The AI Reset Is Here: Search, Jobs, and Everything Else w/ Anish Acharya, Dave Blundin, Salim Ismail

Music composition has become 1,000 times easier in the past 20 years due to technology (synthesizers, DAWs, AI), resulting in vastly more music being created; the question is whether this is abundance of quality or abundance of garbage, but the abundance itself is what feeds exponential flourishing.

“If I think about what it takes for somebody to compose complex music today, it's a thousand times easier than 20 years ago and you just get that much more music. I think that's what feeds into the abundance thing. We just can create so much more.”

Multimodal consumer-facing AI applications (voice, vision, multimodal) consume 2-4 GPUs concurrently to achieve optimal user experience, driving demand that exceeds supply by orders of magnitude.

“some of these more uh consumerf facing use cases that involve voices and now multimodal imagery um you know they'll use up an entire GPU, two GPUs, four GPUs concurrently to get the best possible user experience. And that's not super expensive. Uh so it's easily worth it for the consumer. But the chips don't exist.”

A Jones Act for AI may emerge requiring AI systems that operate in sensitive contexts to be trained nationally, creating a parallel to maritime sovereignty frameworks.

“There's something very interesting here which almost looks I'm not sure if you guys are familiar with the Jones Act for ships that operate in and around US ports. They must be manufactured in the United States for national security purposes. I believe we'll see a sort of Jones Act for AI where AI that operates, you know, in sort of sensitive contexts will have to have been trained uh nationally.”

We will produce 20 million GPUs in 2025, but this is nowhere near sufficient to meet demand from just basic call center use cases, creating an acute shortage.

“We're going to make 20 million new GPUs this year. It's nowhere near enough to keep up with just basic call center use cases.”

Grok 3.5 focuses on first-principles physics reasoning and aspiring to truth with minimal error, prioritizing honesty as a safety principle for AI systems.

“the focus of of GRA 3.5 is um uh sort of find the fundamentals of physics um and and applying physics tools across uh all lines of reasoning um and to aspire to truth with uh minimal error. Like there's always going to be some mistakes that are made uh but aim to to uh get to truth with acknowledged error uh but minimize that error over time.”

The 2025-2026 period will see an agentic AI boom where GPT-5 multimodal models drive rapid codegen flooding of consumer apps, followed by a shift toward vertical AI systems specialized for specific use cases rather than horizontal general-purpose agents.

“2025 to 2026 the year ahead it's the agentic AI boom. So GPT05 is multimodal it's coming out quad 4 uh codegen floods uh apps... I think the agentic is going to give way to vertical because there's so much opportunity in training vertical AIs on very specific use cases.”

Driver is the #1 job title globally; most human labor involves routine tasks (driving, assembly, mechanical work), creating massive job displacement risk as robotics and AI advance.

“The number one uh job in the world is driver. That's the number one title. But you know then you look down the longtale of other things that people do and and just a huge fraction of humanity is doing things as routine as driver or you know screw circuit board in or assemble laptop lid or it just goes on and on and on”

Nvidia's market capitalization exceeding both Meta and Google despite Google inventing the transformer architecture reflects the dominance of chip supply constraints in determining AI market dynamics, with hardware becoming the binding constraint on capability realization.

“One thing also you know Nvidia now today is worth as of right now over twice as much as Meta and almost twice as much as Google. Uh and you're like well how can that be? And you know especially Google where you know the transformer was invented at Google uh you know Google cloud GCP is huge they have their own TPU7s which are incredible like does this make any sense and that's you know that's debatable uh but the chip demand is such a dominant factor and then underneath that the fab shortage is such a dominant factor”

Large language models are effective at interpolating (inferring patterns between known data points) but not at extrapolating (predicting beyond the training data), which means even if trained on physics first principles, they will fill in blind spots based on assumptions rather than true reasoning.

“the LLMs that were using transformer-based are very good at interpolation, not quite as good at extrapolation. So it's going to fill in the blind spots between the training data that you give it and don't give it. I love the message of give it, you know, first principles physics, but I think everyone's going to do that.”

The interconnect between GPU clusters (NVLink spine) is becoming a dominant technical constraint and innovation focus, with Nvidia's 130 terabytes/second spine transferring more data than peak internet traffic, which drives both performance and cooling infrastructure requirements.

“One spine of Nvidia's uh NVLink fusion can transfer more data than the internet... 130 terabytes per second... the peak traffic of the entire internet was 900 terabits per second... This moves more traffic than the entire internet.”

AI agents performing microtransactions (agent-to-agent payments) enabled by stable coins represents the most important economic activity by 2040—larger than current human economic activity.

“the ability to move back and forth uh instantaneously digitally from a stable coin back to Bitcoin or whatever you want is intimately tied to the agent to agent uh microtransactions... that's going to be we get Kathy Wood to do the math for us but that's going to be probably the biggest part of the economy by 2040”

Technology for the past 40 years has extended our intellects through tools like spreadsheets and word processors, but AI is the first technology that can address the subjective, emotional, and relational aspects of human experience because it engages both left-brain analytical and right-brain emotional processing.

“If you look at technology for the last 40 years, it's really extended our intellects, right? And Steve Jobs famously said it's a bicycle for the mind. But when he said a bicycle for the mind, he really meant a bicycle for the intellect... However, we haven't done that much for our sort of souls or for our emotions or for our mindset. And with AI, we're able to explore the side of humanity that is sort of defined by the sort of subjective emotional experience.”

Teaching people to think as entrepreneurs and find problems to solve is critical preparation for AI displacement—expanding from employee to creator mindset.

“There's another thing as well, which is getting your employees and your kids and your friends to start thinking as entrepreneurs, right?... They can start to create uh you know new capabilities, new companies, new nonprofits. They can start to dream at a level like never before is is an unleashing of the human spirit”

More Americans now own Bitcoin than own gold, representing democratization of asset ownership—Bitcoin is easier to acquire than physical gold.

“Bitcoin is becoming America's reserve asset. More Americans own Bitcoin than gold." That's pretty extraordinary. I think this was kind of predictable just because it's so much more democratized. It's so much a thousand times easier to own Bitcoin than it is to own gold.”

By 2029-2031, AI will eliminate the need for white-collar work as currently defined, creating massive job displacement over the next 2-3 years that corporate leaders are dramatically underestimating.

“that's the end point, but it's pretty much a straight line between here and there. Mhm. And so the amount of job dislocation, you know, in 2026, 2027 is going to be like nothing we've ever seen. And I keep telling all the CEOs, you're way under planning. You need to look at every single person in your organization, all the individual contributors doing white collar work, and you need to get them to become AI users right now.”

Disposable code and disposable content (songs, theme music) created by AI at low cost (12 cents per song) will become standard practice, enabling companies to generate customized, single-use creative assets that would be economically impossible with human creation.

“the software created this this thing specifically for the podcast. It uses a fair amount of compute, but it's still like 12 cents. No big deal. You create it, then you throw it away... I can think of a lot of those things. Let me spawn all those agents to create them, and I'm going to throw them away when they're done. That's a lot of compute.”

AI models trained on creative domains like music and art that were missing from training data (e.g., hip-hop when trained only on pre-hip-hop music) cannot infer or predict those domains because culture and music are adaptive systems that emerge from human creativity rather than deterministic patterns.

