(Preview) OpenAI Declares a ‘Code Red,’ Alan Dye Leaves Apple for Meta, Questions on Tranium 3, Substack, and F1

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 Thompson. Ben, how are you doing?

Doing okay, Andrew. I've been very angsty for 2 weeks. I'm still angsty. We're going to explore the angst on this podcast, but that is the theme of the day.

I can't wait to find out the source of the angst. Is it Giannis-related, or is it tech-related?

Oh, no. The Giannis thing, whatever. It's funny you asked me—was it literally the last podcast? I'm like, “Oh, I hope he stays forever,” but—

Yeah.

I guess—

Great timing on that take. Perfect.

When the news breaks that he wants out, I think my heart was already ready. I'm not too impacted. It is what it is. At least we won 1, you know?

Yeah.

That's been my pinned post for a while: “At least we won 1,” with a picture of the court after 2021.

Once again, that's not a feeling that I anticipate experiencing for at least another 10 or 20 years in Washington. I'm happy that you made it to the mountaintop. As somebody who's watched a lot of the Bucks over the last 2 or 3 years, I, too, have made peace with Giannis moving on. It may be for the best, all things considered. I also watched the Bucks lose to my Washington Wizards. The Wizards may have put the nail in the coffin of the Giannis era—

I mean, it was—

—in Milwaukee.

It was pretty rough. I was supposed to be at this game. We got a foot of snow over the weekend, which impacted travel plans, to say the least. I will say, I enjoyed the snow. I enjoyed shoveling. It's 1 degree out today, but it's not too bad. I think I could do this winter thing, and I'm just surprising myself.

Okay.

I'm reacquainting myself with Madison winters.

Yeah. Reacquainting yourself with Milwaukee winters—not bad.

Madison. Madison. Geez, stop it.

I purposely conflate the 2 to mess with you offline, but to make that mistake on the show—yes, Madison. I'm happy that you have seasons again. That's one of the things that people say: “I miss seasons,” and people in—

I know, and I always—

—Los Angeles make fun of it.

—I always laugh at that. It sounds like a rationalization, but I have to say, having seasons has been pretty great.

It's great.

Now, again—

Yeah.

—ask me in February, when it's snowing more. I'm not sure how I'll feel then.

February? More like March and April out there—

Yeah, good point.

—in Wisconsin.

Good point.

That's when it does get pretty rough.

1. OpenAI Enters Code Red

As for the show today, you wrote about OpenAI on Monday this week, and mere hours after that story was published, The Information reported that Sam Altman had told OpenAI employees he was declaring a “code red” to marshal more resources to improve ChatGPT as threats mount from Google and other AI competitors. As a result, OpenAI plans to delay other initiatives, including advertising, according to Altman. “We are at a critical time for ChatGPT,” he said.

So, Ben, as OpenAI enters the code-red era, it seemed like on Monday, with your article, you wanted to come out and give a full-throated “relax” to all the OpenAI skeptics, but you couldn't quite get there in that article.

I'm glad that came across. That's why I've been angsty. I told you before Thanksgiving, “People need to relax here. I need to write an article.” I thought about it all over Thanksgiving break. I got to Sunday and thought, “I'm going to write an article. I need to write an article. It's time to write an article. I need to write about this.”

The article did not publish until 9:00 a.m. on Monday, and that was because I was up all night bleeding this one out. The reason I was bleeding it out was a lack of conviction.

I am really concerned, and the reason I'm concerned is that, by all rights, they ought to have secured the consumer space.

Yeah.

I've been on 2 points with OpenAI again and again. Number 1, they need to focus on the consumer space. I've been worried about the API being a distraction, and I kind of came around on that. I get it, whatever it might be.

You've softened your stance as far as OpenAI's focus is concerned at various points over the last 12 months.

Which maybe was a mistake.

Yeah.

Don't back down.

2. OpenAI Needs An Ad Flywheel

But, number 2, the whole ad thing: I've been on and on about how they need to launch ads. The problem in the long run is that the power of being an aggregator is not just that you have consumers. It's that you have a flywheel that gives you increasing, expanding power in the marketplace that is impossible to compete with.

