More or Less Podcast

OpenAI's Astra Is Scary Good. Why the AI Labs Can't Slow Down

2047 summary words 9 min summary Watch video

Start with the signal

9 min read

Summary

At-a-Glance

  • Verdict: Watch fully
  • Core thesis: The agent market is still pre-breakthrough and rapidly fragmenting, but new frontier capability exemplified by OpenAI's Astra is making both the consumer-agent race and the safety/governance problem materially more urgent.
  • Why it matters: The discussion directly addresses the architecture, distribution, trust, economics, local-versus-cloud deployment, and control implications that will determine which agent platforms become durable.
  • Best use: Use it as a strategic market-sensing conversation: extract its framework for evaluating agent harnesses and its warning that capability acceleration is outpacing credible coordination mechanisms.

Executive Summary

The hosts frame the current agent landscape as analogous to personal computing in 1982–83: many competing machines and interfaces exist, but the broadly usable consumer breakthrough has not yet arrived. Meta's Muse, Instinct, Grok, OpenClaw, and an anticipated Google offering are treated as competing “agent harnesses,” each trying to become the layer that connects models, personal data, tools, and recurring workflows.

Their central market argument is that superior models alone will not decide the winner. Consumer-scale agent adoption depends on three factors: lower operating cost, much better user experience, and a credible answer to where data runs and who controls it. Meta has exceptional distribution and an advertising-funded route to freemium adoption but less trust; Google combines existing user data, broad distribution, and comparatively high trust; Apple may use device-local, always-on intelligence as a privacy-oriented entry point. OpenClaw's stated position is that local/on-prem operation and support for open models serve users and enterprises unwilling to place email, financial, or proprietary data in cloud agent systems.

The most consequential segment is the reaction to Astra. One host says Astra's English-language interaction may not be the main leap, but its computer-use and 3D-spatial capabilities are striking: their 12-year-old reportedly produced a printable 3D model in an hour. This becomes the basis for a more serious safety concern: capable agents, especially when parallelized at thousands or millions of instances, may cross a threshold where misuse has material real-world consequences.

The panel is skeptical that public “slow down” messaging from AI labs is sufficient when labs face competitive races, investor-return pressure, and incentives to self-declare milestones such as AGI. They consider—but do not substantiate—a more cynical interpretation that prominent AI-risk messaging could be intended to drive regulation favorable to closed frontier labs. The useful conclusion is not the conspiracy theory itself, but the governance gap: frontier actors may recognize danger while lacking unilateral ability or incentive to decelerate.

Key Takeaways

  • Claim: The agent sector is expanding rather than converging, and the decisive consumer agent has not been built yet. | Evidence: Dave Morin compares the market to 1982–83 personal computing, before the Macintosh made computing more broadly usable; he argues that even ChatGPT is only at roughly billion-user scale while about six billion people use mobile-phone software. | Implication: Treat the current proliferation of agent products as an early platform race, not proof that the market is saturated; durable differentiation still has room to form. | Caveat: The historical analogy is a strategic framing, not evidence that a single universal agent will necessarily emerge.
  • Claim: Agent-harness winners will be determined by cost, user experience, and data deployment/control—not just model quality. | Evidence: Morin explicitly identifies run-cost reduction, substantial UX work, and the cloud-versus-local decision as the three unresolved issues; he reports daily demand from enterprises and users wanting open-source models and data kept on-premises. | Implication: For OpenClaw and agent-system design, local execution, open-model support, secure enterprise deployment, and an intuitive harness are strategic product primitives rather than secondary features. | Caveat: The discussion gives no comparative cost data or adoption metrics by agent platform.
  • Claim: Distribution, existing data relationships, business model, and trust create asymmetric advantages for incumbents in the personal-agent race. | Evidence: The hosts argue Meta can push Muse through its global product distribution and advertising model; Google already holds users' email and other data and is expected to build a cross-product harness; Apple is positioned to use strong on-device silicon and privacy framing for always-on AI. | Implication: Do not compete with incumbents solely on generic chat or assistant features. Build differentiated control, interoperability, local/private workflows, or vertical utility that cannot be replicated merely by preinstallation. | Caveat: This is forward-looking opinion; neither Meta's Muse nor a unified Google harness is demonstrated as having achieved durable engagement.
  • Claim: Always-on AI is becoming a practical product direction, but it creates a privacy-control problem even when raw audio is not retained. | Evidence: The hosts describe Apple's purported Watch “audio intelligence” as continuously transcribing conversations without retaining audio or identifying speakers; they then note that a user could combine those transcripts with calendar context in an agent to infer who said what. | Implication: Design privacy around downstream inference and cross-source correlation, not only raw-data retention. Consent, data boundaries, retention, identity resolution, and agent connector permissions need to be explicit. | Caveat: The transcript's product-description details and legal/privacy characterization are not independently verified, and the speakers acknowledge the practical tension between privacy claims and user-driven data enrichment.
  • Claim: Astra appears to represent a capability jump in computer use and 3D reasoning that makes misuse risk more tangible than abstract AI-doom rhetoric. | Evidence: A host reports that Astra enabled his 12-year-old to create a 3D model within an hour and print it, and says the model is extremely capable at using computers and operating in three-dimensional space. | Implication: Reassess agent safety based on capability combinations—tool use, spatial reasoning, execution autonomy, and parallel deployment—rather than model text quality alone. | Caveat: This is an anecdotal evaluation rather than a benchmarked comparison, and the specific harmful-use pathways are not described.
  • Claim: Frontier labs have structural incentives that make voluntary slowing unlikely, even while their leaders publicly raise safety concerns. | Evidence: The panel cites the immense capital committed to Anthropic and argues that investor-return expectations, competitive pressure, and the race to claim AGI make growth deceleration financially and strategically difficult. | Implication: Assume acceleration is the base case. Build operating plans, security posture, and investment theses that work under faster capability releases rather than relying on self-restraint by leading labs. | Caveat: The speakers speculate about individual labs' motives and regulatory lobbying without providing documentary support; this should not be treated as a factual allegation.
  • Claim: OpenClaw's local/on-prem and open-model posture addresses a real segment of demand, while agent adoption is already appearing in specialized production workflows. | Evidence: Morin says OpenClaw had its biggest week and claims it is reportedly running the entire film-editing pipeline for a top film at the Telluride Film Festival; he also says privacy-sensitive users seek local data and open-source-model deployments. | Implication: Prioritize credible reference implementations in high-value production environments and turn local/private operation into a demonstrable operational advantage, not an abstract positioning statement. | Caveat: The growth claim and film-production example are not quantified or independently corroborated in the transcript.

Detailed Brief

How the hosts assess the current harness field

  • Claims: Muse is viewed as a reasonable consumer-oriented start rather than a compelling power-user product; its guided ideas and goals may help AI newcomers, but its output quality and unique purpose remain unclear.; The hosts expect agent products to initially segment by audience and workflow, although one participant argues the major players are ultimately all competing for mass-market users.; Google is portrayed as likely building a secure cross-product harness rather than lacking agent work; the panel believes its caution reflects the difficulty of protecting users against data breaches at large-company scale.
  • Evidence: Muse reportedly could not initially summarize a user's Instagram history and could only access text descriptions rather than the images themselves.; The discussion groups Instinct, Grok, Muse, OpenClaw, Town, Hermes, and future Google products into the same broad harness race.; One host says Google has already embedded agentic features in individual products and showed hints of a broader connective experience at I/O.
  • Caveats: The product comparisons are informal and experience-based; no standardized task evaluation, retention data, or security analysis is offered.; The transcript contains repeated passages and some named-product references that are not contextualized enough to support a precise competitive map.
  • Implications: A generic agent interface is becoming table stakes; onboarding, connector quality, permissioning, and workflow reliability are likely to matter more than novelty.; Monitor whether incumbents make the harness a standalone destination or embed it invisibly across their existing surfaces, since those produce different competitive dynamics.

Safety discourse: credible capability concern versus political narrative

  • Claims: The panel distinguishes between sensationalized extinction messaging and a legitimate need for a society-level conversation about increasingly powerful AI systems.; They praise the former Anthropic researcher for leaving ahead of an apparent equity cliff, interpreting the action as more meaningful than signing a public warning while continuing to work on the technology.; They describe AGI as an operationally vague term and suggest public self-declarations can be strategically useful when corporate commitments are tied to reaching AGI first.
  • Evidence: The conversation refers to a former Anthropic researcher, Jacob, who made a high-profile media circuit after leaving the company and raised AI-risk concerns.; The hosts state that OpenAI had or has a provision to merge with another organization that develops AGI first, then speculate that self-declaring AGI could make such a provision irrelevant.
  • Caveats: Claims about the researcher's compensation, exact departure timing, OpenAI's corporate provisions, and labs' lobbying goals are discussed with uncertainty and should be independently verified.; The panel offers no concrete regulatory proposal or technical governance mechanism capable of coordinating the actors it says must be coordinated.
  • Implications: Separate observable capability evidence from rhetoric about existential risk or competitor regulation when forming strategy.; Track operationally testable thresholds—autonomous tool execution, replication/parallelism, access to sensitive systems, and physical-world design capability—rather than relying on AGI labels.