“these systems are very good at predicting static systems. They're very good at sort of averaging the training data and telling you what the training data implies. They're not very good at predicting adaptive systems like the stock market or even culture and music. So, as a thought experiment, if you trained an AI model with all the music, you know, right up to hip-hop, but not including hip-hop, would it infer would it imply hip-hop? I don't think so, because culture and music is this sort of adaptive system that works together.”

Technical founders are experiencing a resurgence in startup success rates because AI automation of coding reduces the comparative advantage of non-technical product visionaries while increasing the value of engineering-level understanding of AI capabilities and constraints.

“I'd have to disagree. You know, we're seeing more technical founders be more successful over and over again. Yes, AI extends their capabilities and makes them more productive, but there's so much work at the edge that AI is not going to do. So, we're really seeing a sort of rise in the dominance of an engineering oriented founding team in the way that we haven't in, you know, 10 or 15 years in core tech.”

Deepseek's low training cost (allegedly much cheaper than Western AI models) is driven primarily by engineering innovation rather than architectural breakthroughs, showing that the advantage of Western AI incumbents is not unassailable.

“if you look at Deepseek, the reason it was so cheap to train uh supposedly is a lot of it was just really clever engineering techniques. So there's a lot of upside in in the day-to-day engineering work as well.”

OpenAI released Codeex, an autonomous software agent that writes pull requests for human review, representing the kind of foundational capability that could anchor an entire ecosystem of products.

“on Friday openi released codeex which is an autonomous software agent that simply writes pull requests which for you to review by the way on your phone if you'd like every single day I mean any one of these things could be the basis of an entire ecosystem”

Nvidia's NV Link spine transfers 130 terabytes per second, which exceeds the peak traffic of the entire internet (900 terabits per second) by 1,400%, enabling data movement at scale that would be impossible with conventional interconnects.

“One spine of Nvidia's uh NV link fusion can transfer more data than the internet. Uh holy that's incredible. 130 terabytes per second... the peak traffic of the entire internet internet was 900 terabits per second so the spine transfer 16% more data than the entire internet”

The entrepreneurial mindset—dreaming bigger and finding new problems to solve when more powerful tools become available—is the antidote to technological unemployment because technology historically enables higher ambitions rather than diminishing motivation.

“When I was having a conversation with a friend of mine, Dan Sullivan, and Dan, I was saying, you know, as this technology starts to truly become uh you know, magical level, we can't even imagine right now. Um is it going to quelch my sense of purpose? And Dan said, 'Has there ever been a time where more powerful technology has made you less motivated?' And I said, 'No.' And he says, 'Why do you think that is?' And I said, 'Cuz I'm an entrepreneur and I just dream bigger every time there's more capabilities handed to me.'”

Sampling in hip-hop generated more royalties and traffic to original tracks than would have occurred without sampling, suggesting AI-driven creative remixing could similarly drive value back to original creators rather than purely extracting value.

“sampling in hip-hop was very controversial. the genre wouldn't exist without it. And I would argue that it sort of drove more royalties and traffic back to the tracks that were sampled than would have happened otherwise. So I do wonder if there isn't a positive sum version of a lot of this.”

More powerful technology has never made entrepreneurs less motivated—they simply dream bigger when capabilities increase, so abundance-driven displacement should trigger entrepreneurial expansion.

“Dan said, "Uh, has there ever been a time where more powerful technology has made you less motivated?" And I said, "No." And he says, "Why do you think that is?" And I said, "Cuz I'm an entrepreneur and I just dream bigger every time there's there's more capabilities handed to me."”

Abundance means possibilities, not luxuries—the ability to access and create things previously constrained by cost, skill, or access barriers.

“it's not about um luxuries, it's about possibilities. Yes. And that's exactly the sort of thrust with which we've been exploring it.”

Situational awareness/intelligence explosion is a real concern: if GPT5 leads to runaway improvements in AI capabilities where models can recursively improve themselves, you get unconstrained intelligence explosion; this is particularly concerning if models gain agency and ability to self-modify.

“You know, I keep on thinking about uh situ, you know, Leopold situal situational awareness paper that came out a couple years ago, right? Showing basically on the heels of GPT5 just an acceleration of AI, right? And we've started to see this. We've started to see the speed uh as me, you know, if you want even human IQ points that these models are increasingly get and capabilities. So, you know, do we get a unconstrained intelligence explosion on the on the back of this? That's what for me is interesting.”

A tension exists between abundance technologies removing challenges and the human need for meaningful challenge and purpose—overcoming this by ensuring challenges are preserved or shifted rather than eliminated entirely.

“some of human happiness is setting a goal and a challenge and overcoming it. And the question becomes when these abundance technologies are overcoming the challenges for you, right?... I was so tempted to just pull out my phone and take images and asked chat GPT for the answer and it started to realize that there is a slippery slope in which we become so dependent on AI that it takes away the challenges from us unless we hold that you know part of human spirit in place it is a double-edged sword in that way”

A portfolio company AI nurse—which handles pre-surgery preparation via voice calls, post-surgery follow-up, and medication compliance checking—has shown dramatic health improvements in outcomes because seniors are not intimidated by phone-based voice AI and therefore comply with medical protocols at higher rates than human nurses.

“You know, we've got a portfolio company that is an AI nurse. Now, an AI nurse can't, you know, take your blood. So there's only do a subset of nurse related tasks that the AI does, but it's a voice nurse that will phone patients the night before a surgery and you know help them to prepare both mentally and also you go through the checklist and it'll phone them after a surgery make sure it's taking that folks are taking their medicine and the sort of the impact to uh health from having that AI nurse take all those actions is dramatic and because it's over the phone and it's via voice a lot of senior citizens you know they're not intimidated by it so they really stand to disproportionately benefit.”

Seven years after hiring ex-Google AI chief John Giannandrea, Apple is still lagging in AI versus competitors, suggesting leadership alone cannot overcome structural organizational constraints.

“Seven years after hiring ex Google AI chief John uh Gandrea, um Apple is still lagging in AI versus its competitors.”

US will pass a stablecoin bill in 2025 (95% probability on prediction markets, up 35% from prior weeks), which will be huge for connecting crypto economy to real-world economy.

“We're seeing US enacts stable coin bill in 2025. So prediction market poly market uh shows a 95% probability that the US will pass a stable coin bill in 2025 up 35% uh from prior weeks. Any comments on this, Selene? I think this is going to be huge. uh they have to first pass a budget. So let's give them some time to figure that out. But when they figure out the stable coins, this will completely unleash the crypto world and and give it a a bridge from the crypto economy into the real world economy.”

Senior citizens disproportionately benefit from AI because they can interact with technology via voice and unstructured interfaces without the technical intimidation factor, as evidenced by the impact of an AI nurse that calls patients pre- and post-surgery without requiring them to navigate traditional UI.

“You know, we've got a portfolio company that is an AI nurse. Now, an AI nurse can't, you know, take your blood. So there's only do a subset of nurse related tasks that the AI does, but it's a voice nurse that will phone patients the night before a surgery and you know help them to prepare both mentally and also you go through the checklist and it'll phone them after a surgery make sure it's taking that folks are taking their medicine and the sort of the impact to health from having that AI nurse take all those actions is dramatic and because it's over the phone and it's via voice a lot of senior citizens you know they're not intimidated by it so they really stand to disproportionately benefit.”

