（预览）OpenAI带来的提醒、Netflix不断扩张的胃口、远程办公问答、Taco Bell，以及太空数据中心

Hello, and welcome to a free preview of Sharp Tech. Hello, and welcome back to another episode of Sharp Tech. I'm Andrew Sharp, and on the other line, Ben, how you doing?

I literally just said this on Dithering, so I hate to reuse my takes, but—

Oh, boy.

I made the analogy: You haven't experienced this yet, but once your kids go to school, the great thing is that everyone's looking forward to Christmas vacation or spring break or whatever it might be.

Mm-hmm.

What happens is, the moment you get there, they immediately get sick.

Ooh.

You probably remember this when your kids were a year old. Your body's like, “Okay, I made it. I'm done now.”

Yeah.

That's my sensation—

That still happens—

That's my sensation right now.

To me in adulthood every now and then.

That's my sensation right now. I've posted the year in review that I do every year last night, and I'm having to remind my body I still have 2 more pieces of content to put out. One's down. This is the last one. It might get loopy, but we're going to do our best.

There you go, the home stretch. I feel like school's already out. I'm in a great mood. It's going to be nice and light—a mailbag episode today.

I know. One of us has to do a good job here, so I'm trying to hold it together for the both of us.

Okay, good. I like that that's the headspace you're in as we venture forth here. First order of business—

I need a lot of mullet today. A lot of party in the back.

Party in the back.

Yeah.

It's going to be a little bit of news up front, and then a lot of partying in the back.

No, but just to be clear—

Yeah.

It happens every year. I will take care of it since you forgot to do it. This is the last episode for a couple of weeks. We'll be back the first week of January.

Mm-hmm.

You will still be podcasting with The GOAT—

I was going to say—

the great Paul Todd.

It's not actually my last piece of content because I'm recording The GOAT later tonight, and we'll keep The GOAT rolling over the holidays because basketball never stops, Ben. People can look forward to that. But here on Sharp Tech, that's right, we won't be back until the first week of January.

1. OpenAI Improves Image Generation

For now, we did have some news this week before we hit mail—some news I wanted to address quickly. The first news item I'll read from Bloomberg: “OpenAI is rolling out an update to ChatGPT that's intended to generate images better and faster. The latest move by the artificial intelligence developer to bolster its flagship chatbot amid heated competition from Alphabet Inc.'s Google. The company is also creating a new section within ChatGPT's mobile app and website meant for people to make images rather than just doing so in an interaction with the chatbot.”

So, Ben, you've had a couple of days to play around with GPT Image 1.5. What are your impressions of this new feature from ChatGPT?

It seems pretty good. I'm enjoying it quite a bit. The—

I am too.

Yeah, I was using Gemini for a little bit, which I think speaks to this being a really important area for ChatGPT to shore up. That was a reason to go use Gemini: The image generation was really good.

It was basically—

I know people got—

The only reason to use Gemini over the last several months.

Right. I mean, if you want to generate a picture of a turkey, go to Gemini. If you want the correct instructions for how to cook it, go to ChatGPT. That was sort of my takeaway.

Yeah.

But, yeah, it seems good. Personally, based on what I've done on both, it's interesting because I wrote the daily update. I had this long thread that had been going on for a little bit, and my single favorite image across both apps was actually from Gemini.

But as a whole, I think I preferred the ChatGPT ones. They're both great, they're both amazing, and it's really important for people who already have ChatGPT not to think about going somewhere else.

Yeah.

I think getting this out ASAP was important. They also did the GPT-5.2 update. It's still the same base model. It's more post-training, but a good response by them. Then we'll see what happens in the new year when, supposedly, their real next-generation model is supposed to come out.

Yeah. It was a really clean user interface. ChatGPT is just good at this. Even the name of this product, ChatGPT Image 1.5—clunky as ever, consistent with the—

Right.

OpenAI house style.

I thought you were about to say, “You said that after they're good at this,” and then you listed another horrific name. I'm like, “Wait, where are you going?”