Notable Concepts & Terms

  • Agent harness: The umbrella interface/control layer that connects a model to user context, data, tools, and workflows; the hosts see this as the main competitive category.
  • Local/on-prem agent deployment: Running agents and models under user or enterprise control rather than sending sensitive data to a cloud service; presented as OpenClaw's core trust and enterprise wedge.
  • Open-source models: Models that organizations can run and control themselves; positioned as important for privacy-sensitive users and enterprises.
  • Always-on AI processing: Devices continuously collecting or processing ambient context to create memory and assistance; its utility depends on broad contextual access, while its risk includes inference from combined data sources.
  • Astra: The frontier model discussed as unusually capable at computer interaction and 3D modeling, triggering the hosts' concern that practical autonomous capability is accelerating.
  • Parallelized agents: Thousands to millions of agent instances operating simultaneously; the hosts identify scale of execution as what turns a powerful individual model into a potentially systemic-risk issue.
  • Regulatory capture: The speculative concern that incumbent labs may use safety messaging to obtain rules that restrain open-source or smaller competitors; presented as one possible interpretation, not established fact.
  • AGI: A contested and poorly defined milestone that labs may compete to claim; the conversation treats the label as strategically consequential despite its ambiguity.

Operator Notes / Why Ken Should Care

  • Create a scored competitive matrix for agent harnesses using distribution, trust, data control, connector breadth, local/on-prem support, run cost, and workflow reliability; avoid evaluating competitors primarily on chat quality.
  • Add a threat-model review for cross-context inference: assess what an agent can deduce when transcripts, calendars, email, contacts, and task systems are combined even if each source independently claims privacy protections.
  • Develop and publish one or more high-value private-deployment reference architectures, with clear boundaries for model hosting, credentials, audit logs, connector permissions, and data retention.
  • Benchmark emerging frontier models on real OpenClaw-relevant agent tasks, especially computer use, multi-step execution, visual/spatial tasks, and safe parallel operation; capture capability deltas release by release.
  • Do not base planning on voluntary lab slowdowns or AGI claims; maintain an acceleration scenario that assumes rapidly improving models and intensifying incumbent distribution.

Source/Metadata

  • Title: OpenAI's Astra Is Scary Good. Why the AI Labs Can't Slow Down
  • Transcript words: 14411
  • Duration seconds: 3077
  • Timestamp note: No timestamps or chapters were provided; transcript includes substantial duplicated passages.
Full transcript 9570 words · 64 min read
0:00

OpenClaw is actually growing faster than ever. And I think what this means is that the world of agents is just expanding. I always tell everyone that it's like we're in 1982, 1983, in the personal computer, and the Mac hasn't been invented yet. Everyone's creating new computers, trying to make a computer that's more usable for everyone, racing towards can there be a consumer version? Because the big prize is an agent that everybody in the world uses.

0:02

When you hear someone like Altman, who now says once a day on some platform, podcast, or conference, we need to slow down the pace, you're like, okay, I think that's good. And the people that were here in the Valley were here to be creative and build cool stuff. And I can't say that for everyone that's here right now. In the early 2010s, people in tech were making things because whatever they were working on was awesome because they loved the creative endeavor of it. And that has changed. There is a lot of people making things in the Valley now that just make money. More or less.

0:12

Is it no or is it yes? We'll debate the tech that's best. When we get more or less. Dave and Britt plus Sam and Jess. Put it all right to the test. More or less. Why, hello, friends. Specifically, Morin, since you're the only one here. Welcome to More or Less. How's everyone doing this week? Hey, Jess. It's a good week. Football kicked off. But are you in any fantasy football leagues? The Information has started one and I'm not ready to participate, but I'm almost ready. I usually do, but this year I skipped it. So I would be open to joining one if you want to invite me.

0:27

Oh, no, I don't have the bandwidth for that. But maybe in the future. Are agents doing fantasy football yet? Is this a thing? I did create a WhatsApp group for the fourth grade cross country team this week. That was my sports organizing. What if all of us just had our agent join together in a fantasy football league and make their most informed choices and argue each week? That would be interesting. I think so. What happens when all the AIs just win all of the fantasy football leagues?

0:44

You know what, Dave? I'm going to assign that story and we will find out. I like that story. There probably are real agents in these leagues right now. If any of you listeners out there are in a fantasy football league, I would not trust your tech friends that are in this league with you. Instinct, Muse, OpenClaw. They're all over this.

0:48

Well, yeah. Speaking of agents, I just did for all parents out there that have ever had this experience. I feel like I pulled off the triple Lutz or whatever, because I managed to get the primo parent teacher conference slots in the first thirty seconds of parent teacher sign up opening. And it's taken me five years to perfect my technique. And bam, bam, bam. Mic drop. Something else an agent could have done for you, Jess. You're living in the manual archaic days right now. No, there's no way an agent could duck the number of logins in the Byzantine system.

1:05

Oh, I disagree. Let's come back to this next semester and see, like six months from now. Okay. I can't wait to ask school support help. I'm still trying to get a Waymo stop. So, but anyway, I feel good about that. And then I also feel good. I managed, and at the end of this episode, we'll cover what's happening in the wide world of tennis because it's very important. But in the middle of a meeting yesterday, I got an email that said that the BNP Paribas Indian Wells tickets were open. What does that mean?

1:12

That was a lot of letters. This basically means you're going to be ten minutes late to your next meeting and you refresh, refresh, refresh, refresh, refresh. And there are other ways to get tickets, but being the delicate flower that I am and the fact that my children don't usually like one hundred degree heat, I need my tickets in a very special way. You need the shade. Yeah. Yeah. I can tell you all now because I just got them.

1:20

Yeah. That does seem to be the thing on tennis things. It's like if you get stuck in the sun, it must be miserable. Same with the forty-niners games. There's a specific section. One hundred seventeen to one hundred twenty, run row T or higher is maximum shade.

1:25

I enjoy taking a lot of people to this tennis tournament, so I buy a lot of tickets. I had two laptops and four browsers to defy the bot detection. It was all me though, no bots. I got all the seats I wanted. I feel like that's fine, but I know. I just feel like that was a lot of wasted energy for you, Jess. Yeah, the agents can do all this for you now, Jess. But now I know. I don't have to check. It's just very satisfying. And I need those exact seats. So anyway, one, the internet over here is a human. But guys, it's been a busy week. We have a lot to talk about. We've got Apple. We've got agents. We've got doom. We've got gloom.

1:33

I want to start with Apple. Doom and gloom. Dave, you, we have, you know, here I am yesterday trying to convince the internet I'm not a bot, mildly running a news organization for those ten minutes and many other things. And my mother texts me, should I get the new iPhone? And I'm like, good Lord, here she is probably just pressing by on this thing. Right. So Dave, bring us inside. Was this Ternus' first Apple big hardware unveil as CEO? What was it like? Yes, it was Ternus' first keynote at the helm. Can we call them keynotes? They just record videos and show them to people. Yeah, it's not really a keynote anymore. Okay. It's more of a... Movie?

1:43

Yeah. Which Apple's audience is many billions of people in every country in the world. So I get it. They got to do this thing on day one in every language. And, yeah, I mean, it was a great day at Steve Jobs Theater. People were in good spirits. And the big news is the launch of the iPhone Duo, which interestingly, I was watching the video and I was very skeptical. I've had one of these folding phones before. I had the Google Pixel because I was really interested in it as mostly actually I carried it to read on because I liked the format for reading. And it never made it to become my primary device. And I ultimately ended up stopping using it. And so when I was watching the video, I was very skeptical, particularly of the size. In my brain, I thought that it was two of these.

1:44

Is it two iPhones? It's much smaller once you hold it in your hands. The best way to think about it is if you were to cut the bottom and the top off of your iPhone, it's that big. Like an inch off the top and an inch off the bottom? Yeah. But then two of them? Yeah. And then it's as thin as the iPhone air. I don't think my fingers can reach that far. Also, will this fit in a woman's jean pocket? Because I only store my phones in pockets. Oh, not in my jeans. Not in my jeans, Brett. No way. The answer to all of these questions is yes.

2:13

Did you see this thing happening where everyone was talking about how Apple used AI on the fingers? Because there's no way the fingers could be that long holding the phone in the ad. Guys, we know an Apple iPhone hand model. We do. And she does not have long fingers. Shout out to Jessica Sittig. We are going to ask her what she thinks of whether she could have hand modeled the Duo. I think the hand model problem is probably just the camera tricks. It's probably just the type of lens that was used to shoot it.

2:20

But that was the first thing that I was thinking about is this thing's going to be too big. And I'm going to hate it. Once I had it in my hands, I was immediately won over. It's the perfect size for what it is. What is it? Is it a primary device? Yes. Because most people don't have two phones plugged to the... Yes, it's a primary device. The screen, when you open it, is hard to describe in pictures or even with words. The quality of the surface, it almost feels like paper. It's more like a Kindle, but even better, if that's possible to describe. The quality of the screen is so good that your hands are like, I've never felt anything like this.

2:35

But the camera isn't there. And basically, I carry my phone for the camera, right? What's the camera sitch? The camera's pretty good. They have two extremely high-quality cameras. We'll have to look up the specs. I mean, I've been carrying the iPhone Air all summer. And I'm going to hate it. Once I had it in my hands, I was immediately won over. It's the perfect size for what it is. What is it? Is it a primary device? Yes. Because most people don't have two phones plugged to the... Yes, it's a primary device. The screen, when you open it, is hard to describe in pictures or even with words. The quality of the surface, it almost feels like paper.