If you're Google in search, it's not just that you have everyone searching on Google. It's that you have ads that people are accustomed to and are fine with. You have a massive universe of advertisers that are used to going there. They buy ads to show all those users. What that does is give you the money to invest to make it better.

The thing about ads is that there's a real spectrum. A lot of ads really suck.

Yeah.

The ads that Google does are, again, a little exploitative because they're just stealing the top spot in many respects. The fact that you have to bid on your own name on Google is really annoying.

Pretty absurd, yep.

Facebook and Meta are the better example. Half the time I go on Instagram just to see what ads I'm going to get and what new crap I can buy, because it's all awesome. It's great.

You don't get there overnight. You start bad and you get better. What really gives me pause right now is that OpenAI is facing this challenge from Google, where Gemini is probably better. Even if OpenAI comes out with a better model—which it probably will soon—there's a whole other discussion there about choices Google made with TPUs versus maybe some challenges that are happening with NVIDIA GPUs, things along those lines.

But the time to be crappy and diminish the user experience was when you had a huge lead. You're facing this situation where, suppose OpenAI decides it actually does need to monetize—which clearly it does. It can't just keep losing $15 or $20 billion, or whatever it is, forever.

But it's going to do so in a context where it's going to risk losing its audience, because Google doesn't need to monetize. It's monetizing on search just fine, and it can let Gemini be free. Imagine a situation where it's not just that Gemini is arguably better, but Gemini is the nice, ad-free experience. It's cleaner.

I feel frustrated because, on the one hand, I strongly believe in the advantage they had with the customer base, and it's hard to move people. But this is like Uber back in the day. I went to battle about Uber with skeptics.

That was a case where you did have conviction 10 years ago with Uber, when there were armies of people saying, “Uber's losing money. Uber's going to fail,” and so on. You were on the other side of all those arguments.

It was really frustrating in 2017, when all those scandals happened, because Lyft was about to go out of business and Uber was going to take the market. Then suddenly Uber faced all these scandals and all these things went down, Lyft got massive infusions of money, and it just basically—

And so I'm like, “So you're saying there's a chance?” I think I was right, but can I actually know if I was right? That's how I feel right now.

I think I was right about OpenAI being positioned to own the consumer market. But because they didn't act quickly enough and didn't secure their position—and their position isn't just about acquiring users, it's not just about having good products, it's about having a sustainable flywheel of revenue generation that actually locks you in more—they might lose.

If they lose, does that mean the theory was wrong? Does that mean they were too slow? I don't know. I'm very angsty. That's the theme. I'm still angsty.

I don't feel great about my article. I'm glad the angst came through, because that's the best I could say: I don't know. I would say “I don't know” is an acceptable answer at this point and, frankly, kind of an exciting answer as we wait to see how some of this plays out.

six to 12 months.

Dude, that is, I'm already thinking about what I'm going to write next. I came back the next day weighing out the bear case very explicitly in the daily update.

Mm-hmm.

But I'm wanting to write about this more, and there was a meta angle, which is that this is very exciting. This is very thrilling. It's great for me. Yeah, it's great for us.

I mean, it sucks if I'm wrong, but I always say that being wrong is actually one of the most exciting times, because you have to step back and examine what assumptions you had about the world and how it works.

Mm-hmm.

And what went wrong. That's a chance to understand the world better. It does entail a little bit of embarrassment along the way, but it is what it is.

Well, additionally, it goes without saying, but I really appreciate OpenAI for continuing to keep us entertained—

Yes.

—month in and month out. Number one on the take, number one on the take index. You—

Number one—

If people didn't read—

Number one with a bullet.

If people—

The takability rankings, yes. I know. Andrew came out with the takability rankings, which was very long—

Yeah.

—very enjoyable, and when you read it, it's all a long preamble for Andrew to express his appreciation for OpenAI. As far as takability, OpenAI is in a universe of its own in 2025. In October, we had Sora and $1.2 trillion in spending commitments, and then in November, they're talking about partnering with the government on loans, this, that, and the other thing. Now we have Code Red. I can't wait for 2026's agenda.

