China Blocks Meta’s $2B Manus Deal, Meta Bets on Space-Based Solar | Diet TBPN

Summary

TBPN: "China Blocks Meta’s $2B Manus Deal, Meta Bets on Space-Based Solar"

Episode Date: April 27, 2026
Podcast Hosts: John Coogan & Jordi Hays (plus Tyler, Dylan, Phil, Rhys featured in this Diet TBPN highlight episode)

---

Overview

This Diet TBPN episode delivers a fast-paced, 30-minute breakdown of the day’s most significant headlines in tech, focusing on China’s unprecedented blocking of Meta’s $2 billion acquisition of AI platform Manus, and Meta’s bold space-based solar power initiative. The hosts also explore broader themes like the global AI talent war, psychological studies on mindset, automation’s impact on jobs, advances in generative UI, and inject humor and pop culture into the conversation.

---

Key Discussion Points & Insights

1. Elon Musk’s $134 Billion Lawsuit Against Sam Altman & OpenAI

• [00:00–01:00] Tyler introduces the major news: Elon Musk’s case goes to court in Oakland, alleging OpenAI abandoned its nonprofit roots in favor of profit, with Microsoft involved in jury selection.
• Quotes & Color:
  • “Musk is alleging that OpenAI abandoned its original nonprofit mission and became a for-profit entity focused on maximizing profit.” — Tyler [00:10]
• OpenAI’s new cloud flexibility agreement: now serving products across AWS, Google, etc., giving them much more flexibility with chips and hosting.

2. China Blocks Meta’s $2B Acquisition of Manus

• [01:25–02:24, 22:24–28:15] The centerpiece story: Chinese regulators intervened to unwind Meta’s $2B purchase of AI app Manus. Even though Manus had moved to Singapore, China used regulatory leverage to halt the deal, signaling a shift in approach to tech M&A.
• Key Facts & Reactions:
  • Some Manus team members were briefly detained in China.
  • The Manus team wanted the deal; some have deep family ties in China—a clear leverage point for the government.
  • Beijing’s order: unwind the acquisition completely, return funds, re-register ownership, and cease all Meta-using Manus algorithm activity. Threats include penalties, business limits, and possible criminal charges if not complied.
  • Quote: “The gesture was pretty harsh and it carries a strong intention to stop follow on deals like Manus. In reality, it's hard to unwind a done deal.” — Dylan [22:39]
• Analysis:
  • China’s move may serve as a warning for future transactions, showing the power to intervene even when a company “goes overseas.”
  • Broader Impact: It deters AI founders from starting in China, fearing future government intervention and lack of access to Western capital.
  • Quote: “This is what happens when you regulate by fiat rather than by rule of law. Ultimately, this is a much larger defeat for the Chinese AI ecosystem than for the United States interest.” — Tyler, paraphrased from Delian’s reaction [27:00]
  • Chinese leverage over Meta: a significant share of Meta’s ad revenue originates from Chinese companies; further escalation could hurt Meta directly.

3. Meta’s Space-Based Solar Power Initiative

• [02:18–02:53, 10:41–12:49] Meta partners with Overview Energy to beam up to 1 GW of solar power from orbit to Earth, plus deploys ultra-long-duration batteries with Noon Energy.
• Details & Insights:
  • Goal: 24/7 clean energy for data centers, overcoming the intermittency problem of Earth-based solar.
  • Partner companies include Overview Energy and Noon Energy (100-hour batteries).
  • Discussion about related ventures in El Segundo (Reflex Orbital) and the concept of beaming energy via mirrors or lasers.
  • Quote: “All of these projects are incredibly difficult to math out and require a lot of different things to go well... But if it's working for them and it winds up working for Meta, you can imagine that there are going to be lots and lots of buyers because energy is, of course, in short supply.” — Tyler [11:21]
  • Implications: If 24/7 “always-on sunlight” is possible, many more solar projects become viable, especially for energy-hungry data centers.
  • Intermittency and the need for constant uptime are critical issues, with battery advances playing a key role.

