
Maybe… the TikTok thing has finally reached a resolution. Again. Maybe Amazon owes us all some money. Maybe that Neon app from yesterday was a bad idea. OpenAI and Meta make some super interesting strategic AI moves. And, of course, The Weekend Longreads Suggestions.
Loading summary
A
Welcome to the Tech Brew Ride home for Friday, September 26th, 2025. I'm Brian McCullough. Today maybe the TikTok thing has finally reached a resolution again. Maybe Amazon owes some of us some money. Maybe that neon app from yesterday was a bad idea. OpenAI and Meta make some super interesting strategic AI moves and of course the weekend long rate suggestions. Here's what you missed today in the world of tech so I think it's official now. Look, don't at me people. I didn't even talk about this yesterday because it's just a drip, drip, drip. But it feels closer to being official than I don't know, whatever. President Trump has signed an executive order declaring that a proposed deal allowing TikTok to continue operating in the US will be a qualified sale which meets national security concerns. The President says Xi Jinping has approved the TikTok deal and Vice President J.D. vance says the transaction values TikTok US at $14 billion. More on that in a second. According to sources, Oracle, Silver lake and Abu Dhabi's MGX will be the main TikTok USA investors with around a 45% stake. ByteDance will own 19.9% and previous ByteDance investors 35%. So why am I still at this late date hedging just a bit? Because Byte ByteDance for one, hasn't officially acknowledged the deal, at least at the time of this writing. And this morning CNBC is reporting that China and its media have been quiet about all this so far. A notable silence as China can still decide the deal's fate. Chinese social media chatter, for example, has been extremely limited. Quoting the AP Much is still unknown about the actual deal in the works, but Trump said at a White House signing ceremony Thursday that Chinese leader Xi Jinping has agreed to move forward with it. Vice President J.D. vance added that, quote, there was some resistance on the Chinese side, but the fundamental thing that we wanted to accomplish is that we wanted to keep TikTok operating, but we also wanted to make sure that we protected Americans data privacy as required by law. The Chinese Embassy in Washington didn't immediately respond to an Associated Press inquiry seeking confirmation of China's approval. The executive order itself is a declaration by the President that the proposed deal meets the security concerns laid out in that law and it gives all negotiating parties an additional 120 day reprieve in order to finalize. Asked if he wants a US owned TikTok algorithm to suggest more content promoting his Make America Great Again movement, Trump said he'd make it 100% MAGA if he felt he could. But he intends for every philosophy, every policy to be treated right, end quote. Vance said the deal ensures that, quote, American investors will actually control the algorithm that determines the content seen on the social media app. He said more information about the deal will be revealed in the coming weeks. Now, let's come back to that $14 billion figure. As soon as I saw that, I thought basically what MG Siegler lays out here, $14 billion. It's a number so comically low that it would seem more believable if it was what one entity was paying for 10% of the company. But no, everyone is reporting the same thing. $14 billion is the amount being paid for 100% of the new company. Well, technically they're paying less than that because of the just under 20% being retained by ByteDance while current ByteDance investors get to roll stakes over while also putting some new money in. But just how small is $14 billion? ByteDance itself is worth north of $300 billion, with secondaries currently trading in the $400 billion range. Yes, there's more to ByteDance than TikTok, but it's a huge part of the value. And yes, this is just the US portion of the service being sold, but. But it's also undoubtedly the most lucrative part given the importance and usage in the American market. Even 30 to $35 billion feels comically low here. ByteDance's overall revenue was $155 billion last year, and again, TikTok drives a large portion of that and presumably most of the international sales, which were just under $40 billion. Again, just the US business is going to be a large portion of that. But even the most conservative estimates there think it's north of $10 billion a year as a business right now. So these investors are paying something like a 1x multiple, which is absolutely silly. So where does $14 billion come from? Nobody knows, but one guess may be Snap, which is currently valued in the public markets at $14 billion. But Snap famously has monetization challenges where TikTok does not. In no way should that be the comp here. And even Snap has a 2.5x revenue multiple. Other would be comps range from 3x all the way past 10x, and those are public companies. TikTok will be a private and still fast growing. It would not be crazy to value it well north of $100 billion. So, yeah, this is perhaps the most sweetheart of sweetheart deals. Are there risks? Of course, and I'm quite skeptical this will actually work out well. But that doesn't change the current value of the property, which is being comically discounted here. Also, one nugget from the Lauren Hirsch trip Mickle