Snap SPECS, Taste Labs Timeline Turmoil, AI Execs at G7 | Diet TBPN

Snap Spectacles. We talked about it a little bit yesterday. Feedback has been mixed. Not good. People don't like it. Pull up the picture from DJ Kaws. Glasses? What glasses?

It's really—

This one.

It's really tough because if a startup shipped these, everyone, like, they would be able to raise capital.

After you've seen this one, go back to the other photo. Looks really normal now.

There you go.

Honestly.

There you go.

The funny big exaggerated version makes me feel like these actually look really cool now. That, not those, that's too much. But you flip back, I'm into it now. It's actually inoculated me to the, oh, they're big because I saw a bigger version and I like these. They're a little bit blocky, yeah, but it's like a style choice. I don't know, I'm getting, I'm getting pilled. I might pick up a pair.

Here's the thing.

I might pick up a couple pair.

If a startup launched this product and was able to do the demos that they can do, we've tried this product, we've done a number of the demos, that startup would be able to raise at, I would say, easily a billion just based on current market conditions. But they're, a startup is evaluated a lot differently, of course, than a public company that has spent somewhere in the range of $3.5 billion building this product. So yeah, the feedback, from the market has not been great, from activists has not been great.

2% over the last 5 years, and just in the last 5 days, down another 8.5%. And Evan Spiegel has been having to defend his decisions, his investment here. We'll see where all this goes. The question's like, how expensive is this effort? How core to the business is it? How many Snap employees are working on this? They have a great ads business, a great social media, with a network effect should be an AI winner. You know, just increase the ad load, increase the ad targeting, run a really lean, thin operation, and you should be able to be a very, very profitable enterprise. The era, clearly a lot of these investments were greenlit in the early days when the stock was up, when the market was booming, and now we're seeing them roll out and everyone's asking a wildly different set of questions because we're in the— In the bear market. AI era, not the wearables era, but—

Tyler in the chat says Snap down 92% since peak.

Yeah, peak ZIRP.

So you can imagine a lot of the work that was done on these was done when they were a much, much, much bigger company.

Yeah, but to be fair, Evan has been acquiring in this category and thinking about this for probably over a decade. I know I actually talked to a founder that he sold his company to Snap, it must have been 10 years ago.. And they bought a couple of companies and been working on this. And then of course they did have the first version of Spectacles, which were like the Meta Ray-Ban displays or the Meta Ray-Bans, no screen, but just camera. And the rollout for that was really well received, but never quite got to escape velocity where it really moved the needle for the business. But very clear, interesting R&D thinking. Anyway, Evan Spiegel is gonna have to defend himself from our own Brandon Gorrell because Brandon Gorrell came up to me after writing the newsletter and said, I don't think I get it. And I'm like, that's fine. We'll read through your piece. We'll take— we'll steelman it. I'll steelman it. No problem. So Snapchat showed off Spectacles, its new augmented reality glasses, at Augmented World Expo 2026 yesterday. Interesting. I didn't realize that this was an industry conference for augmented reality, not a Snap-specific event. The features are a mix of things you'd want in a daily driver pair of glasses that you'd have on all the time everywhere. Maps, HUD, review of restaurants in your visual field, prosumer features like the ability to collaborate on shared virtual whiteboards and more general AI-powered assistance stuff like measuring distances for you so you don't have to use a tape measure. The broad mix of features combined with the facts that specs are fairly pricey, $2,200 basically, and that they look painful to wear. So Brandon Gorell is pointing out the fact that Evan's ear looks a little bit vent from wearing the specs, the, what do they call that? The bar? What's that thing on the glasses that goes in the back? I don't know, whatever that thing is, it's a little thick, it's a little heavy. There's a battery back there, probably some compute. And so that is compressing his ear a little bit. Imagine wearing that for 4 hours. Maybe it gets a little bit tiring. We will see how—

Other scenario, he's getting some cauliflower ear, he's training. He sees Zac has gotten into MMA. He doesn't want to be left behind. So we don't know.