It's amazing. It's going to be great.

But in terms of what you're talking about, one of the things I come back to as a normie—and it's a point that you raised in your article on Monday—is that, for all the hand-wringing over competition with Google or Anthropic or anyone else, ChatGPT is the one AI product that everyone I know uses and likes. There are all kinds of opportunities for everyone else in AI, but ChatGPT is the one that's obviously succeeding right now.

On the other hand, I do think that the Code Red impulse is probably healthy. It's a hilarious way to frame this moment for a CEO.

3. ChatGPT Becomes The Main Thing

Dude, I've been begging them for years to focus on ChatGPT. Stop wasting your time on all this other stuff.

Well, that's the thing: they're trying to be an infrastructure and API business. They're building hardware. They're doing e-commerce deals. They're talking about designing chips. All of that is great and reasonable, but ChatGPT is the linchpin for everything else they want to do.

So, to the extent that this does delay the ad initiative, do you think that's healthy at this moment in time, given the competition from Google?

I mean, that's fine, because they have a real problem here. The initial introduction of ads, in particular, is going to diminish the user experience. That's the last thing they need to have happen right now.

By the way, as long as we're complaining, I'm very excited about the shared chat thing.

Mm-hmm.

It does not work on desktop.

Yeah.

The chats are just gone.

So what's the point, really, if you're trying to improve workflows?

No, it speaks to the fact that they're getting sloppy. One of the things they delivered on again and again was how their apps were always better, and they had native apps on the platform. This is actually the reason why people don't do native apps, because it's harder to keep them up to date and handle all these bits and pieces. But I always thought that set them apart.

Mm.

They seemed to care about the user experience of ChatGPT in a way that was tangibly different from anybody else.

Yeah.

And the whole shared chat thing, which I think is, number one, super useful, and number two, has the promise of a network effect: you start getting people used to using it with other people. Now it's not just, “Oh, I try Gemini.” It's, “Oh, I have to get my friends to blah, blah, blah.”

Mm-hmm.

That's just a real barrier, and they half-ass it. They have thousands of employees at this point. What are people working on? They're working on a lot of stuff, and they're losing sight of the main thing. The main thing needs to be the main thing.

I gave them grace for it for a while because I think this transition from being a research lab to being a consumer tech company was a difficult one. It was a painful one. People were going to leave. There was going to be drama because that's not what people—people didn't sign up to be Facebook.

But you are presented with a Facebook opportunity. You need to take it. They kind of seemed to take it for 2 months, and then they're like, “Oh, actually, we're going to be the—”

We're going to be Apple.

—company of the universe.

Yeah.

4. OpenAI Takes On Apple

Yeah, no, that's actually another great example. Let's go in on that, the whole Johnny Ive thing. OpenAI was Apple's partner. They should be Apple's partner. They are a horizontal services company that needs to have the best experience everywhere. That's what you need to be if you're going to be an aggregator.

They basically signaled, “Oh, in 3 years, Apple, we're coming for you.” All you're doing is pushing Apple into the arms of Google, and rationally so.

This is a mistake that Google made with Android back in the day, where this VP from Microsoft, Vik Gundotra or something, had gotten this very “We're taking on Apple” posture. They'd open their Google I/O keynotes with all this talk about how they were open and Apple was closed.

Great impression. Love it.

Well, it was also a very exciting time.

Yeah.

That's really when Structure got started, when everyone—

Yeah.

—was saying that Apple was doomed, and I'm like, “Eh, I think they're going to be okay.” I felt a lot better about that one—“Eh, they're going to be okay”—than I do about this one right now, to say the least.

Yeah. Well, I mean, related to the ads question, we got this note from Paul. He says:

“It's been interesting hearing Ben's enthusiasm about the urgency to place ads on ChatGPT. While I understand and agree with the long-term necessity of ads in the product, I think he may be overlooking the timing and positioning of ChatGPT compared to his examples of YouTube and Facebook.

