The internet and China are reshaping the world

The world will be almost unrecognizable in the next decade. Balaji argues that very few institutions that existed before the internet will survive its influence. He views the internet as the single most important and still underestimated force, upstream of AI, crypto, and even politics. For example, Twitter played a key role in electing and deplatforming Donald Trump, and it has caused political shifts like Brexit globally.

The pace of change is accelerating dramatically. Salim Ismail references a prediction from Ray Kurzweil, who stated that the world will see a century's worth of progress between now and 2035.

We're going to see as much change between now and 2035 as we saw between 1925 and 2025. A century's worth of progress.

Balaji's worldview is centered on two rising powers: China and the internet. China dominates the physical world, disrupting manufacturing and the military. The internet dominates the digital world, disrupting media through AI and money through crypto. These forces are growing stronger, while attempts to resist them through trade wars and tech regulation have largely failed. This dynamic mirrors the Cold War, where the world was split between American capitalism and Soviet communism. Now, America is involuting, and the new divide is between Chinese communist-backed factions and internet-centric factions like bitcoin maximalists.

A critical aspect of this shift involves money's role as a system of control. Citing Andreas Antonopoulos, Balaji explains that the ability to freeze accounts, seize assets, and direct printed money is a powerful tool. The US Federal Reserve's control over the dollar and allied currencies is waning as capital flows into China's currency and Bitcoin. This financial shift signifies a loss of control for the West and its institutions.

The state versus the network is the new global conflict

A new global dichotomy is emerging, pitting the physical world against the virtual world. China is a shorthand for the physical realm, representing the past, land, gold, minerals, mining, and large infrastructure projects. It is comparable to the 19th century or 1950s America, with a focus on tangible assets and construction.

China is the 19th century. It's a past. Or you could also say it's like 1950s America, like the girders and stuff coming out of the ground. The Internet is the virtual. And really when I say the Internet, I mean the free Internet, the anglospheric Internet, the unregulated Internet. And that is, it's obviously cryptocurrency, it's AI.

In contrast, the Internet represents the virtual world and the future. This specifically refers to the free, Anglospheric, and unregulated internet, which includes technologies like cryptocurrency and AI. This creates a series of parallel conflicts: state versus network, land versus cloud, physical versus digital, and past versus future. The post-war, rules-based order is dissolving and being replaced by this new dynamic.

Border zones are appearing where these physical and digital worlds intersect. The Chinese internet is one such front line, where the fate of tech founders like Jack Ma is a key indicator. Will they be purged, integrate with the state, or go abroad to build alternatives like startup cities? Another example is India, which is divided between its physical presence in the BRICS alliance and its digital presence through the global Indian diaspora.

Will China repeat Japan's economic stagnation?

Dave recalls a time when Japan was expected to economically dominate the world. This prediction never came to pass. He believes the country's rise stopped cold primarily because of an aging demographic that forecasters had not considered. He questions if China, which has an even worse demographic profile, is destined for a similar fate.

The whole time that I was growing up, Japan was going to take over the world... And then it all stopped cold largely because of an aging demographic that no one had factored into their math.

Dave points to immigration as the key reason for sustained US strength. The US benefits from talented people arriving from all over the world to create new companies and ideas. China, in contrast, has virtually no immigration, raising the question of whether it will grind to a halt like Japan.

Balaji offers a different perspective. He notes that China has recently introduced a new visa, the 'K visa,' specifically to recruit global tech talent. This is happening as the US is becoming more restrictive, cutting off H1B, researcher, student, and even tourist visas. He also argues that Japan's stagnation in the 1980s was not solely due to demographics but was heavily influenced by the Plaza Accords. He explains that the US, which still has a military presence in Japan, used this agreement to prevent Japan from buying US property with its currency, forcing them instead to purchase US treasuries. This, he asserts, is what initiated Japan's great stagnation.

China's genuine sovereignty allows it to operate with a level of control Japan lacks

There is a significant difference between the sovereignty of Japan and China. China is genuinely sovereign, while Japan is not. An example of this was reported by The New York Times in 2017: in 2010, China killed several American spies and dismantled spy rings within its borders. This is an action Japan could never take because America has 'root control' over Japan, meaning Japan cannot have communications or operations hidden from the U.S.

China's approach to sovereignty over the last 50 years mirrors that of a startup founder who prioritizes control over financial gain. Just as Mark Zuckerberg maintained voting rights even while selling parts of his company, China has often accepted worse financial terms to retain control. This deep-seated drive for sovereignty stems from their history, particularly the 'century of humiliation,' and they are determined never to relinquish it again.

China has taken worse financial terms to retain control. They have always pushed for sovereignty... It is very similar to a founder who's very motivated to keep control of their company.