4. Diet Coke Deep Dives & Virality Culture

• [02:53–07:48]
  • Unexpectedly, the hosts veer into the viral world of Diet Coke Instagram Reels—$4,000 room-stocked fridges, taste differences between packaging, and inefficiency in soda storage.
  • Notable Quote:
    • “They just came out with this new study that compared people that drink Diet Coke versus people that drink Coke Zero. And what it actually found was that people that drink Coke Zero are idiots. And then people that drink Diet Coke are actually Sigma chads...” — Rhys [03:19]
    • Full-on banter, jokes about “Diet Coke app,” and “jester maxing,” illustrating the lighter side of tech podcasting.
  • Transitioned to a quick explainer on semiconductors by Tyler: “A semiconductor is like geranium or silicon. The current can be turned on and off, so it's semiconductive... that's why they are semiconductors.” [06:12]

5. Placebo Sleep Study & 'Golden Retriever Mindset'

• [07:48–10:41]
  • The team walks through the trending “placebo sleep” study: simply believing you slept well can boost cognitive performance, regardless of reality.
  • Quote: “Just having a delusional golden retriever mindset measurably changes outcomes and physiology. Sleep badly, convince yourself you're well rested; stressful day, convince yourself it's fuel.” — Dylan [08:08]
  • Cultural tie-ins: comparison to movies like Yes Man (Jim Carrey) and Air Bud (golden retriever actual vs metaphorical).
  • Emphasis on the power of positive self-delusion and mindset shift.

6. Impact of AI & Automation on Global Employment

• [13:17–17:51]
  • Explores AI’s impact on jobs, especially in classic outsourcing regions (Philippines, India).
  • Despite automation fears, Filipino employment is up 4% in 2025, suggesting AI’s job impacts lag or are offset by broader trends.
  • Call center job decline in the US predates ChatGPT, likely due to outsourcing, not just automation.
  • Quote: “AI might play the similar role in the sense that... there is some sort of onboarding cost to moving from a US based to a Filipino call center. That's taken a decade to actually decline by, you know, not even half.” — Tyler [16:03]
  • Shoutout to Poland’s economic boom—income on track to outpace the UK—a symbol of shifting economic fortunes.

7. AI Progress: Weights vs. Inference Debate

• [17:51–19:11]
  • Discussion of Noam Brown’s view: model weights (the trained parameters) may become less critical than inference capacity (the ability to run AI at scale).
  • Quote: “Model weights become relatively less important as inference becomes more important, which means securing weights still matters, but securing inference capacity becomes a strategic advantage.” — Summarized by Tyler [17:51]
  • Geopolitical implications: as with nuclear material, owning the “reactor” (data centers for inference) may matter more than owning the “recipe” (weights).

8. Outsourcing Life to AI, the ‘Token Maxers’

• [19:11–21:24]
  • Segue to a Semafor profile of a tech entrepreneur who lets his personal AI run much of his life, including responding to emails and managing parenting.
  • “I didn’t ask it to help me. I asked it to be me.” — Semafor profile [19:11]
  • Tyler and Dylan reflect on how much of daily life is already “mediated by technology”—alarms, maps, appointment calendars.
  • Raises idea of a 'barbell world': some go all-in on tech and AI, while others focus on more traditional, tangible experiences.

9. Generative UI & the Future of Digital Interfaces

• [28:44–30:28]
  • Showcases advances in generative UI: ChatGPT and similar models can now render custom, visually-rich interfaces (maps, dashboards) on the fly.
  • Quote: “You're getting closer and closer to this generative on the fly ui... when you get something that perfectly sums up exactly what you're looking for on the fly.” — Tyler [29:00]
  • Ben Thompson’s bull case: The magic is in dynamic interfaces, reduced chrome (UI clutter), and experiences that adapt instantly to user context.
  • Ends on a humorous note critiquing an AI-generated map: “One Tower Golden Gate is a crime.” — Ryan [30:19]

---

Notable Quotes & Memorable Moments

• “Did they get enough of the company to Singapore in order to clear the hurdles that require the acquisition to go through? Then some members of the Manus team were detained briefly in China on business.”
  — Tyler, on challenges maneuvering past Chinese regulation [00:53]
• “If you are so successful that a non-Chinese firm tries to acquire you for billions of dollars, the Chinese government will lure you back to Beijing, ban you from leaving the country, and take your profits by canceling the acquisition.”
  — Paraphrased summary of Delian’s take [27:40]
• “Sleep badly, convince yourself you're well rested, stressful day, convince yourself it's fuel.”
  — Dylan, on placebo mindset [08:08]
• “I didn’t ask it to help me. I asked it to be me.”
  — Semafor’s AI power user [19:11]
• “China does have a good amount of leverage given that like tens of billions of Meta ad spend originates from Chinese companies.”
  — Dylan [28:15]
• “You're getting closer and closer to this generative on the fly UI... when you get something that perfectly sums up exactly what you're looking for on the fly.”
  — Tyler [29:00]