3:17

It's more like a Kindle, but even better, if that's possible to describe. The quality of the screen is so good that your hands are like, I've never felt anything like this. But the camera isn't there. I carry my phone for the camera, right? What's the camera sitch? The camera's pretty good. They have two extremely high-quality cameras. We'll have to look up the specs. I've been carrying the iPhone Air all summer. Yeah, Britt, what are the specs on it? I think it's 48 megapixels. I watched the whole video. It's a great camera. It's much better. I've been carrying the iPhone Air all summer. I switched to the iPhone Air, and I'm absolutely in love with it.

3:57

But the main problem with it is that it does not have the wide-angle camera. I actually missed the wide-angle camera more than I missed the zoom camera. And this one has, I believe, both wide-angle and normal camera. And I do not think it has the zoom. The zoom is reserved for the pro. That's pro camera work. Okay, so you're sold. You're buying. I really think that this is a truly great product design. This is a classic example of Apple taking something that's a known product and then doing the Apple thing to it. They really solved a lot of key problems. One, they had to make the OS. They had to redesign the navigation of the OS to fit this new form factor.

4:34

And it was really well done when I held it in person. They did some really thoughtful things with how you can set it on the table in folded mode. And you can watch movies. You can use the keyboard like a little miniature laptop. You can flip it over and make it an alarm clock. And it glows when you wake up in the morning. They were really thoughtful about this form factor being foldable. Oh, $3,000. The glow in the morning. That's going to convince me. It's $2,000, isn't it? My favorite feature is that there's a kid use. So when you're trying to get a photo of your toddler, you can play like Snoopy and say, look at me over here.

4:59

And so it gets their eyes to look at the camera. So more ways to manipulate our children. Okay. Or you can turn that on and show it to the person you're taking the photo with. And they can see the framing of the photo while you're taking the photo, which is a cool idea. It's a really interesting device. I think it does exist somewhere between a phone and an iPad. I think it's more computer than phone for what it's worth. And it may not be for everybody. But I do think that a lot of people are going to get this thing. Britt, what about you? Are you going to switch? Are you going to try? Well, I need, the hand thing is real. And the jean pocket is real.

5:40

I'm not a purse girl. I'm a pocket girl. It needs to fit. Have you heard of a fanny pack? There's a middle ground. Just saying. I don't think I can carry a fanny pack every day. It's $2,000. And if you want the two terabyte version, it's $3,000. The Pro is getting up there, though. No, it's $1,199 for the Pro. This is more expensive than a MacBook. It's crazy. I got a new studio display last week for $1,500. So this is two studio displays that can fit in my pocket. So I think that price is going to deter a lot of people. But I do think the novelty will create a lot of buzz, especially for the early adopter types.

6:36

And if you think about it, it's an iPad and an iPhone priced together in one. That's how I'm trying to think about it. There's no keyboard. No, you can make a keyboard when you fold it. A mini. It's a mini computer. But it's going to need a bunch of new accessories. All of the Apple accessories products need to change. The other whisper that I heard was that the markets in Asia, this is the favorite form factor. The why is missing here, right? Because obviously these phones have existed. People use them. People have a lot of phone preferences. But for Apple to enter this category, they have to see a big opportunity or be afraid of something.

7:07

I think the opportunity is China. My understanding is that China, this is the favorite form factor. And it has been for a long time. Samsung did their folded phone many years ago. I went in skeptical. And I'm telling you, when you get this thing in your hands, you're going to consider it. Okay. What about the watch or the AirPods? I was less excited about AirPods. Maybe a little bit more about watch. It seems the TLDR of the watch, from what I gather, is basically now competing directly with Aura and Whoop and all the other fitness trackers. It does your readiness score, your sleep score, your HRV, all the things. It has longer battery. Yada, yada, yada.

7:49

Aren't we scored up the wazoo? I know, but a lot of people are buying all the rings and wristbands still. So Apple, I think, is trying to say we could do it all in one. They're going to do your longevity score now. They're going to use all of your data from all of these different things. The most bullshit number. Well, can they tell me my biological age and how fast I'm aging? 0.85 years. Guys, today I feel 43 and a half. But it doesn't have my blood yet. I still want to compete with Sam on this because I do think I'm going to age slower than Sam. It's so sad that Sam's pounding the pavement, visiting the U.S. Open in New York right now.

8:24

There's actually a really important nuance that a lot of people missed. And for some reason, the internet, at least the Twitterverse, was not paying attention to this, which is the watch has a software change, which is very interesting. They call it audio intelligence, which is that the watch is now listening to all of your conversations all day long and transcribing them for you. It doesn't record audio. It doesn't differentiate speakers. They pitched it as a memory device, which for people like me, I have really terrible rote memory because of my brain type. And so when I was reading or watching this, I was, wow, that's a really thoughtful feature.

8:38

I'm worried about businesses like Granola because Apple just basically put Granola into the Apple Watch running 24-7 and did it at a higher level of privacy than Granola. And so that's an interesting positioning thing. How can I bring up something I forgot about from yesterday if they aren't storing it? It's storing the transcripts with no names and no audio, which I believe is what follows the law, right? And so it transcribes it and then leaves it as text. So you still have the text for it. Can I MCP that into my agent? Yeah, I think so. Okay, great. We're billing something for more privacy. And then we're trying to strip out all the privacy for the maximum utility.

9:03

I'm just pointing out that's what's happening. Well, here's the thing. I'm going to MCP it into my agent. My agent has my calendar info. It's going to be, oh, you were meeting with blah, blah, blah when they said this thing yesterday. So then I'll be able to connect the dots. So even though Apple's not skirting the law, I'm still able to. I actually asked Sam to direct my bot to do something that he told me he wouldn't do out of privacy. So we found the one thing that Sam was, he probably wouldn't do that. It was to read my Gmail. I just want to know. Oh, that's what a lot of people do. I know. I didn't think that would be a huge.

9:47

Anyway, shows everyone's mildly inconsistent. Okay, guys, let's wrap Apple. Last points on Apple. I think this is just an interesting positioning point, Jess, that this is new territory for Apple to be proposing that a device listens all day. What is the billboard going to say? Don't worry, it's iPhone with big ears and an X through it. I don't know.

10:04

So then I'll be able to connect the dots. So even though Apple's not skirting the law, I'm still able to. I actually asked Sam to direct my bot to do something that he told me he wouldn't do out of privacy. So we found the one thing that Sam was like, ah, he probably wouldn't do that. It was to read my Gmail. I just want to know. Oh, that's what a lot of people do. I know. I didn't think that would be huge. Anyway, it shows everyone's mildly inconsistent.

10:09

Okay, guys, let's wrap Apple. Last points on Apple. I think this is just an interesting positioning point, Jess, that this is new territory for Apple to be proposing that a device listens all day. What is the billboard going to say? Like, don't worry, it's iPhone with big ears and an X through it. I don't know. I think it has to do with future plans and that the more of these devices, whether you've got other ones in the home or glasses on your face, to actually make sense of the world with AI, you need to have them turned on. Right?

10:14

I think this is the beginning of Apple moving into always-on AI processing because they're in a great position to do this. Is this a shot across the bow from OpenAI who has hardware ambitions for pucks that are going to sit on their desks? Perhaps. Oh, I don't know. I actually hadn't thought of that. Maybe. I think so. But there's also all the rings and necklaces that are always listening and all of that stuff. It's more of a shot across the bow to OpenAI, to Meta with the glasses, all these kinds of things. This is a really nuanced privacy thing that we're moving towards. Right?

10:28

And everybody in Silicon Valley uses Granola, but most of the people in the world do not. So there's an interesting territory here. Yeah, definitely. And also the number of times AI wasn't mentioned to or crossed. I mean, clearly the Apple marketing team has changed their positioning on. No question. They want all of these things to be AI devices, which they are in many ways. To their credit, they've got great Silicon. Apple Silicon is the best, if not the best, on the inference side of the house. Do you think any podcast talks as much about Apple Silicon as this podcast? No, but we should. Let's just rebrand ourselves to the Apple Silicon podcast.

10:35

Oh, we should talk about our new branding at the end. Okay, I'm going to move us on to other AI topics, but I love that debrief, Dave, and I love it. Okay. So anyone use Muse? Meta's done a lot in AI. They've been through different models. They've been through different teams. They launched this week their next big play at a personal agent. Seems to me, Dave, they're coming after the OpenAI world with this. It's called Muse. I've used it a bit. Anyone got any feedback? Yeah, it's funny. I downloaded it. And then the first question I asked it was, what do you know about me? And that's what I asked mine too. I was expecting a book.

10:42

It got nothing. It knew nothing. Don't you literally know me for 20 years, Muse? Look at all my Facebook and Instagram data and WhatsApp. Muse knows where I ate in college. But yes. Yeah, I went even further and I said, read all of my Instagram posts and tell me about me. And then it's like, I'm sorry, I can't do that. I'm like, they're my posts. Why can't you do that? And then it's like, okay, I will try. So it goes and tries to read all of my Instagram posts. And it comes back and it says, I cannot read the photos, but I can read text description of your photos. Oh, oh. And I was like, oh, come on. Multimodal, multimodal.