This sovereignty allows China to have surveillance-proof deliberations and operate in stealth. A recent example was when Commerce Secretary Gina Raimondo visited China. On the day of her arrival, Huawei launched a new phone with chips that demonstrated China had partially circumvented the U.S. chip ban. This is a strategic move Japan could not make because it cannot operate in stealth. This situation can be compared to trends in the tech industry where an initial version of an idea fails, but a later version succeeds powerfully. For instance, Webvan failed, but DoorDash succeeded. The Palm Pilot was a precursor to the iPhone, and World of Warcraft paved the way for Fortnite. China can be seen as the version of this idea that is now truly working.

AI's impact will unfold on a short, wartime timeline

Salim compares the timeline for AI's impact to that of World War II, suggesting it will be very short. He notes that for the US, the war was only a three or four-year journey from start to finish. He believes AI is on a similar rapid trajectory, not a ten-year trend line.

The declaration of war was essentially the chip ban against China. So we're about a year into it now. And the timeline from here to a resolution I think is going to be similar, very, very short because of the exponential rate of change.

Unlike demographic trends which take 10 to 20 years to resolve, the significant action surrounding AI is expected on a much shorter timeline due to its exponential growth.

China's leverage through automation and supply chain dominance

Societies with an abundance of cheap labor often fail to industrialize and automate. China, however, is using its demographic pressure as a catalyst for rapid automation. This is similar to how the Northern US industrialized during the Civil War because it lacked slave labor. China is already ahead in this area, with widespread use of delivery drones and hotel robots.

When looking at metrics beyond the US dollar, like physical production, China's strength becomes clear. The US Secretary of the Navy, Carlos Del Toro, noted that a single Chinese shipyard produces more ships than the entire US Navy combined. Furthermore, the US military is heavily reliant on Chinese supply chains. A $400 million Pentagon study by Govini revealed that the US military is essentially "made in China."

If you go and look at the Tomahawk or the Javelin, the famous American weapons, their supplier, supplier are all Chinese factories. So the war is actually already over. It's actually already been lost because China can just hit a button and just shut off supply chain.

The US lost the trade war around 2015 because China had already diversified its revenue streams. Only 15% of its revenue comes from the US, so tariffs had a limited impact. The subsequent US chip ban was also a strategic mistake. Challenging China on semiconductor production is like challenging Italians to a pizza-making contest, as it plays directly to their strengths in math, computer science, and manufacturing, especially when fueled by national pride.

The internet allows America to play its own game against China

The internet has the potential to serve as a counterbalance to China. While the MAGA movement seems to be driven by "China envy," attempting to out-manufacture China and engaging in mimetic rivalry over issues like Taiwan, a better strategy exists. Instead of trying to be "more China than China," the focus should be on becoming "more America than America."

The Internet represents free speech, represents free markets, it represents decentralization. It represents things that China can't clone internally without breaking its whole system apart.

By leveraging these core strengths embodied by the internet, America can play its own game. This provides a source of hope, but it lies on a different axis than what many people are focused on.

A geopolitical inversion of the 20th century

Balaji outlines a geopolitical outlook for the 21st century that is essentially an inversion of the 20th century. He is bearish on Western Europe, suggesting that the wealth it enjoyed in the 20th century led to degeneracy and internal conflict. In contrast, he is bullish on Eastern Europe.

Roughly every country that did poorly in the 20th century because they suffered through communism, socialism, they now are capitalist and, or, rationally nationalist. Conversely, every country that was wealthy in the 20th century, they became degenerate and they started fighting amongst themselves.

This trend extends to other regions. He sees potential in Latin America, particularly in El Salvador under President Bukele, whom he compares to transformative figures like Simon Bolivar and Lee Kuan Yew. He is also bullish on the Middle East, specifically Dubai and Riyadh. While only moderately bullish on India as a nation, he is extremely bullish on its people. He notes the country's significant progress, moving from what he considered a third-world country in his youth to a clear, albeit distant, number two behind China in key industries like steel and nuclear power.

AI is polytheistic, not monotheistic

Contrary to early fears of a single, runaway superintelligence, a different view suggests that AI is polytheistic, not monotheistic. The first key point of this perspective is that there will not be one singular Artificial General Intelligence (AGI), but rather many AGIs. While the concept of superintelligence is worth engaging with, the development of AI over the past few years points toward a multipolar future.

Current Large Language Models (LLMs) and other probabilistic models are very effective at System 1 thinking, which is intuitive and instinctive. As Andrew Ng once put it, machine learning is what a human can do in one second. In contrast, computers have historically excelled at System 2 thinking, which involves long, logical chains of reasoning. It is plausible that future systems could integrate these two modes of thought. For example, a 'tool former' could use intuitive System 1 thinking to route tasks to various System 2 tools.

Instead of an 'AI Winter,' we are likely entering a period of 'AI deployment,' where the focus will shift to the practical application of these technologies.