---

Timestamps for Key Segments

• [00:00–01:00] – Headlines rundown: Musk vs. OpenAI, OpenAI cross-cloud moves
• [01:25–02:24; 22:24–28:15] – China blocks Meta's acquisition of Manus, reaction and analysis
• [02:18–12:49] – Meta’s space-based solar power project explained
• [02:53–07:48] – Diet Coke Instagram reels and “Sigma Chad” banter
• [07:48–10:41] – Placebo sleep, mindset science, and pop culture tie-ins
• [13:17–17:51] – AI’s impact on global employment, BPO/outsourcing
• [17:51–19:11] – Model weights vs. inference capacity, AI safety/security
• [19:11–21:24] – Personal AI assistants and the “token maxers”
• [28:44–30:28] – Advances in generative UI, dynamic interfaces

---

Tone, Language & Style

The episode blends deep-dive tech analysis with relaxed, humorous banter, moving fluidly from hard-hitting geopolitical news to viral Instagram memes and philosophical musings. The crew channels classic Bay Area energy—skeptical, curious, irreverent, and occasionally self-deprecating—making the podcast approachable for both tech insiders and interested observers.

---

Summary for the Uninitiated

If you missed the episode, here's what you need to know: Meta’s $2B bet on a promising AI startup was derailed by Chinese regulators in a rare, forceful move, sparking global debate about where AI companies should be built and who really controls global tech innovation. Meta, not standing still, doubled down on cutting-edge space solar projects to fuel its data centers, demonstrating their ambition (and the tech sector’s appetite) for both AI and energy innovation. The hosts also handle viral internet rabbit holes and psychological hacks with the same curiosity as international headlines, giving listeners a full-spectrum view of life at tech’s cutting edge.

Transcript

[Tyler]

We'll take you through some quick headlines of what's going on in the news today. Of course, the Elon Musk for the production team. Yes, we background update. We have a new screen in the background, much better contrast, much brighter. We're excited to explore this. The production team's stoked. There's four key stories going on in the news today. The first is Elon Musk's $134 billion lawsuit against Sam Altman goes to court today in Oakland. Musk is alleging that OpenAI abandoned its original nonprofit mission and became a for profit entity focused on maximizing profit. Microsoft jury selection is going on today. There are a number of reporters on the ground and we'll be checking in on the progress there.

[Dylan]

Already there was some reporting that said certain people were saying it would be difficult for me to be unbiased here because of my general dislike for basically everything going on.

[Tyler]

Sam and Greg were spotted on scene. Elon has not been spotted yet. There's Also other Microsoft OpenAI news. They changed their partnership. OpenAI can now serve all of its products to customers across any cloud provider. Andy Jassy had a post about it saying it was a very interesting update. Obviously he's excited to vend OpenAI technology through AWS, which they also have a partnership. And so in other words, OpenAI can potentially use Google, TPUs, Amazon, Trainium, other chips are on the table now. There's a lot more flexibility that comes from that. Over in Metta World, China has blocked Meta's $2 billion acquisition of artificial intelligence platform Manus after regulators reviewed whether deal violated Beijing's investment rules. This was something that went back and forth. Did they get enough of the company to Singapore in order to clear the hurdles that require the acquisition to go through? Then some members of the Manus team were detained briefly in China on business. There were some issues there and one

[Dylan]

of the challenges is going to be clearly the Manus team wanted to do this deal. They wouldn't have signed up for it. It was a great outcome for the team. They were excited to build with Meta. But one of the challenges is you have all these different team members, many of which were born and raised in China and they still have family members and people back in China that they care a lot about. So that's going to be, I would assume that's going to be a leverage point from the ccp.

[Tyler]

And the last story also from Meta, is that they are planning to use solar power from space at night, beamed from space. They're partnering with Overview Energy, a different company than the other solar space we talked about. Space they've been in. They're planning to beam up to 1 GW of space solar power from orbit to Earth for around the clock power production. They're also deploying 1 gigawatt of ultra long duration storage batteries with noon energy. So some exciting deals coming out of Menlo Park. We gotta start with these Diet Coke videos. Cause I fell down a rabbit hole of Diet Coke Instagram reels.

[Dylan]

I'm surprised you weren't in this.

[Tyler]

Yeah, it took them a long time to figure out that I drink three Diet Cokes a day. And I love diet co. These videos have brought me a lot of joy. So we can play one. There's been a new study about Diet Coke versus Coke Zero. And we can pull it up in just a minute.

