Cited during discussion of OpenAI’s growth ChatGPT context window: 128K+ tokens - New GPT-4 Turbo context size announced at Dev Day GPT-4 Turbo cost reduction: 3x cheaper - OpenAI lowered developer usage costs Humane launch timing: Launched after months of leaks and pre-announcements - Used to emphasize the unusual rollout and lack of surprise MyQ impacted users: 10 million+ users - Chamberlain’s stated user base when explaining the HomeKit integration shutdown MyQ affected percentage: 2% - The company said the third-party integration impacted a small percentage of users MyQ estimated broken garage doors: 200,000 - Math derived from 2% of 10 million users Steam Deck battery test: 2 hours 11 minutes - Sean Hollister’s battery life result on the OLED Steam Deck while running Control at up to 80 fps Steam Deck original test: Just under 2 hours - Original Steam Deck review-unit benchmark for the same game at 60 fps Rivian R1S horsepower: 840 horsepower - Alex Heath’s test-drive impression of the EV SUV Rivian 0–60 time: 3 seconds - Described as extremely quick for a large SUV
Pivotal Quotes: "This thing is real silly." — David Pierce: Summarizing the Humane AI Pin’s underlying concept despite the company’s serious presentation "We think actually computing should be recessed more into the background, and you should interact through voice." — David Pierce: Explaining the philosophical appeal of Humane’s product direction "Don’t build a thin wrapper on top of OpenAI. We are planning to build the obvious features that you would expect for a robust platform over time." — Sam Altman: Describing OpenAI’s intent to own the core AI platform rather than allow others to build shallow wrappers
Implications: AI hardware and AI platforms are converging on a battle over interfaces, distribution, and trust. The winners will need clear utility, credible accuracy, and a workable creator/legal model—or risk becoming expensive demos with short-lived hype.
From the Episode
And last night it finally clicked when I looked at the humane pin, where I was like, oh, it's not on your face. So that changes the whole equation. But then it is all the other things that it is. I was disgusted by my work on the iPhone. So instead, I made a little laser projector that you have to look at your hand. There's no Instagram. You can't do Instagram with lasers. Just didn't work. This thing is real silly. They should have let us and other reviewers actually into it, and they should have been less self-serious about it, is what I'll say. Excited to play with one. I think we're all excited about new gadgets all the time. I want to see how the speakers work. I want to. The personic speakers. Yeah. I want to see. I want you and I both to have one and see if we can hear each other's. That's going to be the worst episode of this show. The worst episode of the show. Just two people listing a title four feet apart. I mean, I will say, again, to go back to the meta things on that one, like I've seen a bunch of people be like, oh, the personic speaker. This is the thing that the smart glasses have been doing for a while. This personal directed audio that you can hear and other people don't.
There's no structured data about the almonds. I mean, there are a million holes you can poke in this thing, and I think we're doing that appropriately. I think the direction Humane is trying to go here is really interesting. And I want to have competing visions for where computing is going in the world because everyone else, Meta, et cetera, Apple, they think the future of computing is more visuals in your eyes, right? Like literally augmenting over the world. And Humane is saying, no, we think actually computing should be recessed more. Into the background, and you should interact through voice. And it's more of that Spike Jones, her future. And I think those different directions need to exist. People need to be building those different directions. It just feels very early. This feels like General Magic or Magic Leap. You know, there's a ton of examples of companies that were just too early to the tech. And I think this is the case here. No, I'll make a big distinction between those things because now we're old and we know a bunch of ex-General Magic people. They were very sincere. Tony Fidel, friend of the show, he was part of the General Magic.
Different modalities, different data sets. At the same time, they're saying, Come build on our platform, we'll share revenue with you, come make your own GPTs, et cetera. And he got asked about this and he said very clearly: he was like, Don't build a thin wrapper on top of OpenAI. We are planning to build the obvious features that you would expect for a robust platform over time, and there's enormous value to build. So that's him just being straight up: like, we will probably surelock you. And I'm going to see, I see a lot of parallels here to the early app store. We're going to build the app. And I think that's okay. Like, you know, the early app store had a bunch of flashlight apps, right? Eventually, that just became a feature on the phone. And it's a better feature as a result. You can go in the app store right now and download a home interior app that is a thin wrapper on Dolly. Like, yeah. Like, we're painting our house, and I wanted an app that would just put paint colors on the wall. And like, half of them are now just Dolly wrappers. And they're.
About The Vergecast
The Vergecast is the flagship podcast from The Verge about small gadgets, Big Tech, and everything in between. Every Friday, hosts Nilay Patel and David Pierce hang out and make sense of the week’s most important technology news. And every Tuesday, David leads a selection of The Verge’s expert staffers in an exploration of how gadgets and software affect our lives – and which ones you should bring into yours.