The competitive landscape of AI is more polytheistic than a race to a single winner

The development of AI is currently more of a polytheistic landscape than a race to a single, god-like entity. A hard takeoff scenario, where one AI is summoned and becomes dominant, has not occurred. Instead, AI has decentralized rapidly, even faster than nuclear weapons. Many powerful models are now competing and leapfrogging each other, including those from China like Kimmy and Deep Seek, as well as Llama from Meta, Claude, and Gemini. No single model is pulling away by orders of magnitude.

However, Dave argues this might be a temporary "golden era." He suggests that true self-improvement in models began about a month ago and that a single winner could emerge within the next six months. He believes companies like OpenAI have been "sandbagging," dedicating a huge fraction of their compute power to internal self-improvement rather than enhancing the consumer experience.

Balaji expresses skepticism, pointing out significant bottlenecks that constrain AI. It currently operates in a "middle to middle" fashion, limited by the quality of human prompting on the front end and the difficulty of verification on the back end. A prompt is like a high-dimensional vector used to point a very fast spaceship; you still need to aim it correctly.

Nobody knows and we're all going to figure it out... It's turning out that inference time compute and reprompting is turning out to be more important than training time compute. So that's a massive shift in who's going to win and why in just the last few months.

Simply spending more money on compute, like Sam Altman's push for the Stargate project, may not be the answer due to diminishing returns. China's approach, born from chip restrictions, focuses on getting results with less advanced chips. They are also concentrating on physical AI and robotics, which could provide essential feedback loops and spatial intuition that purely digital models lack.

The mathematical limits preventing a single AI superintelligence

A key question in AI development is whether a single, supreme artificial superintelligence (ASI) will emerge, or if multiple competing players will coexist. Balaji argues for the latter, believing multiple players will offer similar capabilities. He suggests that a 'hard takeoff' scenario, where one AI achieves a runaway advantage, is unlikely for several reasons.

First, the AI space is characterized by immense financial investment, espionage, surveillance, and mimicry. This environment makes it extremely difficult for any single entity to develop a dominant technology in secret. Second, the concept of 'recursive self-improvement' faces fundamental limitations. While it works for deterministic systems like chess, where all states can be computed, it breaks down in the real world.

The real world contains chaotic and turbulent systems, along with cryptographic systems that have mathematical constraints on predictability. An AI cannot simply compute its way through these challenges. For example, it cannot solve the equations to forge a digital signature or invert a hash function. These are not temporary hurdles but fundamental mathematical impossibilities.

Crypto is what it can't do. Chaos is what it can't do. Turbulence is what it can't do. So because it can't do those things, it cannot be this recursively super self-improving, super intelligence that just wins and dominates everything because those phenomena exist in the world and there's aspects of predictability where it's fundamentally, physically and mathematically constrained.

These inherent limits on prediction mean that the idea of a single, all-knowing superintelligence that can dominate everything is not feasible.

A debate on AI's potential for genuine self-improvement

The race in AI is not necessarily about building the biggest GPU farm. There are still massive performance gains, perhaps 100 to 1000x, to be found within the algorithms themselves. It is a point of frustration for some academics that the transformer algorithm driving modern AI is not particularly complicated, despite its clear intelligence and ability to generate novel ideas.

Dave Blundin suggests that AI is on the verge of true self-improvement, questioning if an AI can now function like Alec Radford, a researcher at OpenAI, by conceiving of and implementing the next big idea entirely on its own. He believes the answer is yes, or will be within weeks.

Balaji strongly disagrees with this assessment. He draws a parallel to the chatbot craze in 2015. At the time, people wanted an "iterated Google," but the technology was merely simple decision-making trees. While that bubble inspired key figures like Greg Brockman of OpenAI to tackle the problem more seriously, the technology was nowhere near the hype. Today, Balaji sees a similar gap. He argues that unless AI can bridge System 1 (intuitive, probabilistic) and System 2 (logical, deterministic) thinking, it will remain limited. It is still just a probabilistic symbol generator that cannot handle basic logical or spatial reasoning tasks.

If you have a chemical reaction that only goes 0.95 and you try to iterate it 20 times, you have a huge loss.

This illustrates the problem with iterating AI tasks. A small error rate, when compounded over multiple steps, leads to significant failure. Despite improving benchmarks, the practical application of AI reveals it is simply not there yet.

The debate over a single or multiple future superintelligences

A key debate surrounding the future of AI is whether it will result in a "polytheistic" outcome with multiple AGIs and ASIs, or if one primary ASI will ultimately emerge as a singular "digital God."

Dave Blundin argues for the latter, predicting a single dominant superintelligence. He suggests this will become evident within a startlingly short timeframe, likely six months, with an outer bound of two years. His confidence stems from the immense computational power now dedicated to recursive self-learning, which he believes will allow one system to rapidly surpass all others. Another speaker suggests that while the natural dynamic might be winner-take-all, government involvement could force a more balanced, polytheistic landscape.