[Rhys]

They just came out with this new study that compared people that drink Diet Coke versus people that drink Coke Zero. And what it actually found was that people that drink Coke Zero are idiots. And then people that drink Diet Coke are actually Sigma chads that are way better than everyone.

[Tyler]

Thank you. Thank you. We needed that. Wait, play the restock video. There's a restock video that is. It claims to be $4,000 of Diet Coke stored all over this person's massive house.

[Dylan]

I think it's about 200.

[Tyler]

Let's count it up. Let's count it up. Let's pull up the restock video. And you tell me, is this $4,000 worth of Diet Coke? Okay, that's maybe $20 of Diet Coke. This is maybe $10 of Diet Coke. Here's another 10. We're what, under $100 still of Diet Coke, I would imagine.

[Dylan]

I'm just so curious. When do they opt for the plastic bottle versus the can?

[Tyler]

It is odd. Some people have preferences. I saw again on Instagram in my Diet Coke Deep dive, someone who insists that the 16 ounce aluminum can of Diet Coke tastes better than the 12 ounce aluminum can. That's the level of Diet Coke efficiency.

[Phil]

Also, check out the organizational inefficiency.

[Tyler]

Yeah. What's going on here? What's going on here?

[Phil]

She's not just putting up frauding in the front.

[Dylan]

Yeah, frauding, frauding.

[Tyler]

Yeah. Because you could stack the Diet Coke much deeper, but once you're that far in the real you call it, Instagram

[Dylan]

is just the Diet Coke app for you now.

[Tyler]

Yes. Well, I'm so deep that I'm getting Diet Coke vibe reels. Let's pull this one up because this one is electric. This is like Peak content. There we go. This one. Adding an accident.

[Phil]

Cause this must be heaven.

[Tyler]

I gotta testify. Come up in the spot. This is like John. Just the fact that somebody took so long to edit this together is.

[Dylan]

That looks like something you could. One shot.

[Tyler]

I mean it's like in capcut and you'd need to like choose the words and place them and add the features. Like you should be able to puppeteer that with an agent. But I don't know of any agents that are really there on the video editing front. Certainly like the next chip to fall, the next opportunity.

[Dylan]

Anyway, Rhys says Chad might mog, but when the jester performs, even the king sits to listen.

[Tyler]

True.

[Dylan]

And this is incredibly true. We have a friend of the show that is 100% a jester and the closest thing we have to kings in this industry will pick up his call and listen to whatever he has to say, even if he's Jester maxing.

[Tyler]

That's true.

[Dylan]

Based 16C. Got to be honest, bro. Oh, I think it deleted,

[Tyler]

but said gotta be honest, bro. I have no idea what a semiconductor is. Do you know why they call them semiconductors?

[Phil]

That made me laugh so hard.

[Tyler]

Do you know why they call them semiconductors?

[Dylan]

Me, I do not.

[Tyler]

So full on conductors like copper, electricity flows through it constantly. A semiconductor is like geranium or silicon. The current can be turned on and off, so it's semiconductive. And that started the computing boom because you can effectively store ones and zeros in it. A little more complicated than that, but that's like the very high level version for why they are semiconductors. That's not why they call it semi analysis. It's because Dylan just says he doesn't want to do full on analysis. He only wants to do semi analysis. I think his analysis is totally full on analysis. But he decided to go with semi analysis anyway. Imagine genuinely believing that the entire human race was going to be wiped out in the next year. And then you just kind of aimlessly argue about it on Twitter. That is a weird, weird phenomenon that's going on. Oh, Phil is looking for a large gong in the Bay Area. If you are in possession of a gong that is over 30 inches, give him a call. He's in the market.

[Dylan]

Is a 30 inch gong a large gong though?

[Tyler]

How big is our gong? I think it's 42. 42. We're, we're around there. We're around there. Anyway. Do you know where we got this gong? Gongsunlimited.com gongsunlimited.com Phil, you have your answer.

[Dylan]

Okay, but here's.

[Tyler]

I couldn't reply to this on X, but I chose to say it to Monday.

[Dylan]

Okay, here's the thing.

[Tyler]

I'll tell it to you in person. Yeah.

[Dylan]

Why did we get the biggest gong? Why did we, like, why is there a limit on gong size?

[Tyler]

We've seen bigger gongs online. They do exist, but they get very expensive. Sort of a. Sort of an exponential relationship between gong size and price, unfortunately.

