What this covers

The conversation was recorded live at the Digital Asset Summit in New York, with Jordi Visser of 22V Research as the only guest. He dates the move from the chatbot era to the agentic era to late November and early December, tying it to the Opus 4.5 release and the daily model releases from Anthropic that followed. From there the talk covers macro frameworks, productivity, labor, geopolitics, compute supply, and a closing section on how to learn to use AI.

The range is wide. On markets, Visser argues that agents will make advertising worthless, since ads are built for human cognition. He also proposes splitting the Mag 7 into three hardware names (Tesla, Apple, Nvidia) and four software names, and says oil shocks now matter little next to AI spending. On crypto, he says Ethereum can be valued with discounted cash flow because it earns protocol revenue, while Bitcoin cannot, and he treats velocity (economic)) as the metric that favors Bitcoin. The labor discussion is contested. He rejects comparisons to 1970s job losses, expects unemployment to stay low because of a structural labor shortage, and locates the real damage in the broken corporate ladder for new graduates. He also describes Labor Arbitrage Inversion, where solo builders gain more than large enterprises.

The most speculative turns concern power and supply. Visser likens AI to a nuclear weapon and expects governments to seek control over private development, which he thinks would hurt private valuations (the Government Control Trajectory)). He points to acute shortages in semiconductors, DRAM and silver, and to Nvidia's Lilly deployment as an early sign of industry-specific data centers. He also argues that current frontier models are already good enough for commercial agent use. The closing advice is practical: use the tools daily, and hold a conversation with them rather than handing them one-off tasks.

Sharpest takeaway

The transition from chatbot to agentic AI represents a fundamental economic inflection point requiring 1,000x more compute, which will destroy traditional equity valuation models, compress profit margins for enterprises, and make Bitcoin the superior growth asset while creating immense wealth through velocity-based cryptocurrency systems.

  • Agentic AI increases compute demand by 1,000x from chatbot era, creating unlimited demand for hardware and infrastructure
  • Traditional discounted cash flow valuation breaks down when AI agents, not humans, make business decisions within 3 years
  • Productivity gains accrue to solo entrepreneurs using LLMs, not enterprises, collapsing the labor arbitrage that sustained software company margins

The claims · ranked46 claims · weighted by value

This asset isn't compiled yet

You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.

0.73

If you need to make long-term investments in any asset (including equities), you must account for humanoid robots and AI agents that are stronger, faster, and smarter than humans, which fundamentally changes assumptions about future corporate value creation.

normativehigh valuecontestednovelty 3/4durability 3/4· Jordi Visser

“If you're thinking of making long-term investments, if you're thinking of anything in long duration, you have to question your mind and actually ask, 'Am I making the right decision on this?' And I think this is why this is so important.”

0.73

Velocity is the critical economic metric in an AI-agent-dominated economy, and cryptocurrency is the only asset class that can capture velocity value because traditional fiat systems have regulatory guardrails preventing fast transaction speeds required by agents.

causalhigh valuecontestednovelty 3/4durability 3/4· Jordi Visser

“Velocity becomes the most important thing of GDP. Velocity is what crypto is about. We cannot have the velocity on the fiat and the guardrails that have been set up for the financial system. It's going to be related to the speed of it.”

0.69

To become proficient with AI, you must practice daily with the tools, similar to learning a sport like skiing or golf, rather than learning abstract principles.

normativehigh valueestablishednovelty 1/4durability 3/4· Jordi Visser

“If you're a 15 handicap in golf and you wake up one day and you went, 'You know what? I want to be a five handicap. How am I going to do that?' Everyone knows what to do. If you don't know how to ski and you want to go ski, you have to go put in the reps. You have to put in the miles.”

0.69

Solving major scientific and engineering challenges (cancer, fusion, autonomous vehicles, humanoids) requires orders-of-magnitude more compute than current models, extending the compute expansion cycle for 4-5 more years.

forecasthigh valueestablishednovelty 1/4durability 3/4· Jordi Visser

“If you want to solve fusion, if you want to have a an army of humanoids, if you want to have autonomous vehicles to the degree that are safe and can go across the country, that takes a lot more mem- a lot more models in terms of power and usage than we've gone through. The compute needs for that to where you'll allow humanoid to be in your house around your kids, we're nowhere near that. That's going to take a long time.”