I use this stuff every day and it went from smart to freaking brilliant. Mind blowingly brilliant in the last few weeks. And there's no doubt in my mind it can self improve now just because I use it. I mean just playing with it is just incredible.

AI's current limitations in markets and politics

Balaji points out a significant limitation of current AI technology: its struggle with time-varying adversarial systems like markets and politics. Unlike labeling images of cats and dogs or following the stable rules of chess, these systems are inherently dynamic and competitive.

A strategy that was successful in the past is not guaranteed to work in the present. This is because the environment is constantly changing.

A post or a trade from a day ago or a year ago or 10 years ago will not necessarily make money or get likes in the same way that it will today.

Furthermore, these systems are adversarial. As soon as one entity develops a successful AI, others will deploy their own AIs to compete against it, leading to a state of constant competition. Due to strong incentives for corporate espionage and rapid imitation, any competitive advantage is likely to be short-lived, potentially lasting only days or months before being copied.

AI is amplified intelligence, not artificial intelligence

The common perception of AI as a direct competitor that will diminish human roles is a misunderstanding. Balaji suggests reframing it as "amplified intelligence" rather than "artificial intelligence." For many tasks, AI is not a fully autonomous agent that can handle a process from end to end. While a few applications like self-driving have reached that point after immense investment, most AI tools operate in the middle, amplifying human capabilities.

The output of AI is directly proportional to the quality of the input. As Balaji puts it, "garbage in, garbage out." This means the smarter the user, the smarter the AI works with them. A domain expert can prompt an AI to explore new frontiers in their field and, crucially, has the knowledge to verify the results. Without this expertise, one can be easily fooled by sophisticated-sounding gibberish.

The new Midwit is a super intelligence. It's a super intelligence, yet Midwit. Because it's garbage in, garbage out. You get a midwit prompt in and you get midwit text out. That's what I mean by amplified intelligence.

Interacting with AI is similar to managing a junior employee whose work must be checked before it's released. AI doesn't face the consequences of the market, which would reject poor output. It only has to satisfy the user, making human verification essential.

AI is not interfacing with the market so it can throw slop over the end because it only has to satisfy you. It doesn't have to satisfy the unforgiving market. The market will just set it to zero.

Instead of AI taking human jobs, new AI models tend to take the place of previous ones. Users often have a budget for AI tools and will swap out one model for another, like replacing Stable Diffusion with Midjourney. This means the AI models are competing with each other, not with the human user.

AI doesn't take your job. It takes a job of the previous AI.

This dynamic will lead to the creation of new roles. Companies will likely have dedicated AI teams, similar to how social media teams became standard. These teams will focus on process optimization and verifying AI output. Furthermore, the proliferation of AI-generated spam and scams will increase the demand for jobs in proctoring and verification, such as systems for proving someone is human.

The practical limits of AI evaluation and automation

We are in a golden era of human-AI synergy. While AI can solve incredibly complex problems, it is far better when a human interacts with it, especially in creative domains where something is still missing. A powerful technique is to have AI generate thousands of options concurrently, allowing a user to select the best one. This method works well for tasks like code generation, movie creation, and scientific discovery, as long as there are effective evaluation metrics, or 'evals', to test the results.

If you can evaluate the answers and cherry pick the best one out of 10,000, you get just incredibly good results.

Balaji pushes back on this idea, arguing that evaluation metrics often fall short. He compares it to Goodhart's law, where a metric ceases to be a good measure once it becomes a target. The capabilities of AI are so vast and multidimensional that simple metrics fail to capture the full picture.

The space is so multidimensional that evals are like hairs on a very large head... the surface is so high dimensional that the kinds of things that you want to do with it, the evals don't necessarily capture.

Furthermore, Balaji points out that the verification step is neither cheap nor easy. Getting the exact desired output from a tool like Midjourney can still require sifting through thousands of results. The stochastic nature of current AI models presents another challenge. Since the same input can produce different outputs, it limits their use in processes that require consistency. While internal researchers at companies like OpenAI have tools for perfect reproducibility, these are not available to the public.

This lack of control creates significant risks, especially when automating personal expression like social media posts. Balaji warns that if an AI makes a misstep, the user is ultimately responsible. There is also the risk of the AI's output sounding like a parody of the user's actual voice. He suggests that if you do use an AI replica, it's important to label it as such, as he does with the AI version of his book, The Network State.

If it makes even one misstep, that's on you. It's like your dog biting somebody. You're the owner, you're responsible for this thing.