10:49

All jokes aside, though, I do have to give Meta props. They've shipped a new product and it's been a while since they've done this. So I just think props to Meta for getting a new agent product out the door. It's a great start. And there are some nuances. I love the actual avatar. Yeah, so cute. The animations that they did around how the agents are thinking is really thoughtful product design. I don't know. I think the product, it's not clear to me what purpose this agent versus any other agent has. Whether or not consumers are going to really get into it. I think that's the question of the next five years. But Meta's got a great product team, so we'll see where it goes.

10:57

What's so interesting about it is Meta's doing exactly what Meta probably should be doing, but they're also picking their lane, which is can we build a pretty good product that can be used both because we can push it to people. Hello, Threads bigger than X. And also bring not just the ChatGPT version of AI, but the more agentic version to more people. Yeah, totally. Does it have to be as fully featured as everything else that a power user is going to do? No. Is it going to have to be great? Yes. It's free. I mean, so I'm like, okay, eventually the investors are going to wonder. Is it free? It actually, I thought it had a subscription. No, it's going to be paid.

11:24

Yeah, it'll probably be freemium. I mean, obviously the ads are going to roll through because Meta runs that. I think all the details haven't been hammered out, but I'm confident that the strategy is to get as free as possible and then ultimately have ways to get paid. Yeah. I mean, their business model supports that, right? They're actually in the best position of all. They've got a great advertising network. I think Google is actually. Well, let's come back to Google. What's Google got for me here?

11:29

Here's my quick take on Muse. Fine. Yes. Agree with Dave. I think for the normies out there, it's nice that there's this thing on the sidebar called ideas and goals and ways to get started. I think a lot of people that have never used AI are just like, what is this? Help me with the task. So that's great. The UI is good. I don't think the content it's giving back to me is great yet, but fine. Good start.

11:30

Second point. This is the total agentic harness war. In the last two weeks, it's been Instinct and Grok and News and everything else. So I'm just like, how many is too many, guys? And where is this all going to net out? But I think they're going after different parts of the bit. I am not as familiar. I haven't. I haven't trialed Instinct. That's what everyone's talking about. No, because I'm not a VC. I mean, this feels like Superhuman all 10 years ago. Yeah, there's no way I'm giving Instinct my email. But also I'm not anti-Instinct. I just haven't used it. But have you used Grok? Kind of. Guys, I also built my own cloud codes, right? Okay. Well, I use all of these.

11:48

No, but my point. I want to pick up on your point, which we are seeing this war, but we're in the phase of the war where not everyone's going to go after the same market with the same feature set. So I actually think Muse has a different intended audience than Instinct, at least at this phase. I don't know, Jess. I actually disagree with that. I think they're all going after the masses. And then there's Town and there's OpenAI and there's Hermes. What is Instinct's business model? So Jess, this is why Meta has such an advantage. You don't know about Hermes? No, guys. I don't know about Hermes.

12:05

The Information writes about it, Jess. It raised at $1.5 billion, just like all the others. They're all valued somewhere between $1 and $5 billion. And then you've got Muse from Meta on top. I'm waiting. And then you've got Grok from XAI slash Cursor slash SpaceX, whatever. Then I'm waiting for Google's to drop because I think once Google's drops, it's game over for me at least. I'm moving everything to Google. But okay. So Britt, why? And I'm interviewing Corey in two weeks or something. But why hasn't Google gone fully agentic here? Because I'm spending a lot of time getting my agents to talk to Google.

12:09

Here's why. Let me tell you why Google hasn't played here. First of all, Google has done a bunch of stuff agentically in all the products themselves. Yeah.

12:13

I don't know about Hermes. The information rights about it, Jess. It raised at $1.5 billion, just like all the others. They're all valued somewhere between $1 and $5 billion. And then you've got Muse from Meta on top. I'm waiting. And then you've got GrokBot from XAI slash Cursor slash SpaceX, whatever. Then I'm waiting for Google's to drop because I think once Google's drop, it's game over for me at least. I'm moving everything to Google. But okay. So Britt, why? And I'm interviewing Corey in, I don't know, two weeks or something. But why hasn't Google gone fully agentic here? Because I'm spending a lot of time getting my agents to talk to Google. Here's why. Let me tell you why Google hasn't played here. First of all, Google has done a bunch of stuff agentically in all the products themselves.

12:15

Yeah. Some steps. Some steps. Yeah.

12:31

I think Google is building a harness. I don't know this. This is what I would imagine. They have to be building a harness that spans. I think in I.O. They showed something that I'm still waiting. But something in the form factor of like a Muse or a GrokBot or OpenClaw, like in its own application that connects all these things together. But, and this is key for Google, does it securely, safely, figures out what the limitations are for data breaches. Like it's a huge company and has to make sure it's protecting its users. That's why, by the way, I will trust it. It already has my email. It already has all my stuff. So I'm excited to move over to it. We'll just trust it because we've already trusted it. So any vulnerabilities we're already exposed to.

12:33

No. Yeah. Of course. I follow this. Exactly.

12:45

And it already has billions of people using it. So it's like the adoption curve is going to just blow these instincts and towns of the world away as soon as it launches. This is right. There's these variables that really matter, right? Like Meta does not have trust, but does have business model, right? Google has trust and business model, right? I said this in passing, but it's worth recognizing, right? Like Meta is claiming, and they don't usually lie on this stuff. People might disagree with that. But they're claiming threads has surpassed X, which is true and a statement of the fact that distribution in certain products that wouldn't be the truth for every product is really, really meaningful. So I think that is significant.

12:48

Oh, yeah. A hundred percent. There's no one better in the world at distribution than Meta, period. They are the best in the world at distribution. So this product will get big. That's just true. And they also have amazing business model. And so I think those two things are massive advantages. They have much lower trust. And so when it comes to giving all of your personal information to a company. That's true. But OpenAI has low trust. Everyone has low trust. So it's of like the industry has come down to Meta's level. I don't know. I think you guys were talking about Google. Like Google does not have low trust. I guess that's true. Yeah.

13:05

I mean, I'm willing to give Google everything. And Apple. If Apple ever made a harness, I would also give it to Apple. I mean, it did, Britt. It's called Siri, right? And it's not as good yet. A functional harness. They're working on it. Guys, Siri's still, it's a source of infinite fun for me to see how little it can do. Okay. So we've got me. So the harness wars continue. Dave, what does this mean for OpenClaw?

13:33

Well, I mean, I think that OpenClaw is actually growing faster than ever. We had our biggest week last week. And I think what this means is that the world of agents is just expanding. I mean, I always tell everyone that it's kind of we're in 1982, 1983 in the personal computer and the Mac hasn't been invented yet. Everyone's creating new computers, trying to make a computer that's more usable for everyone. And racing towards can there be a consumer version, right? And can there be a consumer business model? Because the big prize is an agent that everybody in the world uses. And coding agents aren't going to go to everybody. Even ChatGPT is only at a billion scale when six billion people use mobile phone software. And so we haven't seen a breakthrough yet. I still think there's a ways to go. I think all these advantages that we're talking about are interesting. There's a lot of OpenClaw out there, right? And there's a lot of people using it. I heard last week that OpenClaw is being used to do the film editing for one of the top films at the Telluride Film Festival. And the whole film editing pipeline is being run by OpenClaw. So there's a lot of interesting stuff going on with agents. And we'll see. I think that the entire thing is exactly the two things that we're talking about. Number one, cost, right? Like the price of these things to run has to come down. And number two, user experience. There's immense work to do still with user experience. And number three, whether it's in the cloud or it's local. I still know a ton of people who are like, I am not giving my email to anything that's putting it in the cloud. I'm not putting my financials into anything that could be hacked. And you're seeing the hugging face hacks and everything else going on. And I know we're going to get to AI Doomsday soon. Which is our entire world at OpenClaw. That's our entire world is people that want to run it locally or enterprises. That's the other thing, Jess, is that I hear this every single day at OpenClaw, which is that I want open source models and our data on-prem. And to do that, we're going to use OpenClaw. And so there's a big wide world, a lot of agents, a lot of different reasons to use them.

13:35

Yep. Fun time. Speaking of film industry, sidebar information scoop this week that Mr. Jeffrey Katzenberg is launching a new AI venture with the engineer who was behind Sora, but obviously left. Smart. I didn't hear that. That's such a Katzenberg move. Dave, I texted it to you, but that's okay. I heard about it. But it's not Quibi 2. It is in that it's a big, ambitious Katzenberg, CJ venture. That's what I mean. Like, he's super ambitious. Of course somebody should do this. I like a good Katzenberg swing for the fence. I'm good with it.

14:23

Let me tell you, and all I know is what I read in the information. The vision is that Hollywood should control its own models and build models and control models that push the film industry forward, but on its own terms. And obviously there's a lot that can be with Bill Peebles, who is a Sora researcher, not engineer. And I just, anytime a media company or industry or sector is trying to control its own destiny with technology instead of bowing to the tech overlords, I'm in favor of it. So I wish them well. Apparently they're in talks with Andreessen Horowitz for a big check. Andreessen has no money. Come on.

14:30

It does seem important for what it's worth. This is a really important question. And I think that we're going to see a lot more of this, that you're going to see the proliferation of companies that do the fine tuning and training in particular verticals that are important. And Hollywood needs probably many of these companies. They need the ILMs of this era that can go and make the AI tools, particularly in video, really work. Well, isn't Sean trying to do this at Stability? Am I going to get my AI companies mixed up? At Sean Parker? Yeah, he's very involved in the Hollywood conversation, too. Also, isn't Eleven Labs? Didn't they mention they want to IPO now?