Are guys who golf every other weekend in the summer really going to drop over $2K so they can put on their pair of specs just when they need to see how many yards they are from the pin? I think a lot of golfers do have disposable income. The price tag might not be the issue. The question is, does this look cool on a golf course? Is this something that has badge value if you pull out, like a nice rangefinder, like a Titleist bag or something with a great brand? It feels like to make it cool, it's gotta be on the PGA Tour. The heroes that people look to need to be using this actively for the golf community to really—

Yeah, and so many cool use cases, but are any of them killer use cases?

Is that true?

No, I'm just saying, I'm saying like that's a cool use case. You're trying to understand how a piece of furniture is gonna fit into your room. I don't—

Yeah.

I do that—

Mentally.

When I'm doing— You do it in your head. Like if I'm doing an interior design project, I might need that, but that's like a specific moment in time.

Yeah.

Maybe once every couple years at most.

Yeah.

A lot of people, you know, some people are kind of constantly adding furniture here and there, but a lot of people it's kind of you set it and forget it. Again, unclear why this is something you would want on your face. All the time.

You can buy rangefinders for around $150. They're not fragile. Also, a lot of golf, a lot of golf heads, they're out there for more than the battery life. They're more out there for more than 3, 3.5 hours. 3.5 hours might be enough for 9 on a busy course. They're doing 6 hours out there sometimes. You don't want to be out there with your augmented reality glasses and they die on you. Our DIYer is going to drop this much money just so they can have easy access tips for their home projects. Our startup's going to be willing to drop $2K for every employee who wants to collect collaborate in AR. All of these are examples touted on Snap's specs page as things you can do with the glasses, and the features do seem super cool. It's just hard to imagine any one of them justifying a $2K price tag, especially because they look painful to wear. And so that's your point about killer features. I disagree. I don't think that these products need a killer feature. I think the original killer feature of the iPhone was the phone. Like, people were already carrying phones.. And the iPhone was like, we debated this before, but it had some call dropping problems, but it was a replacement for your dumb phone.

Yeah.

And then the fact that it also was an iPod was an extra feature. And then the fact that it was an internet browser was another feature, but it replaced like very, very basic things.

My thing is like, I don't think as cool as the tech is, I don't think the tech is ready to be a daily driver computer.

Yeah, well, I think it needs to replace a very regular everyday interaction thing like a screen. And so that's why I still think VR is like a replacement for the home theater, maybe a replacement for the 80-inch TV, but 80-inch TVs are like $500 now. And so you gotta get it to be better and you gotta have enough for everyone in your household to have one and it's gotta be a better experience. But in that world—

The other challenge is like a lot of these, I mean, like a lot of these use cases, I don't feel like are that aligned to Snapchat's user base. And that's like the biggest—

Their business.

Like a $2,000 device doesn't really align to their, what I believe is their core demo.

Yeah, and so Bugle Capital bloke's asking the question, how did this happen? Do you know how deeply broken a culture has to be to ship this product and let the CEO walk around like this? Again, I don't think they look that bad, There is this question of, you know, is this a serious product? The fashion part must be addressed first. I guess the taste memo never made it to Snap. If you're enough of a dork to have these on your face, you won't even get the chance to say, "May I meet you?" Wow. People are very, very upset about these. JB says, "I legit think this may be the first product ever to hit the market and not sell a single unit." That's ridiculous. They're going to sell a few to people that want to demo them. There's fans that buy every product. Palmer Luckey has a collection of augmented VR glasses, you know, the collectors will get them.

Let's pull up this post from A Capital.

Yeah.

Because this might be a killer feature.

Yeah, we can watch this. This actually might be the killer feature, honestly. Thanks for joining us here today. You're wearing your new specs you just unveiled. They cost $2,195.

The stock's down more than 5%.

The sound effect, the sound mix is really good. Thanks for joining us here today. You're wearing your new Snap. The sound mix. The fact that her voice gets quieter when you go inside the headset is really what does this. So, so funny.

I feel really bad for the Snap team. I think like the, I want them to win, but I don't think this launch will get them the level of traction that they're gonna need to justify further investment is my current.

No. Unless you're— Yeah.

Spiegel just doesn't care.

Yeah.

And, you know, continues to double or triple down, which is totally possible.

Yeah.

But I think at a normal company, this would kind of be the last shot.