0.69

AI is functionally equivalent to a nuclear weapon and is the most critical military technology, making it the highest priority for government funding and private sector development regardless of economic returns.

normativehigh valueestablishednovelty 1/4durability 3/4· Jordi Visser

“Because uh I mean, AI is a nuclear weapon. There's no other way to say it. It is the most important part of the military, and if you didn't think it was, um you just have to go back and think about what happened in Venezuela.”

0.69

Discounted cash flow valuation becomes impossible for software companies because AI agent adoption will make it unknowable what a company's cash flows will be 3+ years in the future, creating a collapse in software equity multiples as markets cannot price uncertainty.

causalhigh valuecontestednovelty 3/4durability 2/4· Jordi Visser

“When exponential innovation becomes disruptive, when AI agents are a part in the digital employees are now dominating the way workflows go, you cannot value how these companies will be there because humans won't be making the decisions. Agents will.”

0.68

Ethereum can be valued using discounted cash flow analysis because it generates protocol revenues (similar to a yield-bearing asset), while Bitcoin cannot be valued this way, making Ethereum acceptable to institutional asset managers but Bitcoin uniquely positioned to capture velocity-based economic value.

causalhigh valuecontestednovelty 2/4durability 3/4· Jordi Visser

“Bitcoin needs to not be valued or not be talked about from a discounted cash flow cuz it doesn't have any. Ethereum, you'll see hedge fund and asset managers will get very comfortable with Ethereum cuz you can do a discounted cash flow analysis for Ethereum.”

0.68

Bitcoin is the end game for growth assets because it will benefit from the wealth transfer created by AI-driven productivity redistribution and fiat system dynamics

causalhigh valuecontestednovelty 2/4durability 3/4· Jordi Visser

“The reason I'm here is because I'm 100% a believer that Bitcoin is the end game for growth assets.”

0.68

The comparison between 1970s job losses and current AI disruption is not valid because 1970s unemployment rose in major waves while current unemployment remains near historical lows despite AI

causalhigh valuecontestednovelty 2/4durability 3/4· Jordi Visser

“So, when we talk and when you hear all this analysis on oil, just remember, in the 1970s, we saw the unemployment rate go up in two major waves. We have the unemployment rate sitting near all-time lows.”

0.64

Lily Pod, Nvidia's enterprise data center for Eli Lilly, represents the first true enterprise data center deployment and signals the beginning of industry-specific compute infrastructure tailored to major problems like drug discovery.

factualhigh valueestablishednovelty 1/4durability 2/4· Jordi Visser

“The first real enterprise data center, the first true one, finally came out recently. This is on Blackwell. It's Lily Pod. If you get a chance, just go look it up. It's obviously for Eli Lilly.”

0.63

Oil price movements will have negligible impact on the global economy and will not cause recession because oil's relevance to the global economy has declined from 40-50 years ago, and AI spend dominates any macroeconomic impact of fuel price changes.

causalhigh valuecontestednovelty 2/4durability 2/4· Jordi Visser

“When we talk about oil prices and we talk about global recessions, it's not going to happen. I can say that with comfort unless we get into a different kind of stage where there is nuclear war or something along those lines. Because oil doesn't matter to the global economy to the scale that it did 40, 50 years ago, and it certainly doesn't matter relative to artificial intelligence, of which there are trillions of dollars going through it.”

0.63

Governments will inevitably seek control over AI development from the private sector as they realize the military strategic importance, leading to nationalization or heavy regulatory capture that will destroy private sector valuations.

forecasthigh valuecontestednovelty 2/4durability 2/4· Jordi Visser

“If the government, as this fight with Anthropic showed, actually gets to the point where they decide they need to have control of this because they're not able to use it in a way that they want from a military perspective, well, then the government owns AI, which what happens to these companies.”