14:42

I mean, it's in two years. I did read that in the information. Did read that. Caught that. I thought they were running amok with lawsuits with all the music artists. It's Suno, Eleven Labs. The Trump administration said ignoring copyright is in our national interest. I don't know if you saw that. So, yeah. Oh, and so that's when they pushed the pedal on the IPO. Okay. Perhaps. I don't know. Not me saying that. But speaking of IPOs, guys, we are weeks away from Anthropic. The hum.

15:24

I don't know if you VCs feel it, but newsrooms are a hum, a buzz. Not because anyone's going to make a cent, but it's just an exciting time because readers are thirsty for any story about these companies in this big moment. Really? Yeah. I thought they were tired of it by now. No. No. Investors are like... Well, investors and LPs, of course, are waiting. Yeah. Did read that. Caught that. I thought they were running amok with lawsuits with all the music artists. It's Suno, Eleven Labs. The Trump administration said ignoring copyright is in our national interest. I don't know if you saw that. So, yeah. Oh, and so that's when they pushed the pedal on the IPO. Okay. Perhaps.

16:29

I don't know. Not me saying that. But speaking of IPOs, guys, we are weeks away from Anthropic. The hum. I don't know if you VCs feel it, but newsrooms are a hum, a buzz. Not because anyone's going to make a cent, but it's just such an exciting time because readers are thirsty for any story about these companies in this big moment. Really? Yeah. I thought they were tired of it by now. No. No. Investors are like... Well, investors and LPs, of course, are waiting. Yeah. They're a third of our readership. So, we love investors. Or finance is a third of our readership. But amid this walk-up, we have one Anthropic researcher who has decided what Jacob... What's his last name?

17:22

Coxson? Yes. Media outlet of the last 48 hours saying, guys, you have to watch the Anderson Cooper. All over the place. Even the Today Show. First of all, Jacob, are you doing your own booking or did you hire someone? Because I'm very impressed. You also spoke to the information. Thank you very much. But, guys, he thinks there's a 10% chance AI's going to kill us all. More than 10% chance. No, no, no. Someone that worked with him thinks that. He just said there's a chance. And this is scary. Somebody responded from Anthropic that said 10%. And he's gone on a media tour that even Taylor Swift would be envious of. What is going on here? Morrence, give us your take.

18:03

If it bleeds, it leads. Jess, everyone likes to hate on AI. They want to know AI sucks. They understand why AI adoption was 0.4% in the month of August. What does that even mean? It means it's plateaued. People aren't really adopting it right now outside of the folks who already have. Of what, though? I mean, don't tell Sam. Guys, my recursive Sam loops continue. Thanks for checking in on it. Sam is not one of the people who obviously hasn't adopted it. But, yeah, I think it's more of what we've been talking about. The sentiment of AI is worse than ICE and all the terrible Jeffrey Epstein and everything else. Now it's going to kill us all.

18:48

It says that we have until the end of the decade to survive, according to this Jacob guy. So people everywhere just want to sound the alarm. They want to get freaked out. This is the human experience. It's why everyone follows true crime and the Lindsay Clancy trial and everything else going on. It's bad news, crazy things. Yeah. It has this true crime. Yeah, you're right. Yeah. That's what it's all about. That's what the world loves right now. And it's sad. Well, there's two ways to look at this. One, this guy came out of nowhere. He's got no Twitter followers. And he quits Anthropic and goes on this, to your point, Jess, this media tour of all media tours.

19:40

One could look at this as incredibly coordinated. Oh, thank you, David Sachs. Yes, say more. You could look at it this way. If you go with the more conspiracy framing, which I'll then do the second framing. But the first framing is that this was a psyop that has been planned for a long time to create regulatory capture. Anthropic's messaging has been incredibly consistent on this. They've been saying this exact same thing for a long time. They have, exactly. And they're trying to convince the government. And I've been hearing this for months and months and months behind closed doors.

20:05

They've been trying to convince the White House and everybody under the sun to make open source illegal, to stop, stop, stop everything. And so what do you need to really make this happen? Three, four weeks ago, there was this open letter and every researcher under the sun signed it. And I remember on this pod saying, if you really believe this, then quit. Stop working on it. Yeah, we did say that. Yeah, you saw that. You could look at this as part of the plan that to get everyone's attention and to really cause lawmakers to activate. And so that's what this looks like because the scale of the media tour is crazy. Okay.

20:24

I will write that screenplay for whoever wants to make that movie. If you want to make that movie, I'll write the screenplay for you. Okay. That's take one. Take two is slightly different, which is I was playing with Astra last week and over the weekend. It was the first time where my experience of using one of these models, maybe for a year, maybe two years, felt marginally scary. Oh, no. Give us examples of what scared you. Well, it's really good. I don't think it's actually improved much in its ability to speak English language, all that kind of stuff. But it's extremely good at using computers. It's extremely good at 3D modeling.

20:48

Our son took it and made a 3D model in an hour that was then printing on his 3D printer that was scary fast. I was like, wow. This thing is operating in three-dimensional space at a speed and with a 12-year-old that if you were a really bad person, you could figure out ways to use this that are legitimately scary. And it was the first time where I had the experience of using one of these things where I was like, okay, I understand what those people were saying in that letter four weeks ago.

20:53

Because if you're working in one of the labs and you're using something like Astra or the next several models, it's teetering on the point now where these things are actually getting to a level of power. If you can run thousands of them or hundreds of thousands of them or millions of them at once, I get it. And so I do think the society-level conversation is getting to a point where we have to ask some serious questions about where we go. Now, is it the right thing to scare everyone by saying AI is going to kill us all? I don't think that's an effective way to have this conversation. But I don't know any other way, right? How do you get everybody's attention?

21:06

How do you have a conversation about what we should be doing? How we should be approaching this? I'm not working in the labs. I'm working pretty close. This year, I've been as close to the frontier as I've ever been. How does it feel? Nauseating. I'm joking. It's just a funny talk. I just don't know how to think about this. You know, we've talked about this on this pod for a lot of years, right? It does seem to be a moment for serious conversation. But how do you coordinate all of society on something this? I have no idea. I want to give props to this guy, without my conspiracy hat on for a second. I respect this guy for leaving and stopping working on it.

21:36

Rather than just signing a letter or whatever and still collecting his Anthropic stock, this guy was like, I'm out. This conversation really matters and I'm leaving my stock on the table. I don't know. My take on that was he had— In the last three years, he's been at OpenAI and Anthropic. He's cashed in a couple lottery tickets at each place and he's kind of like, I think I'm good. No, Britt, he did not take his Anthropic stock.

21:50

Oh, he didn't take the stock? No. It's not clear. He's left some on the table. I don't know if he said all. I don't know. There's been— No, he left two months before his cliff. Apparently, he left Anthropic on the table. But kept the OpenAI. Hmm. No, just kidding. Yeah. Well, he—yeah, he's already gone. Yeah. Everything you say makes a lot of sense, Dave. And on one hand, when you hear someone like Altman who now says once a day on some platform, podcast, or conference, we need to slow down the pace, you're like, okay, I think that's good. But there's no follow-up. We're going to let Sam Altman slow down the pace? What does that mean? It's a very confusing moment.

22:26

Well, they're asking for help, right? They have to compete in capitalism, right? And think about it. Anthropic—I don't even know what the number is, but the sheer amount of capital that went into Anthropic in the last 12 months is like—

22:26

Apparently, he left the Anthropic on the table. But kept the OpenAI. No, just kidding. Yeah. Well, he—yeah, he's already gone. Yeah. Everything you say makes a lot of sense, Dave. And on one hand, when you hear someone like Altman who now says once a day on some platform, podcast, or conference, we need to slow down the pace, you're like, okay, I think that's good. But there's no follow-up. We're going to let Sam Altman slow down the pace? What does that mean? It's just a very confusing moment. Well, they're asking for help, right? They have to compete in capitalism, right? And think about it. Anthropic—I don't even know what the number is, but the sheer amount of capital that went into Anthropic in the last 12 months is—

22:28

$17 gazillion, yes.

22:33

It's so far beyond. It's an unbelievable amount of capital. And that capital wants to grow. One of my great mentors always told me, Dave, capital wants to get a return. That capital wants to grow. It's going to continue trying to grow. By the way, if Anthropic's ability to grow goes down at all, all of that capital that came in in the last 12 months is going to be underwater. And so people are betting their careers. There are so many people riding on this, right? And so the forces at work of the capital machine are so powerful. And this is a massively amplified version of it. So these guys are, hey, we're using this thing internally, and it's pretty crazy. We can't slow down because we have to keep creating the returns for the investors. Well, they also want to—we're not going to go deep into the Millennium Prize win-upmanship in this pod, but they also want to be the one to claim AGI, which OpenAI claimed. Here is a fun fact. Speak of conspiracy theories.

22:33

OpenAI, which has had many different charters and has many different corporate structures, had a provision that has been in place and, as far as I understand, remains in place, that says if someone else develops AGI first, we have to merge with them to support their effort. So this has existed. What? Really? I heard about this. But Greg Brockman claimed they were first to AGI, so now it's been rendered— What is the qualification to be at AGI? I mean, it's a meaningless term anyway.