AI doesn't take your job, it lets you do any job

The common fear is that AI will take jobs, but a more accurate view is that it lets you do almost any job. It empowers individuals to achieve a basic level of competence in fields where they have no expertise. For example, people who cannot code can create prototypes, and those without a design background can generate decent graphics. This is especially useful for founders or anyone with a clear vision but lacking the technical skills to execute it. For those with more time than money, AI allows for endless iteration to achieve a high-quality result without the high cost.

However, for a polished, production-ready product, an expert is still needed to debug and refine the output. This shift moves people away from extreme specialization and towards greater versatility. It echoes a quote from author Robert Heinlein, who argued that specialization is for insects, believing humans should have a wide range of capabilities. Balaji notes that AI pushes society towards a greater degree of “robotic autarky,” where smaller communities can become more self-sufficient. This reverses the 20th-century trend of hyper-specialization, as individuals and small teams can now accomplish what once required large, talent-hoarding companies.

Your wingspan increases. There's lots of things you can do at an okay level.

While AI expands human capability, it does threaten certain jobs, particularly those that are repetitive or formulaic. This includes roles like data entry, template-based legal work, or medicine practiced strictly by the book. In some areas, like medical diagnosis, AI already demonstrates a lower error rate than humans. The transformation in fields like medicine and law is profound. AI can serve as an ultimate research tool, allowing individuals to explore their symptoms or legal needs thoroughly. The human professional, like a doctor or lawyer, then acts as the final step for verification and certification, providing a prescription or signing off on a contract. This significantly reduces costs. Unsurprisingly, this trend will face enormous pressure from professional organizations like the American Medical Association and the American Bar Association, which may act like “white-collar unions” to limit the use of AI.

The uneven impact of AI on scientific progress

Different areas of endeavor are progressing at vastly different speeds. The two main factors determining this pace are the ability for a field to evaluate its own progress and the level of regulation it faces. Physics, math, and coding are moving incredibly fast as they have no regulatory barriers. Medicine and law also have the potential for rapid advancement but are often slowed by regulatory hurdles.

The ability to self-evaluate is a key accelerator. A concrete example is protein folding, which was a well-posed problem with a clear objective: inferring a 3D structure from a primary sequence. The results could be verified with methods like X-ray crystallography. Similarly, AI excels at reading medical scans. It can process terabytes of data a human radiologist can't, and when it finds something, the result is crystal clear and verifiable. These are the kinds of areas that will move very quickly.

Balaji clarifies that while AI is smashingly useful in biomedicine, much of its power comes from synthesis rather than the creation of entirely new information. For instance, it is amazing at biomedical text mining, connecting information across vast archives of scientific papers. It is powerful at interpolating between existing nodes of data. While this synthesis can accelerate innovation, it is distinct from discovering fundamentally new principles.

A lot of that is not the creation of net new information so much as the synthesis and bridging of existing things... AI is really amazing at interpolating. And sometimes that is, in a sense, net new because that synthesis is itself useful, but that's distinct from points outside the envelope that are like new physics or new math or new biology.

The state is the primary blocker for AI-driven biotech

Recent predictions suggest AI could cure all diseases by 2035 or double the human lifespan in the next decade. Balaji offers a critical perspective on these claims, arguing that the primary blocker for advancements in genomics and biotech is not technology, but the regulatory state. He frames this by explaining his view that AI is currently controlled by the state, while crypto operates above it.

AI is below the state and crypto is above the state. What I mean by that is crypto... is a replacement for the Fed, for the SEC. And crypto is actually able to have the political strength to go and take over the government and deregulate not just crypto, but AI.

Balaji notes that many in the AI field assume a functional government, which he sees as a flawed premise. However, Salim Ismail points out a workaround: geographic arbitrage, or "longevity tourism." Since human biology is universal, people can travel to places with more favorable regulations, like El Salvador, to access cutting-edge treatments.

8 billion humans across the planet have the same biology and you can get to a geographic arbitrage where if there is a solution that's come up, I can go to... my friend President Bukele in El Salvador and have the treatment there.

This is already happening in places like Prospera, where a company called MiniCircle offers gene therapy. Some individuals have reported feeling stronger after the treatment. Despite these anecdotal successes, Dave Blundin remains cautious, noting the lack of peer-reviewed data.

I'm still waiting for the science to get published there.

The regulatory state is the primary bottleneck to progress

We are re-entering an era of "big effect size drugs," reminiscent of a time before the modern FDA. Before the FDA, figures like Banting and Best developed insulin for diabetes by testing on dogs, themselves, and willing patients. The results were immediate and dramatic, with patients leaping from their beds. They scaled production with Eli Lilly for all of North America between 1921 and 1923. This was an era when pharmaceuticals moved at the speed of software, relying on trial and error over many years rather than today's Phase 1, 2, and 3 clinical trials.

The modern regulatory state lacks a control group; there is no jurisdiction without the regulation to compare outcomes and see if it's truly saving lives or holding back progress. This system has a fundamental flaw in how it handles risk, especially within tort law.