0.63

The transition from chatbot era to agentic era occurred in late November and early December and represents the most important economic shift in recent history, requiring compute demand to increase by 1,000 times what was needed in the chatbot era.

factualhigh valuecontestednovelty 2/4durability 2/4· Jordi Visser

“What happened in the end of November and into December was probably the most important point of all of our lives in terms of transitioning from a chatbot era to the agentic era. We have entered the agentic side, which means the amount of compute we need has gone up by a thousand times already from what we went through.”

0.59

Supply constraints for semiconductors, DRAM, and silver are acute because infrastructure growth has not kept pace with compute demand, explaining why DRAM prices are up 400-500% since September and why PMI has risen.

causalhigh valueestablishednovelty 1/4durability 1/4· Jordi Visser

“We are at bottlenecks right now, which is why DRAM prices are up 4 or 500% from where they were in September. Why silver has gone up as much it is. Why all of these things have happened. It's why PMI's finally went higher.”

0.57

Unemployment rate will not increase significantly despite AI-driven layoffs because the US has a structural labor shortage with only 30,000-40,000 new jobs needed to maintain current unemployment levels, a labor shortage exacerbated by immigration restrictions, and critical shortages in healthcare, daycare, supermarket, and service sectors.

forecasthigh valuecontestednovelty 1/4durability 2/4· Jordi Visser

“I don't think the unemployment rate is going to go higher. I'm not one of the doom and gloom people that believes that people are going to lose jobs because we've already seen it. So, the one thing that is a mistake by people is we've already seen job losses from AI over the last 12 months. We've created zero jobs. When you take out healthcare... We need more daycare jobs. We need more people working in supermarkets. We need more people in service jobs.”

0.56

Opus 4.5 release and subsequent Anthropic releases represent the start of the agentic era with daily releases of increasingly capable models

factualhigh valueestablishednovelty 1/4durability 2/4· Jordi Visser

“What happened in the end of November and into the December was probably the most important point of all of our lives in terms of transitioning from a chatbot era to the agentic era. And that started with Opus 4.5. We've now subsequently moved through an enormous amount of Anthropic releases. I mean, they happen daily.”

0.55

The psychological damage from the destruction of the corporate ladder and slower initial job placement for new graduates is more significant than actual job losses, creating a different economic scenario than 1970s-style mass unemployment despite similar surface indicators.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Jordi Visser

“What we do have is insanely psychologically damaging situation in the jobs market. There's no more corporate ladder. AI's going to destroy that. People coming out of schools, they can't get jobs as quickly as they used to. People are working in jobs that they didn't go to school for because they need to pay the bills because affordability is so difficult. That is a very different scenario than people losing jobs in a recession. That is a psychological thing.”

0.54

Current frontier models (as of early 2025) are already sufficient for commercial agentic use cases and further improvements in model capability provide diminishing returns relative to compute costs for most applications.

factualhigh valuecontestednovelty 1/4durability 2/4· Jordi Visser

“In terms of where they are now, you don't need any better thing. I always say that I've been paying the max now for a year. I I don't pay any more money, but the models get a hundred a thousand I can't even tell you how much better they get.”

0.53

Advertisements will become worthless when AI agents make purchasing and business decisions, because advertisements are designed for human cognition and decision-making, which will no longer be the majority decision-maker in the economy.

causalhigh valuespeaker onlynovelty 3/4durability 2/4· Jordi Visser

“When Google's sitting there fighting and people like, 'Yeah, there'll be ads.' I'm like, 'How are ads going to be worth anything when an agent is making the decision?' Those ads and everything is for human beings. Human beings will not be making the the majority of decisions within 3 years.”