[Dylan]

Okay, we have to talk about this study that went viral over the weekend. Yeah, it is. Placebo sleep affects cognitive functioning and the takeaway is that literally just having a delusional golden retriever mindset measurably changes outcomes and physiology. Sleep badly, Convince yourself you're well rested, stressful day. Convince yourself it's fuel failed. Convince yourself it's useful data. So in this study it says the placebo effect is any outcome that is not attributed to a specific treatment, but rather to an individual's mindset. This phenomenon can extend beyond its typical use in pharma pharmaceutical drugs to involve aspects of everyday life such as the effect of sleep on cognitive functioning. In two studies examining whether perceived sleep quality affects cognitive functioning, 164 participants reported their previous night's sleep quality. They were then randomly assigned one of two sleep quality conditions or two control conditions. Those in the above average sleep quality condition were informed that they had spent 28% of their total sleep time in REM, whereas those in the below average sleep quality condition were informed that they had only spent 16.2% of their time in REM. Sleep assigned sleep quality but not self reported sleep quality significantly predicted participants scores on the PACED auditory serial addition test and controlled oral word association task. Assigned sleep quality did not predict participants scores on the digit span test as expected, nor did it predict scores on the Symbol digit modalities test when it was unexpected. The control conditions show that the findings were not due to demand characteristics from the experimental protocol. Those findings supported the hypothesis that mindset can influence cognitive states in both positive and negative directions, suggesting a means of controlling one's health and cognition. Takeaway golden retriever mode. Golden retriever mindset.

[Tyler]

They made a movie about the golden retriever mindset years ago. I know. You haven't seen it. Have you seen it? It's called yes Man.

[Phil]

I was thinking Air bud.

[Tyler]

Airbud is a great answer to that. No, that's about a literal golden retriever. Yes man with Jim Carrey is effectively about the golden retriever mindset. Basically, he's a bank loan officer. He's become Withdrawn, he's going through a divorce. He's having an increasingly negative look outlook on life. He then goes to this seminar with an inspirational guru who has him enter a covenant with the universe and say yes to everything asked of him. And so he just has to say yes to everything and hijinks ensue. But he has a fantastic time and it's a very interesting silver lining story. Bradley Cooper's in it. Zooey Deschanel and Jim Carrey star. Highly recommended. If you're looking for a good uplifting movie this week.

[Dylan]

Should we talk more about Meta's space solar project?

[Tyler]

Absolutely. What's going on there?

[Dylan]

So they announced this morning two new partnerships to bring innovative energy generation and storage to our data centers. We mentioned this earlier. Space Solar partnering with Overview Energy to beam up to 1 GW of space solar power from orbit to Earth for around the clock power production. What is the company? I think they're in El Segundo.

[Tyler]

Reflex. Reflex Orbital, Orbital. We talked about that. I think Sean McGuire did the deal

[Dylan]

and then they're also doing.

[Tyler]

And also. Who is it? Co founder of Robin Hood. Baju. Is that his name?

[Dylan]

Yeah, but I thought that was more of a cute play.

[Tyler]

He has done more compute, but at least at one point a piece of the business was collect energy on solar panels in space and beam it down via laser. And so that was all of these projects are incredibly difficult to math out and require a lot of different things to go. Well, they're very exciting, but this company clearly been working for a long time. But if it's working for them and it winds up working for Meta, you can imagine that there are going to be lots and lots of buyers because energy of course is in short supply. The Mirror in Space is such an interesting solution to what I'd heard before, which was collected on a solar panel and then beam it down on a laser. A mirror is such a simple solution to that. So we'll have to see. It feels like step one is just getting more solar panels down on the ground. You see these data center projects and a lot of natural gas turbines, not a lot of nuclear.

[Dylan]

Well, the question is like if you have this ability to bring basically 24.7sun, can you bring a lot more solar projects online? Because the economics just make more sense because you can power things like data centers, especially if you have batteries. The batteries that they are doing in tandem with this apparently have 100 hours of capacity. So presumably even if you had a few days of cloudy weather, you could still keep energy coming through the system.

[Tyler]

Yeah, yeah, it feels tough because data center wants to run 24 7, needs to run 24 7. The math on depreciation and the cost of the chips completely changes if you have intermittent electricity. And in some cases if you lose power, you can actually damage the data center. And so there's a whole bunch of other things that you need to work through. Did you see this post from Benjamin Todd? This was a very interesting post. We found this.

[Dylan]

You were the first person that kind of brought this up.