and Emmett Lindner report on the matter for the New York Times. Quote MGX has been discussing the TikTok investment for months, according to two people familiar with the talks. It was one of in the TikTok deal that stood to benefit from the Emirati's data center project, Oracle will be building a data center that will cost about $20 billion in the emirates. Silverlake, an investment company, has a stake in one of the partners in that project, G42, an Emirati AI company. End Quote Yes, TikTok is being sold for less than the cost of a single data center, one of which all of these parties are affiliated with. Funny that. End quote Amazon Yesterday reached a 2 and a half billion dollar settlement with the FTC over deceptive prime practices with a $1 billion civil penalty and a 1 1/2 billion dollar consumer redress. This trial had only been going on for two days, so I guess Amazon figured their goose was likely cooked. Quoting the FT the FTC had alleged Amazon duped customers into signing up for its prime service without their consent and had been sabotaging efforts to undo their subscriptions. A Seattle court on Tuesday began a jury trial in the case, which stemmed from a 2023 complaint brought by the agency's consumer protection unit under former FTC chair Lina Khan. Amazon admitted no wrongdoing in agreeing to settle, the agency added. The company on Thursday said it had always followed the law. The agreed settlement enabled it to, quote, move forward and focus on innovating for customers. FTC chair Andrew Ferguson on Thursday said, quote, the evidence showed that Amazon used sophisticated subscription traps designed to manipulate consumers into enrolling in Prime. The settlement is the latest twist in the high profile legal fight against the tech giant. The FTC earlier this year backtracked on a request to delay the trial, rejecting its own lawyer's claim of resource constraints. I have made it clear since day one that we will commit the resources necessary for this case, ferguson said at the time. The Trump Vance FTC will never back down from taking on big tech, he added, referring to President Donald Trump and Vice President J.D. vance. Under a proposed order filed on Thursday, Amazon must include a clear option to decline the prime service while also making renewal terms more obvious to its more than 200 million customers. The FTC had alleged the company had violated federal laws by using manipulative, coercive or deceptive user interface interface designs known as dark patterns to trick consumers into enrolling in automatically renewing prime subscriptions. An expert witness for the FTC had estimated Amazon duped roughly 40 million customers into signing up for the service, which currently costs $139 annually. Under the terms of the settlement, prime subscribers who use the service for less than three shipments will automatically be eligible for $51 in compensation. Those with less than 10 orders will be able to submit a claim for a redress payment. All civil penalties awarded case will be deposited to the treasury, end quote. So I guess some of you are going to get paid something from Amazon 50 bucks basically, or just your money back for signing up. Well, maybe I'm due a bit more because a US Federal judge has preliminarily approved Anthropic's one and a half billion dollar copyright settlement with authors of books who had those books pirated and then used to train Anthropic models. Might get a couple thousand bucks for that. Fingers crossed hey, quick note that that call recording app Neon that I told you about yesterday, which became the number two social app on the US App Store, has gone dark after apparently exposing users phone numbers, call recordings and transcripts. Quoting TechCrunch, Neon has gone offline, at least for now, after a security flaw allowed anyone to access the phone numbers, call recordings and transcripts of any other user user TechCrunch can now report. TechCrunch discovered this security flaw during a short test of the app on Thursday. We alerted the app's founder, Alex Kayem, who previously did not respond to a request for comment about the app, to the flaw soon after our discovery. Khayyam told TechCrunch later Thursday that he took down the app's servers and began notifying users about pausing the app, but fell short of informing his users about the security lapse. The Neon app stopped functioning soon after we contacted Kaim, curious about the digital.
B
Infrastructure that powers today's biggest tech.
A
Then check out the Interconnected Podcast.
B
It's a new series from Equinix that explores emerging tech megatrends and the infrastructure, making them possible through candid conversations with industry experts, the show breaks down highly complex topics in a way that's easy to understand. In the very first episode covering AI in the medical field, the hosts do a great job helping listeners understand the computational models and interconnected data systems powering today's most promising medical developments. From there, they reveal the connection between this infrastructure and the groundbreaking healthcare breakthroughs we're seeing lately, like rapid diagnostics, personalized medicine, and supercharged drug discovery. Stay tuned into the digital infrastructure powering today's biggest tech trends with interconnected Give it a listen and follow the show on YouTube, Apple, Spotify, or wherever you get your podcasts.