0.52

The Mag 7 should be split into three hardware companies (Tesla, Apple, Nvidia) and four software companies (Google, Amazon, Meta, Microsoft) with fundamentally different risk/return profiles and valuations.

normativehigh valuespeaker onlynovelty 2/4durability 3/4· Jordi Visser

“There are three hardware companies in the Mag 7 and there are four software companies. I hate the software companies. So, Google is the best of the bunch, but when you get through Amazon and you get through Meta and you get through Microsoft”

0.52

Enterprise software companies have experienced margin compression in the past 4 months because AI productivity gains accrue entirely to solo entrepreneurs and small teams using LLMs, not to enterprises with large employee bases, creating a negative labor arbitrage for corporations.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Jordi Visser

“The productivity boom is coming to people that have no employees, like myself. Cost me $12,000 a year to pay for all five LLMs. It cost me another 5,000 to buy a couple MacBook Pros that I use with my Mac with my OpenClaw. That is far better than the people I used to pay. And so, the labor arbitrage for enterprises is a negative, but on the flip side, everyone who's an entrepreneur, who has a good idea, can make money.”

0.52

AI agents operating 24/7 to service billions of requests create an infinite feed demand for compute, analogous to running an infinite restaurant with an infinitely long line of customers that must be fed perpetually.

definitionhigh valuespeaker onlynovelty 2/4durability 3/4· Jordi Visser

“It literally is like opening up a restaurant and having a line out the door of trillions of people, and you have to find some way to feed them. There is no way that we will have enough compute for all of the agents until we've at least gotten to the point four or five years from now”

0.51

Household net worth increased by $15 trillion in the past year while the S&P 500 rose 15%, showing resilience of wealth creation independent of traditional stock market dynamics.

factualhigh valueestablishednovelty 1/4durability 1/4· Jordi Visser

“In the last year, the S&P 500 has gone up 15%. Household net worth in the country has gone up 15 trillion dollars.”

0.50

You must build a conversational relationship with AI models rather than using them for isolated task completion, because this interaction pattern develops dependency and competence more effectively than task-based usage.

normativehigh valuespeaker onlynovelty 2/4durability 3/4· Jordi Visser

“you have to build a relationship with it through communication, through verbal communication. These models have gotten so good that whether it's on your AirPods instead of listening to the podcast or the music you're going to listen to, walk down out of this building and go down and have a conversation on any topic you want to have”

0.49

There are hundreds of mispriced companies outside the Mag 7 that could outperform significantly over the next decade due to AI disruption, indicating a massive opportunity set beyond the widely followed names.

forecasthigh valuespeaker onlynovelty 2/4durability 2/4· Jordi Visser

“I could give you guys a hundred names right now that I spend time on that are completely mispriced for the trillions of dollars, as Jensen Huang says, that are going to go through the economy over the course of the next decade.”

0.49

The remedy for psychological labor market damage is for displaced workers to build relationships with AI tools, invest in learning, and develop critical thinking skills, which will make them employable at enterprises because most employees lack AI tool proficiency.

normativehigh valuespeaker onlynovelty 2/4durability 2/4· Jordi Visser

“The good thing is people can go spend time with artificial intelligence. They can invest money in themselves, ask their parents to give them a holiday present, and give them access to the top end of any one of the LLMs. They can start building stuff and start learning how to think critically again as opposed to being trained in school, and I guarantee you then you can go out and get a job at any of these big enterprises cuz they don't have enough people that use the tools at this point.”

0.49

Silver is up 60% over the past 6 months while gold is up only 20% because silver is used in every technology and will be a bottleneck in drone and humanoid production, making it functionally equivalent to a rare earth metal.

causalhigh valuespeaker onlynovelty 2/4durability 2/4· Jordi Visser

“Gold is only up 20% over the last last 6 months. Silver is up 60%. I'll take silver every day over gold for the rest of time because it is a necessity in every single piece of technology that you use. If you go through and do the math on how many drones equal a tank, so it is massive in terms of how much silver we need.”

0.49

Young workers (ages 15-22) should prioritize building AI agent relationships and learning AI-enabled workflows rather than taking traditional internships, as enterprise internship programs cannot teach AI proficiency and will become economically obsolete.

normativehigh valuespeaker onlynovelty 2/4durability 2/4· Jordi Visser

“If he wants to talk to a genius, he can just talk to any LLM, and he'll get an answer that's far better than the practitioners that are trying to teach him how to do things at this point.”