[Tyler]

I had looked at that. I'm aware. So the question is AI's impact on jobs and employment broadly? I looked at overall employment in India, overall employment in the Philippines, and how it was tracking this year because of course there's a lot of outsourcing. There's a lot of, you know, lower skilled, white collar style work call centers, BPOs, outsourced processing centers, like small, you know, atomic tasks that get done abroad. And so I expected that if there was going to be an uptick in unemployment, it would show up in potentially India and the Philippines. First, of course, we heard that somewhat hilarious quote that 90% of the Philippines economy is call centers. Of course it's not, it's closer to 5% or something, maybe 6%. But everyone sort of agrees, at least on the surface level, that as Benjamin Todd put it, it's hard to think of a more AI exposed job than Filipino call centers. But oddly, in 2025, employment was up 4%. And so of course people will say maybe it's earlier, the technology's getting better, all these different things. But there has been a process of automation around. I mean, I was trying to get on the phone with a company. Just earlier I had to go through a whole phone tree. It was even hard to find the phone number. There's a whole bunch of steps that companies take to try and reduce the amount of call center operators that are in the flow. And so this was not necessarily a new trend. US call center worker employment is in decline, but that started before ChatGPT and is probably mainly about outsourcing. So the outsourcing boom, you would think it would have started with like the dawn of the Internet. I would have expected the trend to start in 2005. You know, like internationalization. Globalization was well underway. In fact, US call center business support services, all employees for the United States. The peak happened in 2016 and then declined sort of during COVID and then has been declining ever since, probably as things move offshore. So there, so there is a world where, you know, these technologies, they Take time to diffuse. And so AI might play the similar role in the sense that, like, there is some sort of onboarding cost to moving from a US based to a Filipino call center. That's taken a decade to actually decline by, you know, not even half. It went from 900,000 people to 650,000 people over the past decade. Certainly not good if you're in that. If you're in that industry in America. But interesting, interesting nonetheless. Also, Poland is having a breakout year. Income in Poland is on track to overtake income in the United Kingdom. This is based on a forecast for advanced economies. UK is growing now slower than Poland. So everyone who's a fan of Poland will be excited to hear the news that Poland really is going through a fast takeoff over there in Poland. They are doing some great stuff. Poland was once a communist Third World country. Now it's overtaking Britain. This is in the Telegraph. European superpower is luring a record number of UK immigrants with its restored economy and robust patriotism. Interesting. Three months ago, the British businessman Johnny Mercer advertised a marketing role in his construction firm, Polestrad, based in Poland. Not long ago, people weren't interested in moving here as he sits down with a trendy French bistro. This time, however, Mercer was inundated with Britons eager to work in Poland. 35 applicants for the job were British and happy to relocate permanently, including one without any British links who got the job. People are excited. Noam Brown shared some interesting details about the different constraints on AI progress. He says no. Brown suggesting that model weights become relatively less important as inference becomes more important, which means securing weights still matters, but securing inference capacity becomes a strategic advantage. This is from a slide for a talk he gave which is very, very interesting. Just from an AI safety perspective, the idea of sneaking the weights out on a hard drive that you've smuggled in your suitcase, and that being equivalent to a suitcase bomb or refined uranium.

[Dylan]

Yeah.

[Tyler]

Yeah. It's not quite the same. Maybe the chemical chips or the refined uranium more than the actual weights. And the weights are merely one piece of the puzzle. Do you have any?

[Phil]

Yeah. It's interesting because I think, like, recently, the past few months we've seen this big fuss over, like, distillation.

[Tyler]

Yeah.

[Phil]

But like, you know, maybe there's an angle where, like, distillation actually gets less important because, you know, even if the Chinese can distill our models, they actually can't serve them. So it's like, you know, is that even important?

[Tyler]

Yeah. And even if the models are exactly the same, if I'm able to put 10 agents securing my bank account against your one agent trying to break into it. I will have 10 times the amount of solutions and so I should win that battle almost all the time. And it does seem like we're shifting towards this. The incredible value of inference and capacity which of course makes the whole data center slowdown ban so much more complicated because once you get into like the geopolitical considerations and what, what happens when large inference clusters start coming online elsewhere and you go back to the door cash? The door cash. What was it? Probability density curve where he, he says like if AGI comes soon, America wins. If it comes over long term, China wins. Because he's worried about China ramping up their capacity over time, but they're behind currently.