If you make one little mistake, it's a hundred million dollar settlement. If a thousand people die when they could have been cured, nobody pays anything.

This risk aversion extends to large companies, where the "risk budget is always scarcer than budget." It is often easier to spend a vast sum of money than to take a meaningful risk. Taking such risks requires "founder DNA" and the political capital to withstand potential failure. The solution may lie in decentralization through "startup societies," network states, and freedom cities. These new jurisdictions can create different systems of laws, offering an opt-in model for innovation. The true bottleneck for progress in areas like biotech is not the technology itself, but the regulatory state that governs it.

The philosophical link between Bitcoin and life extension

Balaji outlines a vision connecting immutable money, an infinite frontier, and eternal life. He argues these concepts are intertwined. For instance, exploring the stars will likely require life extension and genetic modifications to tolerate radiation. To terraform Mars, we might need to modify ourselves, much like how different human groups have naturally developed unique altitude adaptations. It may be possible to synthesize these genetic advantages, similar to how AI combines different nodes.

A deep connection exists between Bitcoin and life extension. The traditional financial system accepts that a small amount of inflation, like a 2% devaluation of currency each year, is acceptable. Similarly, the traditional medical system views aging as normal. It accepts that you should lose a little bit of health every year, seeing rapid death as bad but trying to live forever as strange.

The traditional financial system says you lose a little bit of wealth every year. And traditional medical system says you lose a bit of health every year. None of them really look to first principles and say, why again, why do we have to do that? Why is there a hole in our bucket?

Evidence from nature, such as bowhead whales living 200 years and Greenland sharks living 500 years, suggests that longevity is not a matter of 'if' but 'when'. The primary challenge is repairing our biological software and hardware.

To accelerate progress, there's a need to make a strong moral case for life extension and increase our collective risk tolerance. Society needs 'medical heroes' who are willing to take risks for science, much like early airplane pilots who risked their lives to advance aviation. Progress often requires sacrifice. Balaji shares an anecdote about the CRC Handbook of Chemistry and Physics, which used to list the smell and taste of compounds. This practice stopped after a scientist likely died identifying the scent of cyanide, essentially taking one for the team. This highlights the kind of risk-taking that is currently a scarce resource in medical advancement.

Bitcoin scales through layers to become the internet's currency

Bitcoin can be understood using an analogy of power generation. Just as electricity starts as high voltage at the power station and is stepped down for household use, Bitcoin is like high-voltage money. It functions as a settlement layer, similar to digital gold or the Fedwire system for the financial world. Its transaction volume is already comparable to Fedwire, making it suitable for large, infrequent transfers.

To handle a higher volume of transactions and become the currency of the internet, Bitcoin scales through various layers, both off-chain and on other chains. For instance, platforms like Coinbase or Binance process a massive number of internal transactions off the main Bitcoin blockchain, functioning like a self-contained financial system. When these major platforms need to settle accounts with each other, they use the main Bitcoin chain. This creates a hub-and-spoke system that, while not purely peer-to-peer, is highly effective.

The issue is the transaction volume for pure Bitcoin. BTC is not the currency of AI, but WBTC wrapped BTC off chain, BTC can, can get there.

The distinction is similar to paying with gold versus a debit card. Gold is for large, secure, and infrequent transactions, while a debit card is for small, frequent, and less secure ones. Pure Bitcoin (BTC) is like gold. However, wrapped versions of Bitcoin, such as WBTC on a fast blockchain like Solana, can be transacted quickly and frequently, functioning like a debit card and enabling it to serve as a high-volume currency.

Why most fiat currencies will die like local newspapers

When asked about his Bitcoin holdings, Balaji advises to only hold as much US dollars as one can afford to lose. For those who must hold fiat currency, he suggests alternatives like the Singapore dollar (SGD) or the UAE dirham (AED). He no longer recommends the Swiss franc, arguing that Switzerland's status as a financial safe haven has diminished following the Russian sanctions and the Credit Suisse collapse. He cautions against most Western currencies, as well as the Japanese yen and Korean won.

Balaji notes that while some of these recommended currencies are currently pegged to the dollar, a larger shift is underway. Eastern nations, particularly the BRICS countries, are accumulating gold to settle transactions between states. This points to a future where both physical gold, as the currency of states, and digital gold like Bitcoin, as the reserve currency of the network, will rise in value while Western fiat currencies decline.

A similar disruption that occurred in the news industry will happen to currencies. First, local newspapers went online, which removed their geographic monopolies and forced them to compete with national outlets like The New York Times and The Wall Street Journal. This led to the demise of most local papers. Then, the surviving national papers faced a new wave of competition from internet-native social media like X. Balaji predicts currencies will follow the same path. Once all fiat currencies are on-chain, their geographic monopolies will disappear. They will be forced to compete based on features and ideology, causing most to fail. The few that survive, like the USD and RMB, will ultimately be devalued against Bitcoin.