“By the time YouTube and Facebook started really leaning into ads, they already had dominant positioning in their market. There was nowhere for users to turn to as YouTube started turning the enshittification dial, and people's lives had already been entrenched in the Meta products by the time ads started taking up most of the feed.

“Meanwhile, if ChatGPT were to start layering in ads at the cost of a clean user experience, there are no real barriers to a user who wants to hop to a cleaner, ad-free alternative, such as Gemini or Anthropic. Instead, it seems like there's a standoff across all the AI companies. Even though I'm sure they would all love to begin adding ads, nobody wants to be the first and risk the PR drama and potential loss of users.

“But I'm not sure it's as simple as ads now for ChatGPT.”

And related, Aaron says:

“Google can, in theory, delay an ad-integrated Gemini almost indefinitely. If there's a strong negative reaction to OpenAI's ad rollout, is that Google's opportunity to make a power play for OpenAI's enormous consumer user base?”

So that's where we are this week.

I mean, I completely agree. I completely agree. My real angst is the fact that they could have put the world's crappiest ads in ChatGPT—

The time to do it was 18 months ago.

No, more than that. I would say I was writing about this in early 2023. If they started then with crappy ads, the ads would be a lot better now.

There is, by the way, a bit where Google is not going for the kill. There is free Gemini. The better Gemini, you do have to pay. The new Nano Banana Pro is amazing, but it's only on the Pro levels. They're a public company that's worried about their margins.

They raised prices for the Gemini API, for the Gemini 3 API, probably because it's much more expensive to serve. But the point is, they're acting with a certain degree of rationality here that suggests they have their own constraints in this regard, as you would expect. We're still waiting to see to what extent they work through that over time.

But this is why it's hard to write about, because it's not very productive to say, “You should have listened to me 2 years ago.” That's really how I feel right now, and it's very annoying.

But reading The Information piece, your reaction on Tuesday was not, “Okay, you guys are crazy.”

No, it's positive.

It’s positive.

Yeah.

Yeah, no, you have to secure the position of ChatGPT. My reaction was, “Maybe I shouldn’t have softened on the API stuff.” If you’re consistently saying, “We’re not giving the best possible experience because we’re resource-constrained,” why are your resources going to other things?

Spread too thin. Yeah, well, and the other reason to implement ads 2 years ago is that you’re training users to experience the product with ads in it, so that there’s not this whiplash effect that would potentially—

Right, and the ads—the ads will get better.

Drive people to Gemini.

5. Ads Align The Incentives

I think that I’m really annoyed by the enshittification aspect, because the reality is that ads, number 1, make a product much more widely accessible.

Mm-hmm.

And I think it’s really easy to sit here—you, me, anyone who’s paying $15 a month to listen to us talk—is extraordinarily privileged, okay? Number 2, most people are not rational about this. How long did it take you to subscribe? I have a friend that I was talking to—it just ages, right? People don’t pay for this stuff even if it is a huge benefit for them. I just skipped over you. I didn’t let you answer.

I know. It’s funny. I was relying on you while I was in law school, but I think it took me 2 or 3 years before I actually subscribed, and you were a close friend at that point. So, yeah, it takes people a little while to come around.

Did you ever subscribe to YouTube Premium yet?

I still have not subscribed to YouTube Premium.

See? This is—

I do that just to spite you because it makes you so upset, and it really doesn’t make that much of a difference in my life.

It doesn’t upset me personally. It upsets me on your behalf. I feel bad—

Mm.

For you that you’re living this life. This is another thing, by the way. I think the YouTube example came up on Sharp Tech, and I referenced it this week in the bear case update that I did.

Mm-hmm.

You have this real marginal-cost problem, which is, by and large, web services have gotten so cheap that anything users do, no one really worries about it. Yes, serving a Google search costs a lot of money, but Google is not concerned about people doing one more search.

We’re still in an era of AI where that is a tangible concern. You have things like rate limits, which, by the way, are an enshittification of the user experience. If you get into ChatGPT, and you’re using it, and you hit a rate limit, that sucks.

That is a good point.

If you could use it—

Yeah.