[Dylan]

Semaphore posted an article said Meet the man who's outsourcing almost everything in his life to his AI assistant, listens to every conversation, reads every message, emails and schedules meeting for him, all while pretending to be him. I didn't ask it to help me. I asked it to be me. And Taylor says I think this will be normal in five years. She has an excerpt here. So he's made a small fortune selling multiple companies to Apple. Multiple companies to Apple. That's always, that's all I love. I love stories like that. And recently launched a voice recognition startup called Olive Said his personal AI has all but taken over his life now. When he wakes up most mornings, he consults the agenda his AI assistant has crafted for him and then spends his days following its directions. The AI has permission to email people on his behalf and sometimes sets up in person meetings with people he has never met. It listens to conversations he has with his three kids and then suggests parenting advice, which he says has improved his relationship with them. It's a portrait of an emerging class of token maxers, power users who are plunging tens of thousands of dollars to MacGyver level AI assistance not by waiting for the next big model release, but by orchestrating today's models and loops with more computing power, more passes and more automated checking, and a massive dose of risk tolerance. The idea is to give the system an unlimited amount of tokens and access to every conceivable piece of relevant data. I didn't ask it to help me. I asked it to be me.

[Tyler]

So some of this is extremely weird. Some of this is maybe very normal. I'm trying to think of like how many of my interactions, my daily life are like, mediated by technology already. Like my alarm clock comes from my phone. It decides when I wake up more than anything else. In fact, the eight sleep will decide the optimal time that I wake up within a few minutes, right. Because it's like an adaptive alarm. And then I get in the car, I select maps. It sort of tells me what streets to go on. I'm merely like the embodiment of the AI to get where I need to go. And then I have a calendar that tells me what I'm doing when there are. There's some level of intermediation, but I don't know, there's still incredible value in touching grass. And I think that we will be.

[Dylan]

We'll see a bull market.

[Tyler]

I just think we're going to be in this barbell world potentially forever where you have Larry Ellison is buying Oracle data centers and going super long on AGI and also CBS and Foghorn Leghorn and you gotta own Bugs Bunny and you gotta own Superman and Batman, right. And then on the flip side you have Josh Kushner, investor in OpenAI ton of different artificial intelligence companies. And then on the other side, SF giants. And it's like are these diametrically opposed or are they actually. Are they actually both true visions of the future of the world? It seems like something that is gonna continue. We should go back to Manus. Do you want through the Financial Times with Tyler and I'm going to take a quick break.

[Dylan]

Let's do it.

[Tyler]

Can we do a two up with Tyler in place of me and you can read through the Financial Times and sort of some of the reactions because.

[Dylan]

Yeah.

[Tyler]

So.

[Dylan]

Regulators reviewed the deal, reviewed whether deal violated Beijing's investment rules. China has ordered Meta to unwind its $2 billion acquisition of AI app Manus as Washington and Beijing vie for dominance over the emerging technology. The decision marks an extraordinary late stage intervention by Beijing involving two non Chinese companies. Meta had already begun to integrate software from Manus, which was founded in China but relocated to Singapore last year. It was unclear how the acquisition could be unwound at such a late stage. A person briefed on Beijing's decision said the announcement could be intended primarily as a warning for similar deals in the future. The person said the gesture was pretty harsh and it carries a strong intention to stop follow on deals like Manus. In reality, it's hard to unwind a dungeon deal. Manus has been live, I believe in the Facebook ads manager. It is, you know, obviously been heavily branded as a Meta platforms company for some time now. The Meta team has been investing in scaling it and so yeah, very much feels like a done it had been a done deal. I don't, I don't think there's been any reporting on it, but I would assume that full cap table had been paid out in large part already. So it's very unclear how you undo something like this. It's not super surprising given that we obviously forced the sale of TikTok. And this feels like somewhat of a response.

[Phil]

Yeah, it is interesting. Is this the moment when China wakes up? Right? People, they're super like, AGI pilled. They're like, okay, at some point China's gonna, like, wake up. This seems like directionally towards that. But it is interesting because you have this and then you have China approving the sale of Nvidia chips there. So it's like, okay, how much do they really want to disentangle from the US regarding AI?

[Dylan]

Yeah, AI 2027. Had China wakes up in mid-2026.

[Tyler]

Yeah, it feels like, I don't know, Manus is. I mean, it's not like a cash flow acquisition. It's not a highly profitable thing that you're trading on earnings.

[Dylan]

It's a team and it's a wrapper

[Tyler]

and it's a technology, but the technology is somewhat commoditized. There's been code leaks from cloud.

[Dylan]

It's a super talented product team with a demonstrated track record of getting real paying users. And so the question was always, how much does Zuck care about keeping this as a standalone product, an AI assistant

[Tyler]

for business that why even buy the whole company then? Why not do one of those zombie acquisitions where you get the talent and then you get a license and then there's a ghost ship and you leave the ghost ship. Yeah, I would assume, because that's got to be harder to approve.