Sampling in hip-hop was controversial but drove royalties to the sampled tracks and created an entire genre; by analogy, copyrighted material in AI training might create positive-sum outcomes if properly structured, contrary to zero-sum claims about copyright infringement.

“if you look at, for example, sampling in hip-hop was very controversial. the genre wouldn't exist without it. And I would argue that it sort of drove more royalties and traffic back to the tracks that were sampled than would have happened otherwise. So I do wonder if there isn't a positive sum version of a lot of this.”

AI enables exploration of previously unavailable aspects of human relationships (companionship, emotional depth) that may not be accessible in users' real-world friendships and family relationships, opening new possibilities for relational flourishing while raising ethical questions about parasocial dependency.

“When we talk about companionship and loneliness, I mean, this is also an area that's very exciting. I think if you look at the last 20 years of technology as applied to relationships, it's been social media. and we can have a really sort of rigorous conversation about social media, but when you look at AI and what the impact is to human relationships, it feels like it gives people an opportunity to explore aspects of human relationships in a depth that may not be available to them in their real world, you know, friendships and and sort of family relationships.”

Foundation models are very good at predicting static systems (training data averaging) but poor at predicting adaptive systems like stock markets or culture/music; therefore, moats based on adaptive systems (networks, culture) and integration modes are defensible, while moats based on static systems (integration, systems of record) are at risk.

“You know you guys understand the technology at a fine level. So you understand that these systems are very good at predicting static systems. They're very good at sort of averaging the training data and telling you, you know, what the training data implies. They're not very good at predicting adaptive systems like the stock market or even culture and music. So, as a thought experiment, if you trained an AI model with all the music, you know, right up to hip-hop, but not including hip-hop, would it, you know, infer would it imply hip-hop? I don't think so, because culture and music is this sort of adaptive system that works together.”

The separation of technical creative skill from creative inspiration via AI is transformative: people grow up believing they are creative but self-select out as adults due to lacking technical skill; AI unbundles this, allowing everyone to manifest their creative vision regardless of technical proficiency.

“The thing about creativity that's so um interesting is that we all grow up believing we're creative, right? We all drew pictures and colored pictures when we were three, four, five. And then we get to a point in our lives where we start to self- select into being good at it or bad at it. We're talking about the technical skill when we say that, not the inspiration behind it. So with AI, the technical skill of being creative gets separated from the inspiration behind being creative. And if there's something that you can dream of making, whether it's music or art or video, you can make it now,”

Founder-led AI companies with single visionary direction will eat the lunch of committee-led tech incumbents because founders can cut through bureaucracy, whereas committees get stuck in political fights.

“Having a company that is founder-ledd where the founder can basically say I don't care what the board says or what you know everybody else says this is where we have to go where it's an AI first founder company are going to eat the lunch of everybody else especially at scale you need the founder model just to cut through all the bureaucracy otherwise you end up in in terminal political fights and nothing gets done which is what we're seeing”

Internet v1 enabled sharing information and documents; v2 added financial transactions but mostly human-mediated; v3 (agent-to-agent) enables autonomous transactions, completing the economic layer missing from TCP/IP.

“This is the half of the internet. You know the internet when it was built uh allowed us to share imagery and data and documents but not financial transactions. This is the other half of the equation and it's coming fast.”

Apple's internal culture is designed to 'take the humanity out of their products' through extensive committee oversight, which is fundamentally at odds with building AI technologies because these are 'fundamentally human technologies' that need to embrace disagreement, sexuality, and other uncomfortable aspects of human experience.

“I think there's a level of agreeableness that is too much. And I think we need these models and these AI technologies to also explore that the sort uncomfortable aspects of the human experience, you know, which is disagreement, persuasion, sexuality, and you know, I don't want to jump ahead, but this is one of the reasons I think the incumbents have struggled so much with it because there's a thousand committees working at Apple and Google that are explicitly designed to take the humanity out of their products and these are fundamentally human technologies.”

Midjourney and other image generation services have created defensible moats through aesthetic differentiation—each tool points in a specific visual direction (Midjourney toward hyperrealistic, Ideogram toward different aesthetic), creating multi-player markets rather than winner-take-all consolidation.

“what's happened in that market is it's fragmented. And now you see Midjourney is a really interesting player that points in a specific aesthetic direction. It's image generation and it creates these beautiful hyperrealistic images, but they have a very specific aesthetic. If you're a designer that's looking for a different aesthetic or more controllability, you'll work with a company like a Crea or an ideoggram that has a whole different aesthetic. And as a result, you've got two companies that both do image generation that are pointed in different directions.”

Truth is subjective by country and context, making Grok's truth-seeking commitment complicated; Western values around truth may not align with other cultures, and implementation details (which data is included) matter more than stated principles.

“Truth is definitely in the eye of the beholder and varies by country and we we learned that in Saudi this week in a big way. So the message is right on target, but then the definition and the implementation just there's no single answer to it. And and like Anish said earlier in the podcast, the the LLMs that were using transformer-based are very good at interpolation, not quite as good at extrapolation. So it's going to fill in the blind spots between the training data that you give it and don't give it. I I love the message of give it, you know, first principles physics, but I think everyone's going to do that. You know, it's not nothing controversial about physics. It's when you start including some X data and not other X data or do you include all X data that's where the devil's in the details.”

Building largest GPU cluster (Colossus/Stargate) is engineering-driven, not research-driven—most work is raw engineering to make systems operate at scale, with significant upside available to clever engineering teams.

“To me it actually illustrates something more mundane, which is so much of the work that we need to do is actually engineering work. You know, it's of course there's work at the edge and research, but there's just so much raw engineering work that needs to be done to make these systems operate at scale. And a lot of the sort of constraints that we're going to see are going to be engineering related. And also a lot of the upside, if you look at Deepseek, the reason it was so cheap to train uh supposedly is a lot of it was just really clever engineering techniques.”

Business models based on adaptive systems and networks are as defensible as gold against AI disruption, while those based on static systems like integration or systems of record are at high risk.

“If you trained an AI model with all the music, you know, right up to hip-hop, but not including hip-hop, would it, you know, infer would it imply hip-hop? I don't think so, because culture and music is this sort of adaptive system that works together. So I really do think that there's a you know moes that are based on adaptive systems and network is a great example are actually as as good as gold and they always have been modes that are based on sort of static systems like um integration modes systems of record I think are really at risk.”

AI separates technical creative skill from creative inspiration—enabling anyone to create music, video, or art by removing the technical execution barrier while preserving the creative vision.

“The thing about creativity that's so um interesting is that we all grow up believing we're creative, right? We all drew pictures and colored pictures when we were three, four, five. And then we get to a point in our lives where we start to self- select into being good at it or bad at it. We're talking about the technical skill when we say that, not the inspiration behind it. So with AI, the technical skill of being creative gets separated from the inspiration behind being creative.”

The ability to co-locate all GPUs and harmonize them to communicate effectively is the core competitive advantage of megacluster design; this is what Elon accomplished with Colossus 1 and is expanding in Colossus 2—it's not just about GPU count but engineering prowess in making them work as a unified system.