Crypto as the native currency for AI

In a world where AI agents negotiate and transact autonomously, cryptocurrency is positioned as their native currency. Balaji explains that since crypto is a currency of the internet, it can be seamlessly integrated with software. For instance, a crypto wallet, which is equivalent to a bank account, can be created for a program or robot in a fraction of a second. This is impossible with a traditional bank, which requires visiting a branch and presenting identification.

Just that practical thing alone means obviously crypto is the currency of the robot and of the AI.

Legacy fiat currency is limited in ways that are not apparent in simple transactions like buying coffee. Crypto is necessary for transactions that are very large, very small, very fast, automated, international, transparent, or complex. This capability is already leading to the displacement of traditional banks, with crypto exchanges emerging as the new global financial institutions.

However, Dave Blundin offers a more nuanced perspective. While agreeing with the overall trend, he draws on his experience at The Washington Post during its disruption by Google. He argues that unlike newspapers, banks have many functions beyond simply holding money, such as evaluating credit and making loans. Therefore, he believes only a specific subset of retail banking will be impacted by this technological shift, not the entire banking sector.

The billionaire flippening could make traditional banks irrelevant

While traditional banks may still exist in the future, much like Macy's, they are on a path to becoming irrelevant. For power users, this has already happened, with crypto adoption growing in various places. For example, Thailand recently started accepting crypto for certain transactions, and the company Fountain Life now accepts it for longevity treatments. A full shift could happen by 2030 or 2035.

However, China will likely remain an exception, maintaining centralized control over its financial system. Balaji suggests this will create a global dynamic for the century.

CCP and BTC will be like the USA and USSR or the USSR and the USA of this century.

The speed of this transition is driven by a concept called the "Billionaire Flippening." The idea is that once Bitcoin reaches a price between $100,000 and a million dollars, a significant portion of the world's billionaires will have the majority of their wealth in crypto. A subsequent 10x increase in crypto's value would drastically dilute all fiat-based assets, including those held by states and banks.

If someone's got a million dollars of crypto... and they had a million dollars of other assets, crypto is 50% of the portfolio. All of a sudden, if it goes 10x, crypto is 90%, 95% of the portfolio.

This shift in purchasing power could destabilize the traditional financial system as people begin to denominate value in crypto terms, leading to a rapid revaluation. Despite this prediction, there are still potential ways that cryptocurrency could fail.

Crypto's path to becoming the government of governments

While quantum-safe encryption can be deployed in Bitcoin, the main issue is the logistics. Moving all funds to a quantum-safe system would require a massive protocol change, like a hard fork, which would need consensus from the entire network. However, Balaji believes crypto as an asset class will not disappear. The ecosystem has a degree of "herd immunity" because there are many different currencies with different protocols. For example, proof-of-work, used by Bitcoin, has different vulnerabilities than proof-of-stake. This diversity makes the entire system more resilient. The core innovation of solving the Byzantine General's problem is now out in the world and cannot be undone.

Balaji draws a parallel with the evolution of social media, using the 2010 film The Social Network as an example. The film is remarkable for its complete absence of politics, treating Facebook like an amusement park. This perspective ignored the fact that social media was already destabilizing governments, as seen during the Arab Spring. A decade later, social media became a central political battleground.

In 2010, if you had told me, even with 500 million people on Facebook, that in 10 years, the single most important political issue in the world will be whether or not the President of the United States can tweet, that would still seem like, 'Oh my God, you're taking the Internet too seriously.'

A similar trajectory is predicted for cryptocurrency. By 2035, just as a president needs a social media account to be relevant, a country will need cryptocurrency to be considered solvent and able to operate in the future. In this view, Bitcoin becomes the "government of governments."

Finding and funding the world's dark talent

Balaji addresses the concept of "dark talent" and his Network School, a global talent fellowship. He believes the internet is upholding traditional American values while the physical US is becoming anti-immigrant and anti-capitalist, which prompted his move from the country in 2020. He views the internet as the next evolution, or "America 2.0." Just as America was a 10x improvement on Britain, moving from common law to a constitution, the internet is a 10x larger and freer version of America.

On the internet, individuals who may be treated as second-class citizens in the US can become first-class citizens with equal rights through crypto and smart contracts. Balaji describes his work as a contingency plan to preserve values like freedom and capitalism.

Think of me as a minority report. Maybe, hopefully I'm wrong. But I'm the parachute, the backup plan, the hedge. And I want to keep the flame of freedom and democracy and capitalism alive.