It's tough. You're competing with these hyper-efficient Chinese companies. There's this company, XREAL. We demoed this. I got a pair for Tyler, took them for a spin. They're not quite there, but they're— way cheaper. You're looking at a couple hundred dollars and you get a screen that's like not even really augmented reality. It's sort of like sunglasses with a screen inside, but then it projects like a 200-inch TV in front of you. And these are actually sort of— it's more chopping at the daily use cases and less doing like frontier technology. So you can play video games on them because you just plug the HDMI from the Xbox into the device, and then you just have a big screen in front of you. And if you don't have a TV with you for mobile gaming, there's a whole bunch of different things you can watch movies on it, do the basic things that people do with screens. And I think that XREAL is on a path to like commoditizing this in a pretty significant way where it's not—

Not to mention Meta Ray-Ban displays are—

$799.

Yeah. And I do believe that these have more, quite a bit, like more features, functionality. They have a bigger developer network.

But a lot of the killer features of like on-demand AI, instantiating a generative UI, answering a question for you, that can be done with a Call of Duty HUD.

That can be done with a Call of Duty screen. All this, like again, it's cool, but I cannot, I don't know anyone. That would do this.

And you face this crazy cold start problem where the developers don't make anything because there aren't that many users. Like it is so crazy that even in the era of vibe coding and software being like free, we're not seeing breakout Apple Vision Pro development. Like the Apple Vision Pro, which I need to bring in for Scott, he wants to demo it, comes with a really, really impressive demo where it, defines all the walls and then one of the walls opens up and is a portal and through the wall you see this like dinosaur land. You'd love it as a dinosaur expert. Kids love it. And then a dinosaur comes through the portal, comes into your— What kind of dinosaur? It's like a velociraptor type thing.

Let's double check that.

And a butterfly lands on your finger and it feels like it almost— you almost feel it because it's so— It's like tracking your hand and it lands. Very cool. It's like a 5-minute demo that the Apple team clearly worked very hard on. That feels like something that could be expanded on very cheaply in the age of vibe coding, and yet no one sees it as an economic opportunity. Developers are just— they'd rather build an app for the iOS App Store. And so, no one's really going there to compete. So, while these demos look really cool, where's the ecosystem going to come from if they're not selling millions and millions of units? Yeah. And people are ready to purchase games or, you know, see ads or do anything that could monetize a business built on the back of this platform. It's a very, very tricky proposition to get a platform like this up and running. Well, the other story that's burning up the timeline is Taste Labs put the timeline in turmoil, people going back and forth. So yesterday, a former Exa AI Labs founding team member introduced her startup Taste Labs, whose mission is to end AI slop. Quote, "This requires turning a fuzzy subjective domain into something we can measure and codify. We're starting with design," her post says. More specifically, Taste says they're working with Frontier AI Labs to improve their models along taste lines through data labeling and AppLayer startups to improve the aesthetics of their products. This has been a critique of vibe-coded projects. They all sort of look the same. Of course, there are examples of really cool projects, but people were starting to say, oh, this has like the vibe code look to it, or this model's not good at front end, et cetera. Her goal is to fix that. Tai's post, her post was immediately, immediately went viral, generating tons of opinions on X and getting over a million views in 24 hours. People's main complaint is basically you can't program taste. It's impossible, they say. But the steel man is that the AI aesthetic, AI's aesthetic, that the aesthetic output can be improved and that it's perfectly reasonable for a startup to try and capitalize on that opportunity. I wanna talk to you about Taste, about your feed. Is it scalable? Is it not? Give me, take me through some of the critiques here. Tell me what resonates with you. And then I have a take about where the business—

So I think the main thing is— Main thing is people have taste fatigue. They don't want to hear that word anymore.

Yeah.

I don't want to hear that word anymore.

'Cause for the last 6 months, maybe last year, it's been like the code word. Like, what will we do when the AI can do all the technical stuff? Well, we'll have the taste.

Yeah, and so yeah, I think there's fatigue around the usage of the word, even the conversation. We've never even like waded that deeply into the conversation, and I'd like to keep it that way. That being said—

Just to set the table on the critique, like a lot of people outside of tech are critiquing it because a lot of SF people in tech are saying taste is so important and the outsiders don't see San Francisco tech—

As being tasteful people.