A
A couple of super interesting moves here. First, OpenAI has launched ChatGPT Pulse, a mobile feature for ChatGPT Pro users that delivers daily personalized updates based on their chats, feedback and connected apps. Quoting the Verge, the new mobile feature, called ChatGPT Pulse is only available to pro users for now, ahead of a broader rollout. The personalized research comes your way in the form of topical visual cards you can scan quickly or open for more details. So each day starts with a new focus set of updates per the company that can look like Formula One race updates, daily vocabulary lessons for a language you're learning, menu advice for a dinner you're attending that evening, and more. It's part of OpenAI's big push toward AI agents, an area in which the company and virtually all of its competitors, including Anthropic and Google, are investing a lot of time and resources in. For years, executives at Amazon, Meta, Google, Microsoft and a whole host of other companies have talked on earnings calls about their desire to build successful AI agents for consumers, espousing that one day such tools could do things like create business presentations, book travel, find a restaurant reservation, buy gifts, and more. So far, ChatGPT has been reactive, according to Christina Kaplan, who leads personalization and productivity for ChatGPT. The chatbot answers questions that users submit, putting the onus on the users to figure out what to ask and what they need now, she told the Verge during a Wednesday demo. ChatGPT can keep a pulse proactively on things important to you and offer information, ideas and practical next steps. End quote. The personalized Pulse that Kaplan shared during the demo featured prompts based on her calendar, past requests and chat history, such as what she had on tap for that day, her dietary restrictions, and what she had expressed interest in learning more about. You've got a busy evening. Here's how to flow smoothly from your run into dinner Pulse prompted her, based on her plan, to run before a work team dinner, adding that it had mapped a 40 to 50 minute route ending near Kio rooftop with buffer and backup if timing is tight. It also offered a Kyo rooftop dinner strategy to help prepare a potential order for her dairy free diet. It also prompted her to share adjustments for the future with what's on your mind lately? I'll remember it for tomorrow's update and potential answers like I'm curious about my upcoming plans include and I'd like to stay on top of. Other prompts from Pulse included a daily Core or Pilates routine. Kaplan had asked for a fuel Stop plan for an upcoming birthday hike and a recovery exercise since she had traveled the day before. But what about the training data of it all? In order to most effectively personalize this personalization feature, users will be prompted to share more of their data with ChatGPT. Specifically, if you already have reference past history turned on with ChatGPT, it'll look at your past chat transcripts to inform what it researches on your behalf. And if you've already connected your calendar and email in ChatGPT, you'll be prompted to confirm that you want the chatbot to be able to look at those apps in order to help you prepare for your day. And you'll have to click Accept in order to allow it, according to Kaplan. When asked whether a user's own feedback about their pulse would help improve the feature for others, Kaplan said a user's feedback would only help improve their own pulse the next day. End Quote Now I say this is an interesting move for a couple of reasons. Number one, this is a play to become not only something you use every day, but something you rely on part of your daily routine. The first thing that you check every morning to put my history hat on briefly, one of the reasons say Yahoo is still a business with hundreds of millions of users is because 25 years ago they signed people up in their portal days to be a place where you went to check your email, check the weather, check your stocks. The first thing you checked every morning. This is part everything old is new again and part another way AI could obviate not just the open web, but the need for the open web. Second thing I want to point out is isn't this the platonic ideal we assume Apple wants to someday offer in terms of AI on their devices? And what happens if OpenAI beats them to the punch by what, two, three years? The second interesting move is Vibes, a platform. Meta is rolling out where users can create and share short form AI generated videos, but do so inside the Meta AI app and on the Meta AI website. I say this is interesting because we've seen recently how the Veo video app has driven adoption of Google's Gemini. And I said a long time ago, who has more distribution than Mark Zuckerberg? He can hook people on AI by allowing them to create shareable videos and who would have the best distribution platform for that, quoting TechCrunch according to Meta, as you browse the new feed, you'll see AI generated videos from both creators and other users. Over time, Meta's algorithm will begin to