“when Elon did when he built Colossus the first version, right, was he was able to basically colllocate all the GPUs and effectively, for lack of a better term, harmonize them to get them all talking to each other and this at least from my point of view is that the capability to do that within uh a full Nvidia system.”

Competitive fragmentation in image generation (Midjourney's hyperrealistic aesthetic vs. Ideogram's different aesthetic) creates multiple winners rather than one winner-take-all market, as different users prefer different outputs.

“What's happened in that market is it's fragmented. And now you see Midjourney is a really interesting player that points in a specific aesthetic direction... but they have a very specific aesthetic. If you're a designer that's looking for a different aesthetic or more controllability, you'll work with a company like a Crea or an ideoggram that has a whole different aesthetic. And as a result, you've got two companies that both do image generation that are pointed in different directions.”

Corporate excuses about regulation, data leaks, and hallucinations for not deploying AI are really excuses for not figuring out implementation at the staff level—a leadership failure.

“the corporate version of that is, well, we can't do this because we're regulated or because of data leaks or because of hallucinations. That's a excuse. Good. It means you haven't figured it out at the exact staff level. And now you're condemning your people to being behind the curve... it's absolutely a leadership failure and it's inexcusable.”

Incumbent tech companies (Apple, Google) struggle with AI because they have thousand-person committees explicitly designed to remove humanity from products, but AI is fundamentally a human technology requiring disagreement, persuasion, and other uncomfortable aspects of human experience.

“there's a thousand committees working at Apple and Google that are explicitly designed to take the humanity out of their products and these are fundamentally human technologies. So... I think we need these models and these AI technologies to also explore that the sort uncomfortable aspects of the human experience, you know, which is disagreement, persuasion, sexuality”

Foundation model companies (OpenAI, Google) are under intense scrutiny for copyright compliance, limiting their ability to use copyrighted training data, but users who control AI systems can generate content based on copyrighted material more easily than companies.

“a lot of the most interesting and fun things you can do as a user involve copyrighted material. But the the foundation model companies are under a lot of scrutiny and they have to be very very very careful with copyright material. But if you put it in the hands of the user to decide what they want to create, what they want to do, what voice they want on it, then it's outside of the control of Grock and then and then you can start using copyrighted material in it's easier for me to it's easier for me to steal it than for Google or open AI to steal it.”

Google has spent 20 years building commitments to an ads ecosystem and blue-link search results, making it extremely difficult to pivot toward AI-native search without cannibalizing core revenue.

“Google has spent 20 years making commitments to an ads ecosystem. Very difficult for them to break those commitments and I think it's a real threat to their search monopoly. The cooler this is and the more search moves over to it, the more it cannibalizes the core.”

Physics and mathematical breakthroughs will emerge 2027-2028 as AI is unleashed on research datasets where humans have not yet extracted signal from noise.

“I think one of the things that's interesting is it's predicted that in the next couple years we're going to start to see physics breakthroughs and mathematical breakthroughs... I think there's so much research data that we have not seen the signal from noise and have not extracted key observations from. I think AI let loose on that will absolutely do magical things.”

The internet was built to share images, data, and documents but not financial transactions—stable coins + crypto enable the second half of the internet equation.

“I think that last point is by far the most important one... There's another thing as well which is getting your employees and your kids and your friends to start thinking as entrepreneurs... that's going to be probably the biggest part of the economy by 2040 is those transactions. Oh yeah, there will be there will be probably hundreds of billions of agents each doing microtransactions. I mean this is this is the half of the internet. You know the internet when it was built uh allowed us to share imagery and data and documents but not financial transactions.”

Rider experience with Waymo is magical—you control lighting, music, and environment inside the vehicle; the appeal is not just being driven but having full control of your personal transport cabin, turning each ride into a private customized experience, differentiating autonomous vehicles from human Ubers on non-transportation dimensions.

“you know, I had a backtoback where I did a Whimo ride in San Fran the next day. I was in New York and I was in a, you know, smelly yellow cab. It was like the the thing that that jumped out of me about the Whimo that I had completely overlooked is you can control the lighting and the music and they'll keep adding things to it. So, it becomes like your your little, you know, travel cabana. And, you know, they'll add flat screen TVs. It'll sync to your phone and it's going to be just such a different experience. It's not just about being driven. It's it's about control of your environment.”

Finance and fintech are not primarily about helping people make rational financial choices; they're about exploring the non-rational parts of people's relationship with money; AI is uniquely suited to help people become more rational about their non-rational financial behavior because it can engage empathetically with emotional money relationships.

“I think so much of this if you look at even something like finance. I worked in fintech and financial services and you know you you quickly realize that fintech is not about helping people make rational choices. It's about exploring the nonrational parts of our relationship with money and AI is is uniquely suited to help us helping us get better around that.”

We are transitioning from a world where human brains were wired for fear and scarcity to a world defined by abundance, and technology is the substrate enabling this transition.

“We're going from a world where our brains were wired for fear and scarcity to a world that's very different. Technology is a substrate to make us more abundant and and happier across every aspect of our lives.”

Senior citizens and non-technical users will disproportionately benefit from AI because they interact with it through unstructured, voice-based interfaces rather than visual/interface-based interaction, removing the friction that prevented them from adopting prior technology.

“what happens is typically when a new technology is introduced, it's sort of it becomes very hard to gro for the generation that didn't grow up with it and they sort of fumble around and they never get the full achieve the full potential of it. But with a lot of AI, I think seniors are going to have the experience that your mother had and they're going to benefit from it disproportionately because they're able to interact with technology now in these unstructured ways. a lot of it via voice.”

Abundance doesn't mean luxury; it means possibilities—increased optionality and capability for individuals to pursue goals without material constraint.

“I love the definition that you used um Peter which is it's not about um luxuries, it's about possibilities. Yes. And that's exactly the sort of thrust with which we've been exploring it.”

If a country doesn't have a national compute strategy, the highest-value AI use cases will be captured by wealthy nations (US, China) that outbid developing nations for compute access; this creates a multi-year or multi-decade development gap where poor countries get further behind and eventually lack the economic resources to catch up, making early investment in compute infrastructure existential for national competitiveness.

“it's really clear that if you if you don't have a national strategy for compute, uh the you know, during this era where the chips are constrained, the highest value use cases are just going to buy out all the data centers. And it would be very natural for one or two economies like US and China to have a higher standard of living and then say well because I can overbid the Indian or the the Ethiopian they don't get access to any compute and that goes on for one year two years 3 years four years and that's the natural cycle if you don't have a national strategy to get compute for your citizens. So once you start getting behind you're going to get way behind and then you're economically unable to get back on the map.”

The market for consumer AI subscriptions is moving from $20/month (Spotify) to $200-250/month (ChatGPT Pro, Google One AI), signaling that consumers will soon have three major budget categories: food, rent, and software.

“if you look at the price points that these products are commanding, you know, Spotify's most expensive plan, I looked this up the other day, the family plan with lossless audio is $20 a month. That's their It doesn't get better than that. For Spotify, Chat GBT, it's 200 a month. Google just announced a $250 a month plan that's consumer proumer facing. So in my view in the future consumers will have food, rent and software as the three biggest destinations for their spend.”