[Dylan]

I'm assuming that that's kind of what happens. I don't know.

[Tyler]

It seems like it's too late. They already bought the company, right? Like, they already did, like, the proper acquisition. Like, you know, as you said, like, paid out the cap table. I don't know. I don't know the exact terms of the deal, but it would have been. It probably would have been easier to do something like what Grok and Nvidia did, or like the Windsurf Google thing where, like, you're bringing people over with this contract and then like, yes, China blocks it, but it's like, what are they even blocking? It's just people getting a new job and a licensing deal that the money flows through, and then maybe they try and claw that back. I don't know. It is a Tricky, tricky situation.

[Dylan]

Yeah. I'm not familiar with an acquisition that actually closed in venture that was then later fully unwound.

[Tyler]

I don't know. I don't know. Delian's taken a victory lap. He says, wow, so weird that they can do this since it's not a Chinese company. According to Gurley, there's always been back and forth about whether or not China would have any power over the Manus team. It seems like they have some, Some. Some sort of power. Says after China's cancellation of Meta's purchase of Manus. Why would any founder start an AI company in China if they had a choice? I mean, well, you can make money in cash flow in China. You don't necessarily need to sell to an American hyperscaler to have a wonderful life as a founder of an AI company in China. But he makes the argument, in China, you have access to less compute, less capital, and salaries are lower than in the West. And if you are so successful that a non Chinese firm tries to acquire you for billions of dollars, the Chinese government will lure you back to Beijing, ban you from leaving the country, and take your profits by canceling the acquisition. Manus did everything right. They even moved their entire business to Singapore to comply with U.S. outbound investment restrictions. Their only mistake was that they originally founded the company in China. It's not even clear what this means for China to force Meta to unwind the transaction. Is it going to force Manus researchers to return to China and place exit bans on them too? Is it going to force Manus founders and shareholders to pay back $2 billion to Meta? This is what happens when you regulate by fiat rather than by rule of law. Ultimately, this is a much larger defeat for the Chinese AI ecosystem than for the United States interest. META will be fine without Manus, but Chinese nationals looking to AI to found AI companies will increasingly start them overseas. Hmm, that's interesting. The message from the Chinese government here is that every AI company founded in China will forever remain subject to the Chinese government regulatory pressure and manipulation, regardless of its legal status. So he goes on, but you can read that there. What's Bill Bishop up to these days? He says he's quoting from the Financial Times. Did we already read this? A person familiar with the matter said Beijing had told the two companies that the deal must be unwound completely, including returning funds, re registering the company's ownership and halting Meta's unit use of the Meta of the Manus algorithm. The person said that if the parties failed to fully undo the acquisition Beijing could impose penalties on Meta limits, limit its China related business, and possibly pursue criminal charges for individuals involved.

[Dylan]

That is a wild China does have a good amount of leverage given that like tens of billions of meta ad spend originates from Chinese companies and so they could put pressure on Chinese companies to pull back spend, which would hurt meta. So yeah, very, very unclear how this will all sort itself out. But yeah, unfortunate for everyone involved.

[Tyler]

Pull up this post from Michael Chang showing sort of a glimpse of like the future of Generative ui. So the prompt here hey chatgpt, what's the weather like today? Might have been a little bit more complicated than that, but using the new images 2.0 it is rendering sort of a video game style map. I don't even know. This feels like the type of map that you'd see at the front of like the Lord of the Rings book or like a Game of Thrones book, but it's giving you the actual information, like accurately telling you for each neighborhood what the weather is like. Of course you didn't need that much information because every single town is 56 degrees, maybe 57, maybe 55. There's very, very slight differences. It is an interesting world where you're getting closer and closer to this generative on the fly ui. Ben Thompson wrote a big bull case for the Meta augmented reality headsets. Not just Orion, but also the Meta Ray Ban displays today. Talking about has AI models get better on the fly UI generation with less chrome which is like the top bar and the bottom bar and less permanent UI functionality is what actually feels magical. Like when you go to look at something and you get something that perfectly sums up exactly what you're looking for on the fly. And previously was it possible to build something like this? Absolutely. But you would have to hang out in Blender and create the 3D map and and render it and then build some webpage that would go and pull in the data from APIs and place

[Dylan]

it on the Ryan says One Tower Golden Gate is a crime.

[Tyler]

So more work to be done. Jobs not finished. Goodbye.

[Dylan]

Cheers.