But here's the thing: The upshot of that name is that everyone who uses this will refer to it as just ChatGPT, which again reflects good product instincts. I was hearing about Nano Banana for the last several months, and it's like, “Where do I find that? What is that?”

Why is it called Gemini? Yeah.

Exactly.

No, for sure.

And ChatGPT—

It's so funny how bad Google is at stuff like that.

Is just good at this.

No, and I think the new UI, with prompts for what to do, is so—

It's really helpful for someone like me, just for the record.

It's so basic—

Yeah.

And yet no one does it. This is one of those things I come back to on the question of ChatGPT versus Google long term. I always, as I wrote this week, say ChatGPT or OpenAI is better at product. What do I mean by that?

2. OpenAI Wins On Product

It's funny that everyone talks about product in Silicon Valley. There's a reason why Apple just cleaned everyone's clock for decades, because it turns out technical people just suck at products because they know how it works. They know how to do it. How do you put yourself in the mind of a normie?

The thing is, it's not just the capability, right? Everything in that new UI was already possible before. It's not just that people need to think of what can be done or have ideas, which is in there. That's part of it—generate a Christmas card, right? That was one of the options in there. People don't know what's possible. The AI can do so much, and people have no idea, and you have to lead them down that path.

The tech people who do know what's possible will think, “All right. The performance here is going to sell itself,” and they don't go—

Right.

The extra step to make it idiot-proof so that people can discover how amazing all of it is. ChatGPT has cracked that code like nobody else in the space over the last 2.5 years. I thought that the release this week was a continuation—

Yeah, you were churning out images.

Of that effort. Yeah, exactly. I was having fun with it.

As far as the competition with Google is concerned, to go back to your point a couple of minutes ago about the advantages that OpenAI has, I never felt that compelled to download Nano Banana until 4 days ago, when a friend was making fun of another of our friends generating Photoshops with Nano Banana. I was impressed by how good the Photoshops were, and I thought, “Maybe I need to check out Nano Banana.” Then, literally a day later, ChatGPT releases this feature.

That's why they needed to get it out.

Exactly.

They need you not to download Gemini.

Keep the normies in the ecosystem.

Yeah.

I still have not tried out Nano Banana, but I had a lot—

Well, so—

Of fun with images.

The other thing I thought was really interesting is that I redid all the prompts from that other thread in ChatGPT. It's funny because that thread was— It was this ongoing conversation, and there'd be some funny thing, like, “Oh, it'd be funny to illustrate this in AI.”

So I had the idea, and this was AI as a tool: I'm going to go and do this thing, and go to AI and generate this image.

Right.

I did the same thing. I just copied all the prompts exactly word for word. What ChatGPT did that Nano Banana did not—and again, this seems so obvious—

Mm-hmm.

But it actually didn't occur to me to ask for this—it made all the images consistent. There were characters that repeated, and the style was all the same. I didn't have to tell it to do that.

So, number one, that was super delightful. “Oh, yeah, this is actually way more fun. It's more consistent.” Whatever they do with the context window and using old images to make the new ones is interesting.

Right.

It made me want to make more images.

Yeah.

Because it was like I was drawing a comic book of these different things that were happening, and then I wanted to build on it and go from it, whereas with Nano Banana and Gemini, I only went to it when I already had an idea.

Mm-hmm.

So I had an idea, and then I went there to generate it. This was pulling ideas out of me, and that's another really important sort of product capability. It's not just that you know what's possible and it's teaching you, but can it inspire you?

Yeah.

Can it make you want to use it more? And again, these are things that are funny because this doesn't sound like a moat. Well, can't Google do that? Yeah, they could, but they don't.

They're not good at it.

They don't again and again and again.

Yeah.

So I thought it was pretty encouraging from an OpenAI perspective, not just that the model was good—

Mm-hmm.

—because it's not just about models. It's also about products, and I thought this product rollout was more compelling than I think you even realize when you look at relative images.