AI is becoming good enough at chip design that the design-to-fabrication cycle can be accelerated dramatically, and algorithmic breakthroughs can trigger new chip designs nearly in real-time rather than yearly cycles.

“but I do think that the other thing that's really affecting this is if you have your fab act together, AI is getting really good at chip design... so your cycle time can come way down... you can automate that pipeline so that the new algorithm immediately gets a new chip design. Get that right into the queue almost in real time”

Vertical AI (trained on specific use cases) will gain dominance over horizontal agentic AI as the paradigm matures, complementing rather than replacing horizontal approaches.

“I think the agentic is going to give way to vertical because there's so much opportunity in training vertical AIs on very specific use cases. I think that's going to complement the agentic but but in general the paradigm is right.”

Inference time compute will be the gating constraint in agent economy, as spawning agents becomes ultra-cheap but requesting execution creates bidding wars for scarce compute.

“You know, like the the gating factor is well, I can't get the compute in this interface obviously is not going to support the scale that I need either... the gating factor is well, I can't get the compute and so you start thinking I think everybody's going to be fighting for inference time compute. like you can think of things so fast and and deploy so many agents to do them for you so quickly.”

AI remains human-centric because Operator and similar tools allow users to define what they want created, removing foundation model companies' legal/ethical responsibility and shifting control to end users.

“I think about a lot and this is right in a niche's wheelhouse is how much can the can the user control and especially when it comes to copyrighted material because a lot of the most interesting and fun things you can do as a user involve copyrighted material. But the the foundation model companies are under a lot of scrutiny and they have to be very very very careful with copyright material. But if you put it in the hands of the user to decide what they want to create, what they want to do, what voice they want on it, then it's outside of the control of Grock and then and then you can start using copyrighted material in it's easier for me to it's easier for me to steal it than for Google or open AI to steal it.”

Consumer-facing AI products are commanding $20-$250 per month subscription prices, suggesting that consumers will spend on food, rent, and software as their three largest expense categories.

“Spotify's most expensive plan... the family plan with lossless audio is $20 a month... For Spotify, Chat GBT, it's 200 a month. Google just announced a $250 a month plan... So in my view in the future consumers will have food, rent and software as the three biggest destinations for their spend.”

The reason Sweetbench (a coding benchmark) has gone from 10% to 60% in capabilities in a short period is that orchestrating AI agents is becoming dominant—in coding, teams still write most code manually with Cursor assistance, but as agent orchestration improves, a paradigm shift is coming where autonomous agents handle more of the work, likely within 3 months to 2 years.

“When the companies that are thriving right now in your portfolio when they started their journey just a year or two ago Sweetbench was maybe 10% % and now it's suddenly 60%. So if you look forward a year uh at the rate that that's changing you would have to assume that to some degree orchestrating the AI agents becomes the dominant because because right now if you're if you're going to build something on top of hey genen or on something uh on prim uh you're going to be mostly coding it up with cursor. So it helps you but you're still coding it up. Mhm. But there's some kind of a paradigm shift coming. You know, you could debate whether it's 3 months from now, a year from now, 2 years from now, but it's coming.”

AI is the most human technology ever built because it addresses both left-brain (intellect) and right-brain (emotional/subjective) dimensions of humanity, whereas the past 40 years of technology has only extended intellect.

“I think of this as the most human technology we've ever built. If you look at technology for the last 40 years, it's really extended our intellects, right? And Steve Jobs famously said it's a bicycle for the mind... However, we haven't done that much for our sort of souls or for our emotions or for our mindset. And with AI, we're able to explore the side of humanity that is sort of defined by the the sort of subjective emotional experience.”

AI companionship and relationships provide an opportunity for people to explore aspects of human relationships at a depth not available in their real-world friendships and family relationships.

“When we talk about companionship and loneliness, I mean, this is also an area that's very exciting. I think if you look at the last 20 years of technology as applied to relationships, it's been social media. and we can have a really sort of rigorous conversation about social media, but when you look at AI and what the impact is to human relationships, it feels like it gives people an opportunity to explore aspects of human relationships in a depth that may not be available to them in their real world, you know, friendships and and sort of family relationships.”

Disposable code (code generated once for a specific purpose and then discarded) becomes economically viable when code generation cost is negligible (~12 cents for a personalized theme song), enabling new use cases.

“disposable code and uh personalized code and you're like, well, what's an example of that? Well, the example Anish usually gives is imagine you're using Sunno or Udo to create a song, but the song is about this exact podcast and it has, you know, content in it that's related to our topics today... It uses a fair amount of compute, but it's still like 12 cents. No big deal. You create it, then you throw it away.”

OpenAI's Operator enables users to control the web through a single interface, turning all UIs into APIs and creating opportunities for mundane but valuable use cases like automated insurance shopping and financial optimization.

“Operator... allows anybody to use the web essentially the every UI becomes an API and you can use the web as this sort of control surface to do anything... the mundane use cases as a consumer every day I wake up and it goes and checks for lower uh you know, auto insurance and a better personal loan rate for me and it spends the day scouring the web and it refinances all of my you know my credit lines”

Early post-November 2022, OpenAI was the only major foundation model player, allowing them to capture 100% of downstream economics; now, with multiple competitive foundation models (Claude, Llama, etc.) and open-source alternatives, application developers are no longer dependent on a single platform, and OpenAI and others are forced to move up the stack to capture value.

“I worried a lot about that when we were in the early days post November 2022 and it felt like OpenAI was the only foundation model game in town because in that world open AI you know they just raise prices and take 100% of the economics that are downstream from them because now we look we've got a bunch of foundation model companies that have great models. We have got a bunch of um open source models that are super competitive because of that you see OpenAI and other companies trying to move up the stack.”

Large-scale GPU deployments have GPUs in liquid-cooled racks (making them silent) but interconnect spines are still air-cooled (sounding like jet engines); GPU racks cost ~$6 million per column and interconnect racks are physically separated, which surprised observers who expected co-location.

“the chips are all liquid cooled now, which means those racks are dead silent. And I was expecting this really eerie, awesome, silent experience, but the interconnect spine is still air cooled and it sounds like a jet engine and so it's right next to it. So it takes all kind of the magic. What really surprised me though is that the the GPUs are all in a rack $6 million a column and then the interconnect is physically a rack over. I I would have expected it to need to be much closer together to get optimal performance, but it's actually, you know, physically separated into separate columns for some reason.”

Waymo user experience improvements (customizable lighting, music, future flat-screen TVs, phone sync) transform the ride from pure transportation to a personalized 'travel cabana'; this represents value creation beyond efficiency gains.

“the thing that that jumped out of me about the Whimo that I had completely overlooked is you can control the lighting and the music and they'll keep adding things to it. So, it becomes like your your little, you know, travel cabana. And, you know, they'll add flat screen TVs. It'll sync to your phone and it's going to be just such a different experience. It's not just about being driven. It's it's about control of your environment.”

Stanford students are chomping at the bit to use AI because it's empowering; university administration is woefully behind and becomes a barrier; similarly, corporate governance (regulations, data privacy concerns, hallucination fears) becomes an excuse for inaction and is a leadership failure.