As being particularly tasteful people. From a fashion perspective, from an art perspective, or curation perspective, it's sort of known as the t-shirts and athleisure community, and that's— it's optimized. It's devoid of taste by design. It's about efficiency, not taste.

Yeah, and then I'll say one more thing, and then I'll steel man Taste Labs. But so the main thing is, when I think of taste, like product taste, I think of linear, right? Linear has always been very opinionated. Very quality-driven, they want to grow quickly because of how great the product is, like, you know, very, very design-driven. Like, that is a company that I think generally has very high, you know, good taste, right? The problem is when you have good taste and people pick up on it, they just start sort of just like blanket, like, copying you, right? Sure, sure, sure. So then there's an entire generation of companies that just look like Linear, right?

Yeah.

From their website to the actual product, And so taste is something that people curate themselves, but then as soon as it's copied, then it's like fundamentally like not tasteful in my view, right? Then it's not original. I think taste has, you know, you need some originality to be able to combine, you know, do 1 1 2. And another example is like—

Isn't 1 1 3?

Sorry, sorry, sorry. 1 1 2. 1 1 2. That's the ad.

Oh, yeah.

That fake startup ad. No, so another example is like Squarespace.

Sure.

Like Squarespace took like high-end website design and then just democratized it. Yeah, democratized it. Anybody could have a pretty website. And then you started to just— I would just look like, OK, is this a Squarespace website, or did this person make it? OK, it's a Squarespace website. And it's not really that much of a knock, but— like it wasn't like the company's own taste that led to that output or the people that they worked with. So it commoditizes really quickly and then it ceases to be tasteful. That being said, just helping AI Labs create better looking outputs and working on that problem feels like a pretty good way to build a, at least temporarily, a pretty big business.

Yeah.

Because this is something that, users really care about, the labs really care about, hyperscalers even care about, right? And so I think that while Tase Labs got a lot of flack over the last 24 hours, their pipeline probably exploded, and I bet they get a ton of business out of it. And very unclear what this business looks like in 5 years, like a lot of the other companies in the category, but I bet they print in the short term.

Yeah. I feel like a lot of the training data, data labeling—

Also the name, the name is like kind of perfect rage bait. Taste Labs, we're building—

Taste, yeah.

Taste in a lab. We built it, we made it.

Yeah, it is funny because you could do the inverse and say our job is to just identify things that are not tasteful, and the end product would be exactly the same because that's just your negative data set and everything else is positive by that. Design, but a lot of the data labeling projects have just been, does the button work? Does this render properly? Like, is this just, is this functional? And some of that's been able to be, you know, looped in a reinforcement learning environment. Some of it's been able to be encoded, just tagging, okay, does the, does the photo have 6 fingers or 5 fingers? Like that was a useful, useful piece of data labeling that happened probably 2 years ago. Now, there's a bigger question about what actually looks good and then how do you represent like a diversity of tasteful designs such that everything just doesn't collapse into like the new corporate Memphis and everything that's AI-generated has the exact same flavor. Like the, it's not this, it's that, but for design would be the bad outcome.

Yeah, which is like already extremely easy to clock. Right now, yeah. And I got a deck last night friend of mine's company, and my first piece of feedback is like, are you okay with everyone knowing that you didn't put any effort into design? Because it's totally possible the answer is yes, but at least you should go into your fundraise process knowing that everyone is gonna know that you just tried to one-shot this. You're slopping it up. Which again, for some businesses is fine, and some investors is fine, it's gonna turn some people off.