22:39

It's totally meaningless. But if you have that provision, because Sam Altman, nine years ago, when it was a nonprofit with Elon or whatever, I'm getting my timelines off, oh, yeah, this will be good. If someone else gets there, we'll help them. You don't want to get rid of it because it makes you look like an evildoer. What would you do? You would self-declare that you won the race. Of course you would. Rendering such provision obsolete.

22:44

Of course you would. So I'm glad that this is working out. But to be fair, Jess, Astra is an insane leap forward. It's so far in front of Fable and everything. I actually think that's an interesting story of the week. I think this probably shocked people a little bit. It's way ahead. You guys are saying that. Everyone else who's played with it, who I've spoken to, is saying that. That seems like a clear consensus. People are impressed. This feels like the ultimate tennis match where I'm watching it go back from one side to the other side. One side to the other side.

22:59

That's why you have to sit in section 117, Britt. Then you just don't do that. You gotta go with me. You gotta go with you. What else is happening, guys? What else? Oh, should we do some plugs to the new Venmo documentary you guys were at? I'd love to let people know about this. Yeah, I got to see Dave and Sam on stage together. My husband didn't tell me about this, so that's why I wasn't there. Just...

23:08

It was very sweet. They were very cute together, these two boys. They were recalling stories of the late 2000s, early 2010s, of how the idea of Venmo came to be. Obviously, our dear friends, Ikram and Andrew Cortina were the co-founders of that. And it started with just, we're at dinner. We were at dinner at the Ace Hotel at the Breslin. And someone was, ugh, it's annoying how I have to pay the bill, but then I have to now ask all of you guys to pay me back. And then you have to go find a way to PayPal me or something. And that takes forever. I wish you could just text me a payment. Those Breslin burgers were expensive. I'm thinking back. Ace Hotel.

23:16

I love those burgers. I have a New York trip coming up, and I haven't stayed at the Ace in a while, but New York hotel prices are astronomical for $150. Per square feet. So I'm thinking, why don't I go a little more spacious and stay at the Ace and get a nice discount? Well, who knows? I did the math on this, Jess, and New York hotel prices have outpaced inflation by 2x over the last 10 years.

23:18

Sam's staying with our friend. I should probably not reveal his location because he didn't book until he left. He's staying with a mutual friend of ours. Sam was tweeting about it, and so I had my agent do the math because I was curious, you know, because I always stay at the same place. What's your stay? Unless it's a state secret. The Greenwich is my absolute favorite. Oh, too far downtown. My office is in Flatiron. No, Tribeca's amazing. Jess, the Greenwich is amazing. I've stayed at the Greenwich many times, but I need to be near the office. Greenwich is the best hotel in the world. Absolutely not. Oh, my God, Dave. It's the best hotel in the world.

23:38

I disagree, but it is nice. Yeah, the Greenwich is amazing. Oh, my God. I think I'm going back to the Ace, guys. I'm going to relive our 20s. We spent so much. I'm going to go, and I'm going to get a king bed this time. You know, I'm really going to splurge. You guys had a wedding party at the Ace. This is a hill I'll die on. We had an engagement party at the Ace Hotel. Yeah, we spent a lot of time there. We had a friend who got married at the Ace.

24:05

The Ace was the place in the 2000s. It's like Malcolm Gladwell was writing books. John Mayer was there. It was this weird, millennial creative hangout. And if you remember, Koreatown was pretty shady back then. But then Milk Bar opened, and it was cool again. Okay, back to Venmo, you guys. Okay, yes. So you guys are on a panel, and there is a movie coming out. Is that correct?

24:12

Yeah, there's a documentary about the founding story. Colin Luce from Basis Theory, which is a fintech startup that we're involved with, has decided to start doing a documentary series on important moments in fintech. And he asked me if I knew the founders and Sam and everybody. And so Colin did an immense amount of work and put together an entire documentary on the Venmo story. It's still being worked on. But this is a professional, amazing documentary that will come out sometime probably in the next year. And pretty cool. A lot of us sat down for it. They went around and interviewed everybody, people from Braintree, people from PayPal, people from the early days. And so we did a conversation at the SF Jazz Center this week on stage with Sam and Ikram.

24:14

I wish I'd been invited. Sounds delightful. It was a cool vibe. There'll be another one, Jess, when the film is ready to come out. It made me nostalgic for the simpler times in tech that happened 20 years ago. I know. It was so similar. No harnesses, right? We just did meetups and hung out at the Ace. And the people that were here in the Valley were here to just be creative and build cool shit. And I can't say that for everyone that's here right now.

24:23

It's a fun time to be here, too. I think Sam made some great points about this. Just that in the 2000s and somewhat the early 2010s, people in tech were making things because whatever they were working on was awesome. They were doing it for fun because they sort of loved the creative endeavor of it. And that has changed. There is a lot of people making things in the Valley now to just make money. There'll be another one, Jess, when the film is ready to come out. It made me nostalgic for the simpler times in tech that happened 20 years ago. I know. It was so similar. No harnesses, right? We just did meetups and hung out at the Ace.

24:38

And the people that were here in the Valley were here to just be creative and build cool shit. And I can't say that for everyone that's here right now. It's a fun time to be here, too. I think Sam made some great points about this. Just that in the 2000s and somewhat the early 2010s, people in tech were making things because whatever they were working on was awesome. They were doing it for fun because they loved the creative endeavor of it. And that has changed. There is a lot of people making things in the Valley now to just make money. They also wanted to make money back in the day. But yes, the different ethos and parts of it.

24:53

But back in the day, a $30 million acquisition was the talk of the town. It was like Flickr got acquired for $20 million and everyone lost their shit. But it is amazing. Venmo ended up not selling for a relatively large amount of money if I... No, none of us made any money. Yeah. I use it twice a day to this day. It's extraordinary on its cultural impact. And one of the great stories that they were telling that I just thought was hilarious that I had forgotten about was this story of in the time period between when Braintree and PayPal were getting acquired, Venmo wanted to do a marketing campaign. And so they were asking for money to do marketing.

25:17

And the CEO was like, no, we can't do this right now. We're in the middle of acquisition. And so I think it was Cortina walked into the break room one day and was like, we should just do a marketing campaign about Lucas. And Lucas was a random engineer on the Venmo engineering team. And so they put these billboards all over the subways and everywhere. And it was just like, Lucas takes the subway. And they didn't even say Venmo on them. And they were all over the city. But it went massively viral. It went massively viral. That was cool. And it caused Venmo to grow enormously with almost no marketing budget.

26:03

And so I think one of the cool underlying stories of Venmo was this really creative ingenuity that they had with limited resources to convince people to use this product. And, Jess, you're right. Britt and I were at Stagecoach Festival last year. And it was Venmo everywhere. And I was just like, man, the ongoing cultural impact of this, the unlikely impact of this crazy little company still to this day is pretty amazing. Well, the irony, I think, is also that Stripe wanting to acquire PayPal for Venmo. So that zero-dollar value that Dave and Sam got has definitely grown over time. Yeah. Okay, we're going to wrap in a sec.

26:33

But we should, speaking of movies, do you guys—by my count, in the next several months, we are going to have an Altman movie. A Musk movie by Alex Gibney. A Social Reckoning movie about Facebook slash Meta. Oh, really? I didn't know that. And a Theranos movie, guys. Yeah. Before she went to jail, she did Nathan for You. Insane. Whatever his name is. Did Ben Stiller produce that or something? He's involved somehow. I mean, probably. So I don't know. Who wants to do some screenings? This is sort of what I... Oh, that's a fun idea. We should make fancy events in SF with red carpet and things like that. But for tech movies.

27:37

There's also apparently a movie which, you know, I asked Jesse Sittig, who's our best filmmaking friend—she's the spirit animal of the pod—it won't be this to come on. I asked her what the best film from Telluride was, and she said Wild Horse 9. And I watched the trailer, and it looks amazing. It's like John Malkovich, Steve Buscemi. It looks amazing. And she's like, it's going to win everything. And so that looks cool, too. Speaking of watching, guys, happy U.S. Open Finals weekend. We're here. We are now in the semifinals. We have four Americans who have shots. Who's going to win, Jess? Who is going to win? I really hope Ben Shelton wins after toppling my always.

28:27

I'm an Alcaraz ride or die. For all of us West Coasters who stayed up late, the East Coasters were staying up until 4 a.m. in the morning watching an epic five-setter. I really hope Shelton goes on to win it. And I'm rooting for Pagula or Coco. I'm rooting for Coco. But I don't know if they can pull it off if she can pull it off in the end. But it's been epic. The tennis blogs are all aflutter with the how dare tennis matches last until 3:30 in the morning. I kind of say if they're great matches, folks, what are you going to do? Make the sport shorter to have less drama? Why is that? I saw some of that drama, but it actually just made me think, why is that?

29:09

Why do they go till? Isn't that hard on the athletes? It's very hard on the athletes. But there is an explanation, which is it usually happens when you have a combination of very long, tight, competitive matches that go long with tiebreaks. And the USTA argues you could, in theory, start earlier. But it's actually hard to move 25,000 people around and have multiple sessions. So I am not an expert on this. And I think maybe some adjustments will be made. But I also think the other side of this is you're watching winner after winner and tiebreak after tiebreak. And that's what tennis is about.