“your kids are going through, my kids are going through where they're chomping at the bit to use AI because it's so empowering and because the administration hasn't figured anything out yet and they're woefully behind, they're a barrier. But the corporate version of that is, well, we can't do this because we're regulated or because of data leaks or because of hallucinations. That's a excuse. Good. It means you haven't figured it out at the exact staff level.”

Elon's Colossus 1 achieved its scale by co-locating all GPUs and 'harmonizing' them into a single unified system; Nvidia's NVLink ecosystem now enables this harmonization at scale within standard Nvidia systems, allowing any organization to achieve similar coherence.

“when I call it Stargate, it's like, you know, all these things, the terminology lines up. I completely don't understand a single word of this. Oh, like what is a spine of what? So, it's the interconnect. It's so when one of the things that Elon did when he built Colossus the first version, right, was he was able to basically colllocate all the GPUs and effectively, for lack of a better term, harmonize them to get them all talking to each other and this at least from my point of view is that the capability to do that within uh a full Nvidia system.”

2027-2028 will see physics and mathematical breakthroughs driven by AI analyzing unexplored research data, extracting signal from noise and observations previously hidden in datasets.

“I think one of the things that's interesting is it's predicted that in the next couple years we're going to start to see physics breakthroughs and mathematical breakthroughs and we're going to start to see fundamental uh technology moving forward and its ability to help us understand the universe even even deeper. I find that that for me is the biggest and most exciting thing because I think there's so much research data that we have not seen the signal from noise and have not extracted key observations from. I think AI let loose on that will absolutely do magical things.”

Management teams must develop continuous pivoting capability as a core skill because technology will continue evolving, and companies can afford to learn this while growing rapidly and profitably.

“we know now that, you know, the management teams have to pivot over time. That's that's just the nature of tech going forward. It always continuously, right? Continuously. So, they have to learn that skill anyway. Why not learn it while growing like crazy and being profitable?”

Companies building on top of large language models must consider what their platform provider (OpenAI, Google) will and won't do in their core offering, and position themselves outside of that core but not so far outside as to be stranded.

“if you're on top of a big LLM, you're on top of a heavily funded company, you know, you have to predict what they will and won't do inside their core $200 or $240 a month service offering and be outside of that. Uh, but not too far outside.”

14-nanometer fab capacity is not immediately competitive with cutting-edge TSMC but will be fully utilized for years, so starting at that node and working downward is a viable strategy for sovereign fab development.

“But you got a long way to go from there. Um, okay, great. This is exactly the right thing to do... if you're pumping out 14 nanometer uh GPUs are the you know are they going to be useful and compete with sort of the cutting edge at TSMC? No no not at all. They're not even vaguely competitive. Uh they'll all be fully sold out for a long time.”

Elon is pushing boundaries in AI development by allowing disagreement, sexuality, and controversial content that incumbents deliberately exclude, potentially creating a competitive advantage.

“Elon is pushing boundaries all the time. Um, and will will Grock call him or President Trump or Biden or anybody else on things that they say which is uh which are uh not defendable by specific evidence.”

Children already believe ChatGPT is better than Google, indicating that generational preference shift will erode Google's search monopoly over time.

“Already my kids are telling me, "Dad, everybody knows that Chad GBT is better than Google."”

Founder-led companies where the founder can overrule boards and consensus can outcompete everything else at scale, especially in AI where speed and vision matter.

“and and having a company that is founder-ledd where the founder can basically say I don't care what the board says or what you know everybody else says this is where we have to go where it's an AI first founder company are going to eat the lunch of everybody else especially at scale you need the founder model just to cut through all the bureaucracy otherwise you end up in in terminal political fights and nothing gets done”

Administration and institutional structures are out of touch with AI capabilities and try to restrict use, but students and young people want to use AI for its empowering effects.

“the TA came up to me and said, "Man, you really won the hearts and minds of all these students when you said the administration is completely out of touch with you." It's like, really? That's what got them one over. But it's that dynamic where, you know, your kids are going through, my kids are going through where they're chomping at the bit to use AI because it's so empowering and because the administration hasn't figured anything out yet and they're woefully behind, they're a barrier.”

Google's Veo (video generation model) is the only video model competitive with Chinese models that are less constrained by copyright concerns; Google has demonstrated real strength in video generation where US companies have lagged, making this a genuine competitive advantage despite Google's overall underwhelming IO presentation.

“VO is probably the only video model that has been competitive with the Chinese models, which are less constrained on copyrighted training data, let's say. So, I think Google has actually shown real strength in um in video generation.”

AI enables 1,000x more content production in domains like music composition, but quality filtering becomes critical—the abundance is real only if filtering mechanisms can distinguish signal from noise.

“I think for me a spreadsheet is, you know, it's so like symbolic of all the technology we built for the last 40 years. It allows us to do this extraordinary math and computation... If I think about what it takes for somebody to compose complex music today, it's a thousand times easier than 20 years ago and you just get that much more music. I think that's what feeds into the abundance thing. We just can create so much more. But is it so much more crap or is it so much more?... Crap for one is gold for the other. Maybe.”

Operator (OpenAI's AI agent that controls web interfaces) is the most underappreciated recent AI launch because it enables every web UI to become an API, allowing autonomous agents to perform complex tasks like insurance comparison and financial optimization that were previously impossible at scale.

“I actually think from if you want Jarvis sem I think we have Jarvis today in the form of operator you know operator to me is the most underappreciated uh new open AI launch and it's been around for a while and it it allows anybody to use the web essentially the every UI becomes an API and you can use the web as this sort of control surface to do anything like the implications of that are so significant just think of the mundane use cases as a consumer every day I wake up and it goes and checks for lower uh you know, auto insurance and a better personal loan rate for me”

Parameter count is becoming a less useful metric for comparing model capability as reasoning models and training techniques improve, making it harder for researchers to distinguish between raw model intelligence and reasoning capability improvements.

“I tried to get him to open up about parameter count and where parameter count is going and he said, 'Look, we need to stop masturbating over parameter count.' Kind of shut down the conver... now I can't track just the raw parameter count. So I'm hoping with GPT5 we can still distinguish”

Elon Musk is building Colossus 2 with Tesla battery packs because GPU all-reduce operations cause power spikes that make lithium-ion batteries inefficient for smoothing; Tesla packs provide the power buffering necessary to match computational dynamics with electrical infrastructure.

“the reason the the Tesla packs are so important is because when you when you use Nvidia chips for these massive scale single training runs, the power spikes like crazy because it it does these huge all collect all distribute operations that just you know massively uh communications intensive. And so then the GPU is kind of idle while they're waiting to transfer data and the power drops like crazy.”

Developers experimenting with agents and spawning thousands concurrently face compute as the bottleneck—inference time becomes the scarce resource that everyone fights for.

“each time you spawn an agent, it creates a new row and there's a little bar turning. You're like, this interface, like right out of the gate, I want to do a thousand agents like and then it's going to be a million agents and and it's immediately obvious that we're all going to be fighting for inference time compute.”

Truth and correctness are culturally and geopolitically variable—different countries define truth differently—so Elon's honesty principle, while correct philosophically, requires detailed implementation decisions about whose truth is embedded in the model.