0.48

The Iran situation will be partially or fully resolved within 3-6 months, following historical patterns where geopolitical conflicts are either resolved or plateau in impact, and market impact will not alter the AI spending trajectory.

forecasthigh valuecontestednovelty 0/4durability 2/4· Jordi Visser

“Three months from now, the situation in Iran will be either over or it will be less bad than it is today. And that's the way markets work. That's the way war works”

0.48

The lack of compute infrastructure is analogous to traffic congestion at an airport: just as JFK lacks capacity for the volume of travelers trying to use it, the compute infrastructure cannot handle the volume of AI agents trying to operate simultaneously.

definitionhigh valuespeaker onlynovelty 1/4durability 3/4· Jordi Visser

“But that doesn't change the way that we go through things. We compare the oil time of the past, and we go to the '90s, and we go to 2000, and we it's different for AI. There is no comparison. We've never needed this much compute ever.”

0.48

Families should involve all members in AI-assisted problem-solving (like planning a trip) rather than individuals learning alone, creating household familiarity and removing fear

normativehigh valuespeaker onlynovelty 1/4durability 3/4· Jordi Visser

“If you have young kids, have a conversation at the table about where you want to travel to. Do an investigative thing and say, 'Hey, we want to go on a trip in the summertime.' Do something that involves everyone in the house cuz as you get used to it and as you start to use it, it becomes easier and easier.”

0.47

Most people do not understand the distinction between chatbots and agentic AI, which explains why market sentiment has not yet rotated despite the massive implications of agent adoption.

normativehigh valuespeaker onlynovelty 1/4durability 2/4· Jordi Visser

“I just don't think most people know what the difference is between a chatbot era and agentic era, but it is a massive difference.”

0.45

Bitcoin will attract investment once software companies (Mag 7) stop producing returns and wealth managers search for alternative growth assets, creating a narrative-driven capital flow rather than fundamental value discovery.

forecasthigh valuespeaker onlynovelty 1/4durability 2/4· Jordi Visser

“The narrative for Bitcoin may not be attractive now because you've been able to make money in the Mag 7. That's not working anymore. When you're looking for something and you call your wealth manager and say, 'Hey, how do I make money? What's in the growth asset? What's working?' And Bitcoin is working, all of a sudden you'll see people come up with a new narrative that basically is what I'm talking about.”

0.45

Micron is trading at a 4x PE on next year's earnings because the market has priced in the assumption that memory demand will not be sustainable, but this represents a buying opportunity if you believe in sustained compute expansion.

causalhigh valuespeaker onlynovelty 1/4durability 2/4· Jordi Visser

“Micron is the biggest position in my own personal portfolio. It's trading now at a four PE off next year's earnings, somewhere below a four at this point. So, they've discounted the fact that memory is not going to be there forever.”

0.45

Gold is not a good investment relative to Bitcoin, semiconductors, and silver because gold does not benefit from the agentic AI economy while the other three commodities/assets are critical to compute infrastructure and value velocity.

normativehigh valuespeaker onlynovelty 1/4durability 2/4· Jordi Visser

“I will take from this point Bitcoin and I will take semiconductors and I will take silver over anything related to gold at this point cuz I think the agentic world has changed the equation.”

0.45

Last year people confidently dismissed AI as a bubble despite not using it themselves, creating an information asymmetry that identified those with genuine understanding and real upside opportunities.

normativehigh valuespeaker onlynovelty 1/4durability 2/4· Jordi Visser

“last year, the entire year was spent with people doubting it, calling it a bubble, not using it, but saying it with this veracity of of comfort. Um the confidence level that I heard from people who literally do not use it at all uh just made me understand better that there was a huge arbitrage here”

0.45

Perplexity Computer is a superior tool to ChatGPT for financial analysis and building custom dashboards, websites, and tools because it operates as an agentic system that executes multi-step tasks autonomously.

normativehigh valuespeaker onlynovelty 1/4durability 2/4· Jordi Visser

“Perplexity Computer is an agentic system that will do things for you that are incredibly simple. It will build dashboards, it will build websites, it will build whatever you want.”