show you personalize content. You have the option to generate a video from scratch or remix a video that you see on your feed before publishing. You can add new visuals, layer in music, and adjust styles. You can then post the video directly to the Vibes feed, DM it to others, or cross post to Instagram and Facebook Stories and Reels Meta's Chief AI Officer Alexander Wang shared in a post that the company has partnered with AI image generators Midjourney and Black Forest Labs for the early versions of Vive. While Meta continues developing its own AI models, the new feed likely won't be welcomed by users, especially since the rise of AI technology has caused social media platforms to become flooded with AI slop. The problem has become so widespread that companies like YouTube are now looking to crack down on the issue. This makes Meta's move particularly puzzling, given that the company said earlier this year that it was tackling unoriginal Facebook content and advised creators that they should focus on authentic storytelling, not short videos offering little value. Well, again, I don't know about that. Zuck has always been unafraid to give users what they say they don't want, but he knows they do want because he can see the actual usage data and listen. You see a lot of AI stuff on social media all over the place. And it's popular. At least the algorithms make it popular. Maybe. But also, the increasing tie up with midjourney is interesting to me. Because here's the thing. Midjourney, unlike some other big AI players, is affordable. It's not that big. No inside information on this. I'm just saying. I wonder if we might not see our first big AI acquisition soon. Gonna squeeze in two long reads for you here real quick. First up, you know how when people talk about AI in medicine, they often mention how AI can reportedly read x rays better than humans can? Well, from the substack works in progress, a look at why AI isn't yet replacing radiologists. This piece claims the AI models underperform in hospital settings. AI use still faces legal hurdles, and the job is much more than simple image recognition. There's still room for humans. But hey, if you think AI is a busted flush, remember there's always quantum computing coming down the hype pike right after it. So from the ft, HSBC says it has tested an IBM quantum computing tool on European bond market data and saw a 34% improvement over traditional methods in predicting order fills. While the results from the HSBC and IBM team are only theoretical and based on historic data. The findings suggest quantum computing could have a significant impact on trading in over the counter markets without a centralized broker. You think things are getting disrupted now? Wait until we unleash both AI and quantum computing on, say, the stock market at the same time. The bonus episode this weekend is an interview with the godfather of New York City VC investing. Jerry Colana was GP of the first ever VC fund raised to invest in the Internet. He started the legendary Flatiron Partners with Fred Wilson. He's been an executive coach to almost any founder you can name for the last 20 years. But also, this is one of the most profound interviews I've ever done about the founder mentality and crucially, the mental health of founders. In fact, I clipped part of that conversation and put it at the front of the episode to highlight it. But listen to the end because the conversation is, as I say, profound. That will be coming to your ears on this feed on Saturday. But if you don't want to wait for that, the bottom link in the show notes today is to a YouTube video of that same conversation that is live on YouTube now. It's worth watching if you want to see what the studio looks like. The new studio that I have access to now that I'm with Morning Brew. More interviews from that studio coming to you in coming weeks. Talk to you on my next Monday.
This episode dives into the breaking news around TikTok’s proposed US sale – claimed to be worth a surprisingly low $14 billion – and what it means for tech, geopolitics, and valuation. Host Brian McCullough scrutinizes the outlandishness of that price, unpacks the major players in the deal, and covers other top tech news, including Amazon’s Prime settlement, Anthropic’s copyright fallout, major AI updates from OpenAI and Meta, and fresh stories at the bleeding edge of tech.
[00:04–09:55]
Announcement: President Trump signed an executive order approving a deal letting TikTok keep operating in the US, calling it a “qualified sale” that addresses national security.
Deal Structure:
Lingering Skepticism:
Security & Algorithm Questions:
Valuation Head-Scratcher:
Connected business interests:
[09:56–11:04]
[~11:05]
[11:05–11:50]
[11:05–end]
[~End]
This episode is dense with breaking tech news and sharp analysis. Brian’s skepticism about the TikTok deal’s value, his detailed breakdowns (especially around AI innovation and big-tech regulation), and his eye for the big-picture make it an essential listen (or read!) for anyone following the intersection of tech, politics, and new digital economies.