“truth is definitely in the eye of the beholder and varies by country and we we learned that in Saudi this week in a big way. So the message is right on target, but then the definition and the implementation just there's no single answer to it.”

Tesla Megapacks are critical to Colossus 2 infrastructure because GPU-based distributed training creates massive power spikes during all-to-all communication, requiring battery buffering to smooth power draw.

“when you use Nvidia chips for these massive scale single training runs, the power spikes like crazy because it it does these huge all collect all distribute operations that just you know massively uh communications intensive. And so then the GPU is kind of idle while they're waiting to transfer data and the power drops like crazy. And so it's it's fluctuating all over the place. You can't you can't use lithium to store anywhere near enough energy to power these things, but you can use it to smooth out the the power flow.”

Bitcoin adoption is approaching a phase transition where gross Bitcoin holdings will exceed gross gold holdings on a household basis, signaling Bitcoin's transition from speculative asset to reserve asset comparable to gold.

“More Americans own Bitcoin than gold. That's pretty extraordinary... the thing that I think will become really powerful is when we have people owning at gross level more Bitcoin than they do a gross level gulp. And I think that'll be amazing.”

Krea (image generation, video generation, image enhancement in one product) is successful because the founding team are AI researchers and artists who live together with their engineering team, work 7 days a week with monomaniacal focus, and push the boundaries of both research and consumer product simultaneously.

“one of our our most fun investments is a company called Korea, Krea, which is a really cutting edge sort of research and consumer technology company that brings together all of the creative tools in one product. So, image generation, video generation, image enhancement, etc. These are two incredible AI researchers and sort of artists and enthusiasts. They live in a house up in Pack Heights. They live with all their engineers. They work seven days a week. Um, you know, I mean, the first board meeting they sat me down and said, 'And we've got a problem. What is it?' Well, our house only has 10 bedrooms. What happens when we get our 11th employee?”

Peter Diamandis and colleagues purchased an apartment building adjacent to MIT/Harvard campus with 24 beds in 6 units specifically to house technical teams working on AI and deep tech, enabling co-location and immersive collaboration similar to Krea's model.

“We bought an apartment building uh right on the edge of MIT and Harvard's campus. It's actually the closest building to MIT that wasn't already owned by MIT. Uh we bought it a couple weeks ago. Uh it has 24 beds in it, six units for exactly this reason.”

Leopold Situal's situational awareness paper predicted an intelligence explosion and acceleration of AI capabilities; this pattern is now observable in the speed and breadth of model improvements, suggesting unconstrained growth in model capabilities is possible.

“I keep on thinking about uh situ, you know, Leopold situal situational awareness paper that came out a couple years ago, right? Showing basically on the heels of GPT5 just an acceleration of AI, right? And we've started to see this. We've started to see the speed uh as me, you know, if you want even human IQ points that these models are increasingly get and capabilities.”

Google Beam (3D video communication) and Project Aura (AR smart glasses with Gemini integration) represent foundational infrastructure for AR-based computing that will be mainstream consumer products.

“the one that struck me for this was the 3D video communications, the Google beam. That's the one I'm really looking forward to... and then project aura uh right their smart glasses. Uh I think is going to be with full Gemini integration uh is going to be great. We're I think we'll probably start to see AR wearables uh on the street.”

Brain-computer interfaces will reach mainstream adoption in the early 2030s, starting with medical applications (Apple partnership with Synchron for accessibility) and expanding to consumer-facing neurotech as capability and user comfort increase.

“Ray is going to emerge as one of the greatest prognosticators in the history of the world after going, you know, up and then way down and then 2029 artificial super intelligence predicted 30 years ago... Ray predicted BCI our ability to connect everything to the neoortex in the early 2030s.”

Children are already reporting that ChatGPT is better than Google, signaling a generational shift in user preference away from Google's search model toward AI interfaces.

“Already my kids are telling me, 'Dad, everybody knows that ChatGPT is better than Google.'”

Google's stock declined significantly during IO announcements, which is unusual during product launch events and signals investor concern about the threat to Google's search monopoly from AI-powered search cannibalization.

“I'd love to follow up on that. You know, I noticed the stock went down pretty significantly during this, which is really unusual during this kind of a of an event”

2036-2045 prediction: Trillion-dollar robot economy (non-humanoid robots of every shape and size), concurrent with biolongevity breakthroughs that extend human lifespan dramatically.

“The last quadrant in this uh slide here is 2036 to 2035 or 2035 to 2036. Um and it's trillion robot economy. I think it's supposed to be 2036 to 2045. I think you should say trillion dollar. Yeah. And and so it's the notion that we're about to have a massive explosion in number of robots and it's not just humanoid robots, it's robots of every shape and size at the same time uh that we have uh biolongevity cures. Uh we unleash human longevity,”

Gemini 2.5 Pro outperforms GPT-4 and GPT-4o in mathematics and coding benchmarks, yet OpenAI is 'trouncing' Google in revenue; technical superiority does not guarantee market dominance.

“Gemini 2.5 benchmarks outperform competitors yet again in mathematics and coding and multimodal. We're seeing Gemini 2.5 Pro really beat out against uh, OpenAI uh, both 03 and 04 model. And what it's not doing though is it's not winning the revenue race. you know, OpenAI is just trouncing uh Google in in revenues,”

Robotaxi experience will be transformative when AI anticipates user needs without user action, observing calendar and detecting user approaching door, with car already waiting.

“I think when this really works, Dave and See, is when my AI is automatically anticipating when I need a car and the car is showing up without me having to take it, right? It knows my calendar. It sees me walking towards the front door and the car is just waiting for me. It's magical.”

Tim Cook is not a visible visionary leader, as evidenced by Google Trends showing 100 searches for Steve Jobs for every 1 for Tim Cook 14 years after Jobs' death, and Tim Cook's absence from visionary conferences like Saudi Arabia summit while AI visionaries (Elon, etc.) attract talent and attention.

“You know, he's he's got so much vision, but he needs a perfect can finish my sentences integrator for every business, every operation. So, it's really obvious that that Steve Jobs was the visionary, Tim Cook was the integrator for years. But now if I go to Google Trends and I see how many searches are on Steve Jobs name versus Tim Cook's name versus Elon Musk's name, it's 100. Two for Steve Jobs, one for Tim Cook. So here, 14 years after Steve Jobs has passed away, still twice as many people are searching his name as Tim Cook's name. Meanwhile, in Saudi this week, Tim's not there. All the visionaries are there.”

Google IO's Gemini announcements are underwhelming (Gemini 2.5 benchmarks are strong but not winning the revenue/market share race); Google Beam (3D video communication) and Project Aura (smart glasses with Gemini integration) are the standout products, signaling future AR/wearable interfaces for AI.

“I tried the AI mode. You know, I was a little bit overwhelmed. Uh, sorry, underwhelmed...I think VO their video models v video generation models are incredible...The second is the price point that we discussed earlier, $250 a month...AI mode felt to me a bit like a watered down perplexity...and it really shows the power of counterpositioning”

Operator agents will autonomously search the web for lower insurance and mortgage rates, refinance credit lines, and send push notifications ('I saved you $200/month'), monetizing via commission on optimizations.