Yeah. Yeah, people are going all back and forth on this. This highlights a fundamental misunderstanding in tech. Not everything can be codified and analyzed. Even if you make AI imitate taste, whatever that means, it still won't mean anything. Taste emerges from craft, context, meaning, subjectivity, and genuine care. That's sort of true, but just increasing the quality of design is valuable. But yes, it is a tall order when you use the word taste because so much baggage has been assigned to that. There's also a bunch of back and forth going on about Netflix buying Lionsgate potentially. Sources familiar with the matter disputed Semaphore's reporting. There was a whole bunch of back and forth about is Netflix gonna buy something else? Were they in the bidding war for Roku? And they're taking shots at each other. Semaphore, what you got? Ben Smith says, congrats on helping with the cleanup. They could have gone to Variety and chose you. I'm not even sure what this tweet means, Sharon. Cleanup on aisle Semaphore, my dear Ben. All I'm saying is they are now saying on the record that they're not interested in Lionsgate, never were. Would have been happy to amplify your scoop too. They're fighting, they're fighting. The timeline's in turmoil. In market news, more Fed officials have signaled a rate increase rate increase as the next move. The central bank held interest rates steady as Kevin Warsh's first meeting as chairman, but 9 of 19 Federal Reserve officials penciled in at least one rate increase by year's end, up from none in March. And so, the market is selling off, down about 1% today based on that. [FOREIGN LANGUAGE] There's also a G7 meeting that's happening between AI leaders. Donald Trump was seated next to Sam Altman and, and Demis Hassabis from Google's DeepMind at a summit in France on Wednesday discussing AI export controls and all things AI. The AI leaders are huddling at lunch with the G7 in France. And I'm sure there'll be more things to come out of that. There's still more back and forth on export controls. Fable 5 is still embargoed in some way and they're working through that. There's been a little bit more according to the reporting, but not much major movement there. And then also US has held off on blacklisting China's Deepseek, more than 100 firms, and more than 100 firms that are deemed security risks. That was part of the Fable 5 rollout was that there was a South Korean telecom company, according to the reporting, that had access to one of the most advanced models from Anthropic. And that, firm, that Chinese, that South Korean telecom company had potentially ties to China in some way. And so the US government was skeptical of that South Korean telecom company. And so there was a debate over that and whether or not that crossed a bright line. And so from the Washington Post, Anthropic later disclosed that the list had ballooned. The list of companies that would be getting the most advanced models had ballooned and roughly 50 additional entities had Already received access, senior officials began to consider using export controls to claw back the technology after the company did not identify new recipients for days. When Anthropic finally turned over names, the administration discovered that one recipient was a South Korean telecommunications company the administration suspected, alleged, of having ties to China, officials said. And so that is the key of the dustup there. But we'll be continuing to follow that.

In other news, much more tragically. And lastly, some very sad news. Many of you will have seen this by now, but Joshua Baier, who is the CEO and founder of Capital Factory, was in a—

Plane crash. Plane crash.

In Laredo, Texas. I guess early, early this morning coming back from Mexico to Austin. I unfortunately never got to meet Joshua, but only heard tremendous things about him, and he really was an important figure in the Austin startup community. So sending our prayers to Joshua's family and friends. And yeah, really, really tragic.

Rest in peace. Yeah.

Rest in peace.

Well, we should close on a positive note in some way. There's a whole discussion over SpaceX potentially using the high share price, the incredible valuation to acquire more companies, create a roll-up. A piece in the Financial Times we can run through another day, but Bill Ackman shares a Hall of Fame opening sentence. One of the things that makes SpaceX so valuable is how valuable it is. A tautological value argument. Of course, what he's actually getting at is that while the stock price is so high, that serves as a currency for acquisition, and when you're a public company, you can acquire companies very easily with your public stock. And so there's a very interesting window. Ben Thompson wrote about it on the back of the Cursor acquisition closing or being announced that the option has been exercised. But it is a very interesting debate. We touched on it a little bit yesterday. Is there going to be an acquisition spree? Will there be a roll-up? Will SpaceX buy NeoCloud assets, energy assets? Mm-hmm. Chip assets like TerraFab has been talking, what's involved in that? They're trading at what, 10 times Intel at this point or something? I don't know, what is Intel market cap? But there's so much that they could do. It's not 10 times Intel. Intel's a $600 billion company. So, that would be a big one. But there's a lot in the supply chain. There's a lot in the AI world and the space world that they could partner up with if that's the direction that they want to go, but it would be a very different direction. Everyone will be watching it very, very closely. Anyway, thank you for tuning into TVPN today.

And have a wonderful afternoon and evening, folks. We'll see you tomorrow. Goodbye.

Goodbye.