29:29

So is that the story that if center court goes super long during the day, then the match starts later because of that? It can. But what's basically happening is that the night sessions are going quite long. And then you also can get in a rhythm because there's rules how frequently players can play. So play, I don't know exactly what it is, but if someone was on until 4 a.m., you can't put them on at 6 p.m. And so there— That makes sense. I am no expert in this, and my tennis friends are laughing at me right now. But it's almost got as much attention in the tennis world as Jacob did for his doomsdayer. So every outlet has been writing an op-ed about 4 a.m.

29:56

And then I have two tennis-watching threads. I have an East Coast WhatsApp group and a West Coast iMessage group. And so they were really enjoying the back and forth. But anyway, for tennis lovers, it should be a fun weekend and some great matches, I hope. So fingers crossed. Okay, Morinz, what else have we got? Did we do it? Did we give the people some smarts for the week? That's it. I think we did it. That's all the smarts we have. We covered a lot of ground. It's always a pleasure and relatively calm being here with you two, although we'll miss Sam. We miss the frenetic Sam energy. He'll be back next week.

30:22

I could use a little help on the home front, although I'm keeping things going. I've got to do three back-to-school nights as one person. Let's see if I can pull this off. But with that, dear listeners, we wish you a very happy weekend, and we'll see you back here next week for another episode of More or Less. Bye. Bye. See you later. If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for @moreorlss, @DaveMorin, @Lesson, @JayLesson. And as for me, I'm @Brit.

30:43

See you guys next time. Bye. And he's gone on a media tour that even Taylor Swift would be envious of. What is going on here? Morrence, give us your take. If it bleeds, it leads. Jess, everyone likes to hate on AI. They want to know AI sucks. They understand why AI adoption was 0.4% in the month of August. What does that even mean? It means it's, like, plateaued. Like, people aren't really adopting it right now outside of the folks who already have. Of what, though? I mean, don't tell Sam. Guys, my recursive Sam loops continue. Thanks for checking in on it. Sam is not one of the people who obviously hasn't adopted it.

31:26

But, yeah, I think it's just more of what we've been talking about. The sentiment of AI is worse than ICE and all the terrible Jeffrey Epstein and everything else. Now it's going to kill us all. It says that we have until the end of the decade to survive, according to this Jacob guy. So people everywhere just want to sound the alarm. They want to get freaked out. This is, like, the human experience. It's, like, why everyone follows true crime and the Lindsay Clancy trial and everything else going on. It's bad news, crazy things. Yeah. It has this true crime. Yeah, you're right. Yeah. That's what it's all about. That's what the world loves right now. And it's kind of sad.

32:04

Well, there's two ways to look at this. One, this guy came out of nowhere. He's got, like, no Twitter followers. And he quits Anthropic and goes on this, to your point, Jess, this media tour of all media tours. One could look at this as incredibly coordinated. Oh, thank you, David Sachs. Yes, say more. You could look at it this way. If you go with the more conspiracy sort of framing, which I'll then do the second framing. But the first framing is that this was a psyop that has been planned for a long time to create regulatory capture. Anthropic's messaging has been incredibly consistent on this. They've been saying this exact same thing for a long, long time.

32:47

They have, exactly. And they're going and they're trying to convince the government. And I've been hearing this for months and months and months behind closed doors. They've been trying to convince the White House and everybody under the sun to make open source illegal, to stop, stop, stop everything. And so what do you need to really make this happen? Three, four weeks ago, there was, like, this open letter and every researcher under the sun signed it. And I remember on this pod saying, if you really believe this, then quit. Stop working on it. Yeah, we did say that. Yeah, you saw that.

33:19

You could look at this as a part of the sort of plan that to get everyone's attention and to really cause lawmakers to activate. And so that's what this kind of looks like because the scale of the media tour is crazy. Okay. I will write that screenplay for whoever wants to make that movie. If you want to make that movie, I'll write the screenplay for you. Okay. That's take one. Take two is slightly different, which is I was playing with Astra last week and over the weekend. It was the first time where my experience of using one of these models, maybe for a year, maybe two years, felt marginally scary. Oh, no. Give us examples of what scared you. Well, it's really good.

34:05

I don't think it's actually improved much in its ability to speak English language, all that kind of stuff. But it's extremely good at using computers. It's extremely good at 3D modeling. Our son took it and made a 3D model in an hour that was then printing on his 3D printer that was, like, scary fast. Like, I was like, wow. Like, this thing is operating in three-dimensional space at a speed and with a 12-year-old that if you were a really bad person, you could figure out ways to use this that are legitimately scary.

34:38

And it was the first time where I had the experience, I guess, of using one of these things where I was like, okay, I understand what those people were saying in that letter four weeks ago. Because if you're working in one of the labs and you're using something like Astra or the next several models, it's teetering on the point now where these things are actually getting to a level of power. If you can run thousands of them or hundreds of thousands of them or millions of them at once, I get it. And so I do think the society-level conversation is getting to a point where we have to ask some pretty serious questions about where we go.

35:17

Now, is it the right thing to scare everyone by saying, like, AI is going to kill us all? I kind of think that's, like, not really an effective way to have this conversation. But I don't know any other way, right? Like, how do you get everybody's attention? How do you have a conversation about what we should be doing? How we should be approaching this? I'm not working in the labs. I'm working pretty close. This year, I've been as close to the frontier as I've ever been. How does it feel? Nauseating. I'm joking. It's just a funny talk. I just don't know how to think about this. You know, we've talked about this on this pod for a lot of years, right?

35:51

It does seem to be, like, a moment for serious conversation. But how do you coordinate all of society on something like this? Like, I have no idea. I kind of want to give props to this guy, like, without my conspiracy hat on for a second. I respect this guy for, like, leaving and stopping working on it. Rather than just, like, signing a letter or whatever and, like, still collecting his Anthropic stock, this guy was like, I'm out. Like, this conversation really matters and I'm leaving my stock on the table. I don't know. My take on that was he had— In the last three years, he's been at OpenAI and Anthropic.

36:24

He's cashed in a couple lottery tickets at each place and he's kind of like, I think I'm good. No, Britt, he did not take his Anthropic stock. Oh, he didn't take the stock? No. It's not clear. He's left some on the table. I don't know if he said all. I don't know. I—again, there's been— No, he left two months before his cliff. Apparently, he left the Anthropic on the table. But kept the OpenAI. Hmm. No, just kidding. Yeah. Well, he—yeah, he's already gone. Yeah. Everything you say makes a lot of sense, Dave.

36:51

And on one hand, when you hear someone like Altman who now says once a day on some platform, podcast, or conference, we need to slow down the pace, you're like, okay, I think that's good. But there's no follow-up. We're going to let Sam Altman slow down the pace? Like, what does that mean? Like, it's just a very confusing moment. Well, they're asking for help, right? Like, they have to compete in capitalism, right? Like, and think about it. Anthropic—I don't even know what the number is, but the sheer amount of capital that went into Anthropic in the last 12 months is like— $17 gazillion, yes.

37:30

It's so far beyond. It's an unbelievable amount of capital. And that capital wants to grow. One of my great mentors always told me, Dave, capital wants to get a return. That capital wants to grow. It's going to continue trying to grow. By the way, if Anthropic's ability to grow goes down at all, like, all of that capital that came in in the last 12 months is going to be underwater.

37:51

And so people are, like, have bet their careers. There's, like, so many people riding on this, right? And so the forces at work of the capital machine are so powerful. And this is, like, massively amplified version of it. So these guys are kind of like, hey, we're using this thing internally, and it's pretty crazy. We can't slow down because we have to keep creating the returns for the investors. Well, they also want to, like—we're not going to go deep into the Millennium Prize win-upmanship in this pod, but they're also—they want to be the one to claim AGI, which OpenAI claimed. Here is a fun fact. Speak of conspiracy theories.

38:29

OpenAI, which has had many different charters and has many different corporate structures, had a provision that has been in place and, as far as I understand, remains in place, that says, if someone else develops AGI first, we have to merge with them to support their effort. So this has existed. What? Really? I heard about this. But Greg Brockman claimed they were first to AGI, so now it's been rendered— What is the qualification to be at AGI? I mean, it's a meaningless term anyway.

39:02

It's just totally meaningless. But if you have that provision, because, like, Sam Altman, nine years ago, when it was a nonprofit with Elon or whatever, I'm getting my timelines off, like, oh, yeah, this will be good. If someone else gets there, we'll help them. You don't want to get rid of it because it makes you look like an evildoer. What would you do? Oh, you would self-declare that you won the race. Of course you would. Rendering such provision obsolete. Of course you would. So I'm glad that this is working out.

39:31

But to be fair, Jess, Astra is, like, an insane leap forward. It's so far in front of Fable and everything. Like, I actually think that's kind of an interesting story of the week, which is that I think this probably shocked people a little bit. It's, like, way ahead. You guys are saying that. Everyone else who's played with it, who I've spoken to, is saying that. Like, that seems like a clear consensus. People are impressed. This feels like the ultimate tennis match where I'm watching it go back from one side to the other side. One side to the other side.

40:01

That's why you have to sit in section 117, Britt. Then you just don't do that. You gotta go with me. You gotta go with you. What else is happening, guys? What else? Oh, should we do some plugs to the new Venmo documentary you guys were at? I'd love to let people know about this. Yeah, I got to see Dave and Sam on stage together. My husband didn't tell me about this, so that's why I wasn't there like that. Just...