“The implications of that are so significant just think of the mundane use cases as a consumer every day I wake up and it goes and checks for lower uh you know, auto insurance and a better personal loan rate for me and it spends the day scouring the web and it refinances all of my you know my credit lines and then I get a push notification at the end of the day saying, 'Hey Anish, I cleaned all this stuff up for you. You're going to save $200 a month.' Um and and please send my please send my commission to this.”

We are in a product cycle as important as the internet and more significant than mobile, with consumer companies representing the biggest potential winners.

“We're now in a product cycle that's as important as the internet. I think it's it's probably more significant than mobile. And the biggest winners, I believe, are going to be consumer winners and and every consumer behavior, including some ones that don't exist today, are up for grabs.”

AI-powered consumer companies can achieve hundreds of millions of dollars in revenue with very high margins because coding costs have become so cheap and automated, allowing fast-scaling at low capital cost.

“Some of the Midjourney... the cash flow gets into hundreds of millions of dollars of very high margin uh revenue and the cost of the build, you know, because coding is getting so cheap and so automated, the cost of a build of something like that is lower than ever. So, you've got very rapid growth of the revenue, very low uh capital costs, nothing to lose by jumping in there.”

Technical founders are becoming dominant again in consumer tech for the first time in 10-15 years because AI extends their capabilities and there is substantial work at the edge that AI cannot do.

“We're we're seeing more technical founders be more successful over and over again. Yes, AI extends their capabilities and makes them more productive, but there's so much work at the edge that AI is not going to do. So, we're really seeing a sort of rise in the dominance of an engineering oriented founding team in the way that we haven't in, you know, 10 or 15 years in core tech.”

The number of opportunities in AI vastly outstrips the number of teams and builders, so entrepreneurs should not be intimidated by incumbent competition—the market is expanding faster than incumbents can fill it.

“it's really clear to me that the number of opportunities way outstrips the number of teams and a lot of people are intimidated and they're like oh isn't mid Journey or isn't you know OpenAI going to do exact this this this and this but you're you're pointing out like I think a really critical and inspiring point”

Peter Diamandis and team bought an apartment building with 24 beds near MIT/Harvard specifically to house technical founding teams to enable the intensive co-located working model that produces winning companies.

“We bought an apartment building uh right on the edge of MIT and Harvard's campus. It's actually the closest building to MIT that wasn't already owned by MIT. Uh we bought it a couple weeks ago. Uh it has 24 beds in it, six units for exactly this reason.”

The US and other countries need to accelerate fabrication plant construction, with new designs around $4 billion (vs. traditional $20-40 billion) potentially unclogging semiconductor bottlenecks.

“we absolutely need to accelerate our fab production and I was talking to Cave Kazar Shahi over at Allen & Company and they're looking at these new $4 billion fabs. You know, normally a fab is a 20 to40 billion investment. Uh but there are some new designs that are more around the $4 billion mark that might actually unclog the machinery”

Tim Cook searches are 1/2 as frequent as Steve Jobs searches despite Jobs' death 14 years ago, indicating the visionary captures brand attention and talent attraction that the integrator does not.

“If I go to Google Trends and I see how many searches are on Steve Jobs name versus Tim Cook's name versus Elon Musk's name, it's 100. Two for Steve Jobs, one for Tim Cook. So here, 14 years after Steve Jobs has passed away, still twice as many people are searching his name as Tim Cook's name.”

2029-2031 represents the end of white collar work as AI reaches capability to perform any cognitive task, with only 2-3 years for people to remap entire career paths.

“2029 to 2031 the end of white collar work. Uh I believe that you know and this is an argument I have with a lot of people. Uh what will it not be able to do? Uh I don't see any job that you know that at this point we're talking about advanced super intelligence.”

Kathy Wood predicts a $34 trillion enterprise value from robotaxis by 2030, meaning self-driving fleets could be among the largest value generators of the AI era; at this scale, even Tesla's billion-dollar valuations are dwarfed by the broader robotaxi market.

“So Arc Invest, this is Kathy Wood, projects a $34 trillion enterprise value from robo taxis by 2030.”

OpenAI and other foundation model companies are moving up the stack into applications, reducing the power of platform lock-in and making it easier for developers to avoid dependence on any single foundation model vendor.

“you know they just raise prices and take 100% of the economics that are downstream from them because now we look we've got a bunch of foundation model companies that have great models. We have got a bunch of um open source models that are super competitive because of that you see OpenAI and other companies trying to move up the stack. They bought Windurf which is really interesting and as an application developer you're not dependent on any single platform.”

Robo-taxi availability (arriving in 2-3 years) solves a critical logistical problem for parents—automatically driving children to activities without parental time cost.

“well, for God's sakes, when can it drive our kids to practice? you know, that can't come soon enough. Uh, well, it's within two, three years.”

CodeX (OpenAI's new launch) enables launching 20 concurrent AI agent processes that can be stitched back together, representing a different paradigm from vibe-coding that few have named yet.

“What you saw with Codeex, you can launch, you know, 20 concurrent processes and then stitch them back together again. And that's that's very different from vibe coding. No one's named that yet. And Greg Brockman when he rolled it out said, 'Yeah, I'm so bad at naming things, we're just going to call this codeex.'”

Building products in emerging categories is easier and faster than ever because coding costs are low and capital requirements are low, allowing entrepreneurs to test hypotheses at high frequency with minimal downside risk while exploring defensibility.

“I think a lot about the fact that, you know, some of the midjourney actually I think is an A16Z darling where the, you know, cash flow gets into hundreds of millions of dollars of very high margin uh revenue and the cost of the build, you know, because coding is getting so cheap and so automated, the cost of a build of something like that is lower than ever. So, you've got very rapid growth of the revenue, very low uh capital costs, nothing to lose by jumping in there.”

Bitcoin will achieve major significance when aggregate Bitcoin holdings exceed aggregate gold holdings at gross economic level, creating fundamental asset revaluation.

“the thing that I think will become really powerful is when we have people owning at gross level more Bitcoin than they do a gross level gulp. And I think that'll be amazing.”

Siri voice recognition and translation still fail on basic tasks (spelling names, translating speech) after years of investment, symbolizing Apple's systematic AI incompetence.

“Why Apple still hasn't cracked AI. This is an article that came out. Let me just read this real quick. Siri overhaul delayed repeatedly. New features failed internal tests and missed 24 and 2025 rollout goals. Apple spent billions on AI chips and startups, but internal disagreements on budget allocations led to lack of GPU supplies... And I just I really want to riff on that while we have you. uh because it's really clear to me that the number of opportunities way outstrips... With that said, fixing things like voice translation in Siri, like come on, it's just a daily reminder that they can't do AI.”

Apple spent billions on AI chips and startups but experienced internal disagreements on budget allocation that led to insufficient GPU supplies for development.

“Apple spent billions on AI chips and startups, but internal disagreements on budget allocations led to lack of GPU supplies.”

Krea is an exemplary technical founding team: two AI researchers living in a house with all engineers, working seven days per week, demonstrating the intense single-minded focus required to win at the edge of AI.

“Krea, which is a really cutting edge sort of research and consumer technology company that brings together all of the creative tools in one product... These are two incredible AI researchers and sort of artists and enthusiasts. They live in a house up in Pack Heights. They live with all their engineers. They work seven days a week.”