Totally. And I sat there the other night—I think it was Tuesday night—and spent about an hour just generating images of my kids in different scenes and having a great time. I do wonder how OpenAI is going to pay for all this over the next several years, because I'm sure it's expensive—

Yeah, well, they had that—

—to generate—

—super weird video about how they don't have enough compute. I mean, you know what? I've been complaining about the prospects of not having enough compute for ages and ages and ages.

Mm-hmm.

This is my whole thing about the API.

Mm-hmm.

Just let Microsoft handle the API. You need to be delivering the most compelling user experience for your consumer app, which you have the chance to rule the world with.

Yeah.

So I feel bad that I backed down on my critique there—

—

—because I was totally right. So there's my 2025 lesson number 1.

Yeah.

I should've been even harsher on ads. If they had introduced ads in 2023, the ad product today would be good, and consumers would be habituated to it.

I feel like you've made up for that omission or blind spot—not, it wasn't a blind spot, but you did back down, and you've made up for it over the last 2 and a half months on the podcast.

With vociferousness.

The vociferous—

It's going to be even more vociferous.

The vociferous, emphatic—

That's right.

—admonition for Sam and the gang out there at OpenAI.

Yeah.

But finance questions aside and compute questions aside, this week was another reminder of the very real advantages that OpenAI—

They're good. They're good at this stuff.

—has in the space.

Yeah. No, they are.

Yeah.

And I think the whole compute question—it's possible that OpenAI's biggest problem is actually that they got too popular too fast. It's a real catch-22. On one hand, that's their biggest advantage.

Mm-hmm.

On the other hand, I can't remember if I've made this analogy before, but a problem Apple has with the iPhone is, “Whoa, look at these other companies coming out with these new, innovative features,” whether it be a flip phone or whatever it might be. It's a lot easier to roll something like that out when you're looking to sell single-digit millions or maybe low double-digit millions.

Yeah.

If you're selling at the scale Apple is, it's really, really hard, because you don't just need to invent it. It has to be reproducible at the scale that they're doing it at.

Mm-hmm.

And that just takes longer.

And an error affects 800 million people as opposed to whoever—

But, yeah, but just to be able to do it, right?

Yeah.

Apple's supply chain almost killed themselves when they introduced Face ID.

Mm-hmm.

They had so many difficulties getting that out and getting it working, and I heard some unbelievable stories about heroic missions by Apple's team to get that out the door.

Yeah.

And that's because they sell so many freaking phones. The Face ID worked. It wasn't an issue of it working. It was: could you get it working at the scale of tens of millions of units?

Mm-hmm.

And so they launched that. They could not have launched that a month earlier than they did.

Yeah.

I think even that phone was somewhat delayed. I can't remember exactly. I think that was during COVID. I don't know, I can't remember exactly. Maybe it wasn't. But that's just an example of how it's really hard at scale. As long as AI has a marginal cost attached to it, OpenAI feels that to a massively greater scale than anyone else, including Google.

Mm-hmm.

If they roll out a ChatGPT feature, it's rolling out to 900 million people.

Yeah.

That's really, really hard.

And people are using it every day—

Yeah.

—all day. And that's—

Yeah.

—expensive. So another thing that OpenAI has rolled out—the other news that we haven't hit on this show, you hit it on Dithering—is that Disney licensed its characters to Sora. My question: Does that matter? What do you think?

No, probably not. Sora was another viral moment, and the fact that a company can produce multiple viral moments, I think, is also a positive indicator.

Mm-hmm.

It's number 25 in the App Store now, or whatever. Is it going to be a thing? I don't know. But the point I came back to—and it's funny, I made this, this is the end-of-the-year problem—

Yeah.

—because I wrote this long thing about the importance of totems for content, and then I was doing my year in review, and I wrote an entire article about this 6 months ago—

—

—that I totally forgot about. I could just link to myself.

Great.

But this idea that you need something to organize around—I was talking about it on the podcast last week. And in a world of AI, I think that's going to give more value to us, right?

Yeah.

Because there are tens of thousands of people listening to this podcast. They can talk about it. They can say, “We should do more about this.” Stay tuned. We'll see what goes on. But the things that people have in common are really valuable. And you see—

I mean, it's why Warner Bros. is valued at $80 billion now, as—

For sure.