—without hitting a rate limit and there were some ads there, that’s actually better.

It would enshittify the YouTube experience if there were buffering half the time you tried to watch YouTube videos.

Right. No, exactly.

Sure.

Regular people all over the world can just use YouTube. The beautiful thing about ads is that the more YouTube videos they watch, the more it costs Google to serve. But the more YouTube videos you watch, the more ads you see—

The more they’re making.

—and the more money they make off of you, right? It’s the same thing with AI. If you had a productive ad experience, longer usage—more usage—is more inventory.

Yeah.

This is why I’m passionate about advertising being an essential business model. It aligns incentives in so many different ways. Yes, I’m not saying it’s perfect. It can go wrong with sensationalism and all those sorts of things.

But, by and large, there is a very healthy alignment, and the problem with charging for things is price elasticity. To make more money to cover your costs, you’re going to reduce the number of people who are willing to pay—economics 101.

The amazing thing about the advertising model is: Who bears the increase in price? Advertisers do. Why do they pay more? Because it’s working. They’re getting results. And—

Yeah.

So you’re shielding that—you’re cutting that off from the user experience. This is why you can grow to cover the whole world. This is why Facebook and Google are so powerful, because it’s not just that they’ve aggregated the users. They’ve also aggregated the advertisers, and they’ve offloaded all of the price rises—raises, whatever it is—to the advertisers. And, by the way, they don’t explicitly raise prices.

Mm-hmm.

One of the things your people in the anti-monopoly movement can’t seem to understand is that prices go up and down as a function of supply and demand. It’s a pure market. The reason prices go up on Facebook is because the ad product works so well; more and more people want to buy ads there. It’s a—

Right.

—beautiful business. It’s the most powerful business possible, and it was sitting there. It was sitting there for OpenAI, and they’re just fucking around doing all this other shit.

Well, lest we cry over spilled milk, the game’s not over yet, and OpenAI pausing the ad integration is probably healthy for the short-term prospects, and—

But it speaks to—

They still have 800 million weekly users.

They need to stop doing what they had to do to get back in a position to do what they need to do.

You’ve convinced me. They should’ve incorporated ads 2 years ago.

No, if anything, I regret not being more forceful about it. I like to think I was pretty forceful about it. But—

Yeah.

6. Google Could Still Win

Well, as far as going forward, I gotta say I’m pretty haunted by the digression that you had on Dithering, where you mentioned that Google was not first in search, not first in smartphones, and not first in browsers or email, and came to dominate in each of those sectors.

Yep.

It does feel kind of ominous when you look—

It—

—at the history here.

Yeah. No, so there’s been this great series of ads that Canal+, the French broadcaster, has put out for the last couple of F1 races.

Oh, yeah. Canal+.

Yeah.

Yep.

They did one with the Jaws theme. They did one—was it Terminator? They did a Star Wars one.

And Star Wars. Yeah.

The implication being it should be this easy, cute little McLaren or Orlando winning the title. And then here comes Max—duh, duh, duh, duh.

Max—the shark—

Like, Google is Max—

—coming through the tunnel. That’s right.

—in these commercials. And it’s just like, no. I’m sitting here as someone who believed that OpenAI was going to win the market, and I just hear Google. Google has always been my nemesis. I consistently underrate them.

Uh-huh.

And I’m like, “Well, looks like maybe I did it again.” It keeps me humble, I guess. There you go.

Indeed. Well, Efficient Market Hype on Twitter tweeted this week, “Gemini winning the AI race is super bearish for the market if you think about it.” I read that and thought to myself it probably is bad news—

Yeah.

—for everybody else.

For sure. Yep.

I mentioned this on Sharp China about a month ago, but if Google wins, they could effectively play the role of China in the AI race, where Google Search is like the CCP subsidizing the AI business and allowing Gemini to just run at a loss, destroying everybody’s ability to make a profit and driving all competition out of business. Is that a reasonable concern to harbor?

Yeah. Yes, it is.

Okay, great. Well, we’ll see. We'll see whether OpenAI... 'Cause look, you weren’t completely bearish on OpenAI either. 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.