0.41

Nvidia's valuation multiple is compressing while revenue remains flat, but the company will lead the market recovery in the first half of 2025 because demand for semiconductors remains huge and Nvidia is not subject to the same disruption as software companies.

forecasthigh valuespeaker onlynovelty 1/4durability 1/4· Jordi Visser

“they'll probably be the company that leads us from whatever low we make in the first six months of this year, they'll again be a story going forward because the demand for them is is huge.”

0.39

CPI will exceed 4% year-over-year within 2 months because import price inflation numbers are already extremely high, making this a tactical opportunity to exit S&P 500 positions before the market reprices inflation risk.

forecasthigh valuespeaker onlynovelty 1/4durability 1/4· Jordi Visser

“we got an import price inflation number today, which was extremely high, and I'll be showing this on my weekly video just to show that CPI over the next 2 months will get above 4% year-over-year. You don't want to be long the S&P 500 when CPI is that high.”

0.39

The market bottom will occur within 6 weeks (from early 2025), when recession fears peak, and this will be the optimal time to buy equities because markets are most negative just before recovery.

forecasthigh valuespeaker onlynovelty 1/4durability 1/4· Jordi Visser

“When the pendulum shifts to recession is the best time for people to start buying into stocks. We'll be there sometime in the next 6 weeks.”

0.37

Visser personally uses five different LLMs and pays $12,000 annually for subscriptions, demonstrating his conviction about AI's importance and practical superiority to hiring traditional employees or contractors.

factualhigh valuespeaker onlynovelty 0/4durability 1/4· Jordi Visser

“Cost me $12,000 a year to pay for all five LLMs. It cost me another 5,000 to buy a couple MacBook Pros”

0.36

Jensen Huang stated Nvidia will achieve $1 trillion in annual revenue, representing 2.5x growth from current levels, but the market is pricing in only ~$600 billion, indicating massive underpricing of Nvidia's growth trajectory.

normativehigh valuespeaker onlynovelty 1/4durability 1/4· Jordi Visser

“And just think about what Jensen Huang said this week at GTC or last week at GTC. He basically said we'll have a trillion dollars of revenues. Well, that's an increase of almost two and a half times where they are today. The market's not priced for that. The market has their revenue of being closer to about 600 billion in in a couple years.”

0.34

The pace of frontier AI model releases (Anthropic, OpenAI, Google) suggests a competitive flywheel where leaders generate resources to remain leaders, but new entrants cannot catch up without orders-of-magnitude more capital than startup funding provides.

causalcontestednovelty 1/4durability 2/4· Jordi Visser

“I don't think for So, let's separate two things with the frontier models and let's get into where I think they're all going to eventually go and this gets back into the government.”

0.33

Everyone is fundamentally motivated to make money, making this the universal principle driving adoption of new assets and technologies regardless of ideological or technical merit.

normativeestablishednovelty 0/4durability 3/4· Jordi Visser

“And I just know the one quality in every human being I've ever met, everyone's trying to make money.”

0.27

Visser has driven 5 hours in a Tesla with Grok without manual input, having extended verbal conversations on arbitrary topics, demonstrating that current AI models can sustain deep conversational engagement.

factualestablishednovelty 0/4durability 1/4· Jordi Visser

“I have driven in my Tesla for 5 hours with Grok, didn't put in a single thing, and I'll have conversations about anything that enters my mind.”

0.13

Microsoft is a disaster as an AI company and will not win the AI competition despite Copilot integration, in contrast to Tesla and Nvidia which have genuine competitive advantages.

normativespeaker onlynovelty 0/4durability 1/4· Jordi Visser

“Microsoft to me is a disaster. I don't know even I don't I never liked it when I worked at a company. I couldn't stand it, but the fact that they were supposed to win this whole battle with Copilot and now you've got the Satya's taking over to do this.”