—people fight over the IP that was generated 30 years ago.

Yep. And so what did you see the day Sora launched? Or the first week, right? It was all with a bunch of nerds—

No copyright enforcement, and—

Well, that's true.

—everybody loved it.

That was fun. Actually, no. So a few things. What I was going to go with is that everyone who had it initially was in tech and interested in AI.

Mm-hmm.

And so all the cameos were all of Sam Altman.

Yeah.

Because why? He's a common figure who was available, and so that's 1, 2—

And anybody who wound up in the Sora ecosystem knew exactly who Sam Altman was, and so—

That's right.

—watching him do funny things was amusing to everybody who was—

That's right.

—quirky enough to have Sora.

And to your point, they also did no copyright to start.

Mm-hmm.

And it was funny to make Pikachu do absurd things. And so this is a very bullish signal for Disney, even if Sora doesn't amount to anything, because what it's speaking to is not just the enduring value of IP, but I would argue the increasing value of IP that's widely recognized and widely—

Mm-hmm.

—seen. In a world where everything can be customized and individualized, things that are in common are going to be more and more valuable, and that applies to Disney pretty much more than anyone.

Yep. That makes sense. I'm skeptical that adding Darth Vader is going to bring Sora back to life all of a sudden, but it's an interesting marker to lay down for Disney and a sign that they're going to be forward-thinking in that space. In any event, speaking of Hollywood—

Well, the other thing, too, is they dropped the cease-and-desist letter on Google the same day, or the day before.

Mm-hmm.

And it reinforces this Google-against-the-world aspect to AI.

Yeah.

Google has its own chips. It has its own model. It's all Google all the way up and down. Google doesn't like doing deals. Deals don't scale. Google is the ultimate everything-must-scale sort of entity.

Yeah.

And on the other side, you have OpenAI, NVIDIA, and now Disney, and this is where Apple should end up. But OpenAI hired Jony Ive, or whatever it might be. It really is Google versus the world.

Mm-hmm.

And Google might win.

Yes. I look forward to tracking all of it in 2026. To keep it moving, we have a ton of great questions related to the last episode on Netflix and Warner Brothers. We have some new Netflix news to discuss.

Actually, speaking of Google, just one thing that is not on the rundown but happened this week: Google Gemini 3 Flash did come out. It's an amazing model that's extremely cheap. All along, the Gemini Flash models have arguably been the more interesting and compelling ones, just because they're closer to the idea of AI. It's not zero marginal cost, but it's a whole lot closer.

So it's cheaper for Google to run when you say it's cheaper.

It's cheaper for Google to run, and it's cheaper for developers to use.

Okay.

The API costs are drastically lower, and that is almost the more interesting performance to look at. How good is the super-duper cheap model that you could actually build a business on? It's really cheap, and it's really good.

Mm-hmm.

Interesting.

Okay. We got a ton of great questions related to the last episode. I'll read one note that we got on the antitrust stuff. Full disclosure, a handful of people were striking this note, and I chose the meanest email of them.

So Grady says, “Andrew and Ben, listening to last week's discussion of the Netflix acquisition of Warner Bros., Ben's choice to focus on attention as the scarce resource is facially arbitrary. Why is attention the marketplace on which Netflix should be evaluated? Why not the marketplace for protons, neutrons, and electrons? By Ben's logic, surely that is the only scarce resource for which all companies compete.”

Do I wait until the end, or should I just start at the top? Or do you want to get this whole thing out?

No, wait until the end. Wait until the end—

Okay, fine.

—out of respect for Grady.

No, sorry, there's not a lot of respect for Grady.

Okay.

But continue. I'll continue. I've got to be honest here.

Yeah. Spoiler alert. While Netflix may perceive itself to be within a competitive marketplace for attention, where in law or common sense does it follow that the company's current strategic positioning is the only market in which anticompetitive effects can manifest?