40:24

It was very sweet. They were very cute together, these two boys. They were recalling stories of the late 2000s, early 2010s, of how the idea of Venmo came to be. Obviously, our dear friends, Ikram and Andrew Cortina were the co-founders of that. And it started with just, like, we're at dinner. We were at dinner at the Ace Hotel at the Breslin. And someone was like, ugh, it's annoying how I have to pay the bill, but then I have to, like, now ask all of you guys to pay me back. And then you have to, like, go find a way to PayPal me or something. And that takes forever. I wish you could just text me a payment.

41:00

Those Breslin burgers were expensive. I'm thinking back. Ace Hotel. I love those burgers. I have a New York trip coming up, and I haven't stayed at the Ace in a while, but New York hotel prices are, like, astronomical for, like, $150. D square feet. So I'm like, why don't I go a little more spacious and stay at the Ace and get a nice discount? Well, who knows? I did the math on this, Jess, and New York hotel prices have outpaced inflation by 2x over the last 10 years. Sam's staying with our friend. I should probably not reveal his location because he didn't book until he left. He's staying with a mutual friend of ours.

41:35

Sam was tweeting about it, and so I had my agent do the math because I was curious, you know, because I always stay at the same place. What's your stay? Unless it's a state secret. The Greenwich is my absolute favorite. Oh, too far downtown. My office is in Flatiron. No, Tribeca's amazing. Jess, the Greenwich is amazing. I've stayed at the Greenwich many times, but I need to be near the office. Greenwich is the best hotel in the world. Absolutely not. Oh, my God, Dave. It's the best hotel in the world. I disagree, but it is nice. Yeah, the Greenwich is amazing. Oh, my God. I think I'm going back to the Ace, guys. I'm going to relive our 20s.

42:10

We spent so much. I'm going to go, and I'm going to get, like, a king bed this time. You know, I'm really going to spot. You guys had, like, a wedding party at the Ace. This is a hill I'll die on. We had an engagement party at the Ace Hotel. Yeah, we spent a lot of time there. We had a friend who got married at the Ace. The Ace was, like, the place in the 2000s. Like, it's like Malcolm Gladwell was writing books. John Mayer was there. Like, it was like this weird, you know, millennial creative hangout. And if you remember, Koreatown was pretty shady back then. But then Milk Bar opened, and I was just like, it's like, it was cool again. Okay, back to Venmo, you guys.

42:46

Okay, yes. So you guys are on a panel, and there is a movie coming out. Is that correct? Yeah, there's a documentary about the founding story. Colin Luce from Basis Theory, which is a fintech startup that we're involved with, has decided to start doing a documentary series on sort of important moments in fintech. And he asked me if I knew the founders and Sam and, you know, everybody. And so Colin did an immense amount of work and put together an entire documentary on the Venmo story. It's still being worked on. But, I mean, this is like a professional, amazing documentary that will come out sometime probably in the next year. And pretty cool.

43:26

Like, a lot of us sat down for it. They went around and interviewed everybody, people from Braintree, people from PayPal, people from the early days. And so we did a conversation at the SF Jazz Center this week on stage with Sam and Ikram. I wish I'd been invited. Sounds delightful. It was a cool vibe. There'll be another one, Jess, when the film is ready to come out. It made me nostalgic for the simpler times in tech that happened 20 years ago. I know. It was so similar. No harnesses, right? We just did meetups and, like, hung out at the Ace. And the people that were here in the Valley were here to, like, just be creative and build cool shit.

44:07

And I can't say that for everyone that's here right now. It's a fun time to be here, too. I think Sam made some great points about this. Just that in the 2000s and somewhat the early 2010s, like, people in tech were making things because whatever they were working on was awesome. They were doing it for fun because they sort of loved the creative endeavor of it. And that has changed. Like, there is a lot of people making things in the Valley now to just make money. They also wanted to make money back in the day. But, yes, the different ethos and parts of it. But back in the day, a $30 million acquisition was, like, the talk of the town.

44:44

It was like Flickr got acquired for $20 million and everyone lost their shit. But it is amazing. You know, Venlo ended up not selling for a relatively large amount of money if I... No, none of us made any money. Yeah. I use it twice a day to this day. It's extraordinary on its cultural impact. And one of the great stories that they were telling that I just thought was hilarious that I had forgotten about was this story of in the time period between when Braintree and PayPal were getting acquired, Venmo wanted to do a marketing campaign. And so they were asking for money to do marketing. And the CEO was like, no, we can't do this right now.

45:20

We're, like, in the middle of acquisition. And so I think it was Cortina walked into the break room one day and was like, we should just do a marketing campaign about Lucas. And Lucas was, like, a random engineer on the Venmo engineering team. And so they put these billboards all over the subways and everywhere. And it was just like, Lucas takes the subway. And they didn't even say Venmo on them. And they were all over the city. But it went massively viral. It went massively viral. That was cool. And it caused Venmo to, like, grow enormously with almost no marketing budget.

45:53

And so I think one of the cool underlying stories of Venmo was, like, this really creative ingenuity that they had with limited resources to convince people to use this product. And, Jess, you're right. Like, Britt and I were at Stagecoach Festival last year. And it was, like, Venmo everywhere. And I was just like, man, the ongoing cultural impact of this, the unlikely impact of this crazy little company still to this day is pretty amazing. Well, the irony, I think, is also that Stripe wanting to acquire PayPal for Venmo. So that zero-dollar value that Dave and Sam got has definitely grown over time. Yeah. Okay, we're going to wrap in a sec.

46:33

But we should, speaking of movies, do you guys, by my count, in the next several months, we are going to have an Altman movie. A Musk movie by Alex Kidney. A Social Reckoning movie about Facebook. Slash Meta. Oh, really? I didn't know that. And a Theranos movie, guys. Yeah. Before she went to jail, she did Nathan for You. Insane. Whatever his name is. Did Ben Stiller produce that or something? He's involved somehow. I mean, probably. So, I don't know. Like, who wants to do some screenings? This is sort of what I... Oh, that's a fun idea. We should make fancy events in SF with red carpet and things like that. But, like, for tech movies.

47:15

There's also apparently a movie which, you know, I asked Jesse Sittig, who's our best filmmaking friend. She's the spirit animal of the pod. It won't be this to come on. I asked her what the best film from Telluride was, and she said Wild Horse 9. And I watched the trailer, and it looks amazing. It's like John Malkovich, Steve Buscemi. It looks just, like, amazing. And she's like, it's going to win everything. And so, that looks cool, too. Speaking of watching, guys, happy U.S. Open Finals weekend. We're here. We are now in the semifinals. We have four Americans who have shots. Who's going to win, Jess? Who is going to win?

47:56

I really hope Ben Shelton wins after toppling my always. I'm an Alcaraz ride or die. For all of us West Coasters who stayed up late, the East Coasters were staying up until 4 a.m. in the morning watching an epic five-setter. I really hope Shelton goes on to win it. And I'm rooting for Pagula or Coco. I'm rooting for Coco. But I don't know if they can pull it off, if she can pull it off in the end. But it's been epic. The tennis blogs are a flutter with the how dare tennis matches last until 3.30 in the morning. I kind of say if they're great matches, folks, what are you going to do? Make the sport shorter to have less drama? Why is that?

48:36

I saw some of that drama, but it actually just made me think, like, why is that? Why do they go till? Isn't that hard on the athletes? It's very hard on the athletes. But there is an explanation, which is it usually happens when you have a combination of very long, tight, competitive matches that go long with tie breaks. And the USTA argues you could, in theory, start earlier. But it's actually hard to move 25,000 people around and have multiple sessions. So I am not an expert on this. And I think maybe some adjustments will be made. But I also think the other side of this is, like, you're watching winner after winner and tie break after tie break.

49:15

And that's kind of what tennis is about. So is that the story that if, like, if center court goes super long during the day, then the match starts later because of that? It can. But what's basically happening is that the night sessions are going quite long. And then you also can get in a rhythm because there's rules how frequently players can play. So play, I don't know exactly what it is, but, like, if someone was on until 4 a.m., you can't put them on at 6 p.m. And so there— That makes sense. I am no expert in this, and my tennis friends are laughing at me right now. But it's almost got as much attention in the tennis world as Jacob did for his doomsdayer.

49:52

So every outlet has been writing an op-ed about 4 a.m. And then I have two tennis-watching threads. I have an East Coast WhatsApp group and a West Coast iMessage group. And so they were really enjoying the back and forth. But anyway, for tennis lovers, it should be a fun weekend and some great matches, I hope. So fingers crossed. Okay, Morinz, what else have we got? Did we do it? Did we give the people some smarts for the week? That's it. I think we did it. That's all the smarts we have. We covered a lot of ground. It's always a pleasure and relatively calm being here with you two, although we'll miss Sam. We miss the frenetic Sam energy. He'll be back next week.

50:28

I could use—there's a little help on the home front, although I'm keeping things going. I've got to do three back-to-school nights as one person. Let's see if I can pull this off. But with that, dear listeners, we wish you a very happy weekend, and we'll see you back here next week for another episode of More or Less. Bye. Bye. See you later. If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for at more or less, at Dave Morin, at Lesson, at Jay Lesson.

51:11

And as for me, I'm at Brit. See you guys next time. Bye.

Reading tools

Type to find a passage

Appearance
Ask this transcript

Add a note