Netflix engages in many marketplaces: for creative talent, for subscription video services, and for network bandwidth. Ben paid lip service to this, but why does the company's competitive participation or positioning within one market make it so that anticompetitive effects in other markets are not worthy of public oversight and intervention?

By that logic, Amazon's weak positioning within the market for streaming could justify its acquisition of Walmart, despite that obviously being a contraction of competitiveness in other markets. I sincerely hope you read this because I would prefer not to write off Ben's analysis of antitrust in the future. Please don't make me into an anti-monopolist.”

All right. So what do you think?

No, just to be clear, it's the last line that prompted the disrespectful comment.

Okay.

You can write off whatever analysis you want. Framing it as, “Ben better answer this or you're out of my mind,” that's not how—

Get out.

—it works at scale. Yeah, sorry. I'm going to drop my takes, and I don't care what you think about it.

Mm-hmm.

So I would abstain from the implied threats in the emails in the future if you want a respectful answer. I don't think it's facially arbitrary.

Yeah.

The confusion I have with this email is that it sounds like Grady's been reading for a long time. It's the single most predictable take in the world that I say the only scarce resource is time or attention.

It was awfully predictable.

So where have you been, Grady? I'm sorry I need to also insult your reading comprehension and memory. This is all in good fun, Grady, but this is literally a thesis of Stratechery.

Sure.

Back in the day, the gating factor was actual physical distribution. In the case of movies, you literally had large reels, and you had a few of them that had to be transported to a theater. They would run the theater; a guy had to be up there to switch it over. If you watched it—and if you have very good eyesight and perception—you could see the switchover when it happened.

Wow. Yeah.

There would be marker frames in there and things like that. The key in that market was controlling distribution. This is why you go back to the 1948 consent decree about the studios not being able to own theaters. It was a problem that they had that top-to-bottom control because distribution was the most important thing.

Right.

On the internet, you can watch any video that you want. I've had this knock-down, drag-out fight before with people. They'd be like, “No, distribution is not a zero-marginal-cost thing. You have to pay Google and Facebook.”

I'm like, “You're not paying Google and Facebook to have your content online. Anyone can go to your website and see your content. You're paying them for attention, for getting people's eyeballs on that content.”

Yeah.

Is it functionally, from a business perspective and in your accounting, kind of the same? Sure, I get why people call it distribution, but it's really important to key in on the specific details: the dollars you were paying on the internet were not for people to be able to access your content, like you used to pay for printing presses, CD presses, DVD presses, shelf space, trucks to deliver it, or all these sorts of things. You get all that for free now.

Yeah.

And it's because you get all that for free that the competition is infinitely larger than it used to be. Netflix is competing with YouTube. There's no ifs, ands, or buts about it. You have a TV in front of you, and the question is, what's on that TV?

Hmm.

It could be Netflix. It could be YouTube. It could be linear TV. It could be Disney+. It could be the Apple Aerial screensaver, which I continue to love. But only one of them is going to be on there.

Yeah.

That is the plane of competition. It doesn't matter if I have a game coming over a cable wire, or I have a game coming as packets over the internet, or a show or a stream or whatever.

Setting aside the antitrust political questions, if I'm a businessperson, if I'm trying to make strategic decisions, you have to drill down to what is the constraint.

Right.

Where is the bottleneck? That is the bottleneck. It just is. Now, as I said last episode, that might not matter in court because this is a political process, and people are going to draw these, if I can use the word, arbitrary lines.

Mm-hmm.

But you're the one drawing the arbitrary lines, not me. By the way, you don't get to say, “Oh, Ben said this,” but he didn't really say this. I said they compete in other areas. I acknowledged that. You don't get to dismiss it because I didn't say it strongly enough the way you wanted it to.

Yes, there are monopsony-type questions, like how many buyers there are for all these sorts of things. That's all true and valid, but that doesn't change the fact that, from a consumer perspective, Netflix is 100% competing with YouTube. YouTube is by far their biggest threat. I think it's impossible to make sense of this acquisition without acknowledging that.

I think that is all well said. I wanted to give you some room to respond.

I'm fired up now.

Yeah. Well, look—

Now, if you don't like it, that's fine. Cancel the subscription. I'll be fine.

No, stay, Grady. We enjoy the emails. We enjoy takes from everybody, takes from all sides. But, yeah, we have to wait and see what the government does on any of this to talk with more—

No, by the way, just to be clear—

Okay.

—and I'm over being mad at Grady. It's fine if you disagree with me on this. People disagree with me about stuff all the time.

Yeah.

I'm going to think you're wrong, and we can all be okay with that, and you can think I'm wrong.

Well, and I think—

Like, so, but this is—

—the Clayton Act, the Clayton Act does—

Well, just go back.

Okay.

For 12 years, if you want to distill what Stratechery is, I wrote my 13th-year interview this week.

Mm-hmm.

This has arguably been the central theme from beginning to end.

So it is what it is.

Totally. And look, if I were making Netflix’s argument, I would point out that hundreds of millions of people around the world have TVs that have the YouTube app right next to the Netflix app in the tiles.

Well, just wait.

And more people are watching—

Once you watch a great Warner Bros.—

…YouTube on TV now.

…once you watch a great Warner Bros. movie on Netflix, you can see it win awards on YouTube.

Oh, that’s right. It’s all accelerating toward Hollywood’s endgame here and a new, really bizarre future.

3. Netflix Wants All Of Television

Thinking about Netflix, I think part of what’s interesting about that company is that, on one hand, they’ve come to dominate what was their original market, streaming on demand. On the other hand, it’s a case where the appetite grows in the eating. People say that about China sometimes. I think with Netflix, the endgame for them for the last 10 years or so has been that they want to just be a stand-in for all TV and all entertainment.

That’s right.

And that’s where YouTube is encroaching on their growth potential as we speak. YouTube—more people watch YouTube on TV than Netflix on TV.

That’s right. No, I wrote that a decade ago, or early on. The bull case for Netflix is: You went from the cable bundle to Netflix.

Right.

And it’s the same experience.

Netflix was seeking to replace Hollywood for distribution. YouTube is seeking to extinguish it.

Yeah, exactly. And I don’t think YouTube has had any huge impact on Netflix—its pricing power or its ability to retain customers—thus far, but it’s growing the business and monopolizing attention where YouTube—

Yeah, but you draw a line from YouTube five years ago to five years in the future, or 10 years ago to 10 years in the future, and everyone in Hollywood should be way more scared of YouTube than they are of Netflix.

Yeah. And Netflix is clearly thinking of YouTube. Another deal that was announced this week: The New York Times wrote, “The number of video podcasts set to flood Netflix next year is growing. On Tuesday, the streaming giant announced a partnership with iHeartMedia, furthering its long-awaited push into podcasting and yanking a few more shows off YouTube, its rival in the modern entertainment wars. And on Wednesday, Netflix announced a similar deal with Barstool Sports. Under these deals, at least 18 new podcasts will publish their video episodes exclusively on Netflix beginning in early 2026, the company said.”

Then a reaction from Tommy Vietor, host of Pod Save America: “Huge news for Barstool, Big Cat, PFTCommenter, and @Henneyz”—I believe that guy’s name is Hank—“who built by far the biggest and best sports podcast in the country from nothing. This move gets the show in front of a ton of new people and drives actual revenue to grow.”

That’s a reference to Pardon My Take, which is, I think, either number 1 or number 2—the biggest sports podcast in the country. It’s them or Bill Simmons. I think it alternates. It’ll be an interesting experiment to put all these shows on Netflix. I know Pardon My Take’s audience was all pissed off about it. Do you have any reaction to some of the deals that Netflix is doing in that space? All right, and that is the end of the free preview. If you'd like to hear more from Ben and I, there are links to subscribe in the show notes, or you can also go to sharptech.fm. Either option will get you access to a personalized feed that has all the shows we do every week, plus lots more great content from Stratechery and the Stratechery Plus bundle. Check it out, and if you've got feedback, please email us at email@sharptech.fm.