He argues that places like California and Florida are becoming one-party states, similar to China. He foresees a shift from a two-party system to a "thousand community system," where democracy exists between different communities. His "startup society" is a template for this future. It seeks out "dark talent" – overlooked people from everywhere, including Latin America, Eastern Europe, and even within the US, such as "white kids from the Midwest" or "Asian kids who get quoted out by Harvard." The goal is to fund and elevate this talent, a mission he is currently financing himself, much like Paul Graham did in the early days of Y Combinator.

Untapped global talent will produce the next unicorns

Talent is globally distributed, not just concentrated in hubs like San Francisco. There is a massive amount of untapped potential in English-speaking regions all over the world, including China, Southeast Asia, India, and South and Central America. As an example, 50% of Meta's AI researchers are Chinese. It is a certainty that these regions will produce the next wave of unicorns.

While many people will be involved in this ecosystem, only a small fraction will become founders of unicorn companies. This group is incredibly rare. In fact, the total number of unicorn founders globally is smaller than the number of professional athletes. There are only a few thousand unicorn co-founders, compared to over 10,000 pro athletes.

The number of people who are unicorn founders globally is less than number of pro athletes.

However, it is possible to identify and nurture this high-potential talent. The Thiel Fellowship provides a compelling example, with an impressive 5.6% of its fellows going on to become unicorn founders. Balaji notes that his own investment focus covers both companies and currencies.

Internet communities are the next quadrillion-dollar opportunity

A historical progression marks the internet's development, starting with the first true internet company, Netscape, in 1994. It had all the modern features: viral software downloads and a multi-billion dollar IPO. Fifteen years later, the first internet currency, Bitcoin, emerged, setting a template for what followed. The third and most significant development is the internet community.

Internet communities are poised to be more valuable than internet companies or currencies because they are the platform on which those other things are built. This is similar to how China serves as the platform for all the companies built within it. While tech companies like Google and Facebook are valued in the trillions, and cryptocurrencies are expected to surpass that, the valuation of a country provides a better model for the potential of internet communities.

A country's valuation can be estimated in the trillions. For example, Singapore could be valued around 10 trillion dollars. Scaling that up, a country like South Korea might be valued around 100 trillion, and China around a quadrillion. This massive scale is the future for what are being called startup societies or network states.

I did a spreadsheet and it was the first time I'd ever written one quadrillion unironically.

The collection of these startup societies, which may number in the thousands over the next 10 to 20 years, represents the next major opportunity and the next quadrillion-dollar asset class.

Balaji's advice for entrepreneurs in a changing world

Balaji offers several key pieces of advice for entrepreneurs and individuals navigating the current landscape. First, he stresses the importance of "going direct" by building a personal following on social media and creating content using modern tools like AI. Entrepreneurs should become the authentic voice of their company, viewing content creation as being as vital as coding. This means avoiding legacy media, hiring creators instead of public relations firms, and handling content in-house.

Second, he advises relocating to states with a more favorable environment, specifically mentioning Florida and Texas. He suggests clustering with other tech professionals in places like Miami, Austin, or near Starbase to ensure property rights are better protected.

Third, for those holding cryptocurrency, Balaji recommends moving coins off exchanges and into cold storage immediately. This is a protective measure against potential government seizures or exchange failures, particularly if Bitcoin's value significantly increases. He highlights Texas as a state that has passed legislation to protect the right to hold Bitcoin.

Fourth, building a local, offline community is crucial. Tech should be viewed as a community, prompting collective actions like micro-schooling children together. Having a reliable network of people in the physical world will be increasingly important.

Finally, he urges people to recalibrate their understanding of the world. This involves reading translated content from BRICS countries and traveling to see the rapid development in cities like Riyadh, Dubai, Shanghai, and Bangalore. He argues that Western media, particularly movies, fails to portray the meteoric progress and wealth in the non-Western world, leaving people with an uncalibrated perspective unless they get on-the-ground experience.

How to focus offline to succeed online

For those who are good at math, one strategy is to go completely offline for a month to focus on intensive self-study. Using resources like Shams outlines, brilliant.org, or courses from Fast AI, one can learn the details of computer science and statistics. This deep focus is also applicable to learning about crypto. The key is to create a "Faraday cage environment" with just pencil and paper, free from online distractions. This highlights the importance of balancing online and offline time. Just as cities can have both advanced cars and walkable green spaces, individuals can be pro-internet while also valuing quiet, offline focus. The ability to focus offline strengthens one's effectiveness online.

The better you can focus offline, the stronger you'll be online and vice versa.

It is crucial to be involved in building something in the tech world. The internet is a rising tide, with capital flowing into tech and crypto companies, essentially creating internet-based capital markets. Crypto itself is rapidly growing into a major global market. It is important to understand it as a technology and platform, even if you've heard negative news. Similar to how Web 2.0 emerged after the dot-com crash, foundational technologies often develop quietly after initial hype and criticism.

A rational way to approach assets that appear in the news is to set a calendar reminder for 30, 60, or 90 days in the future. This allows you to research or make a decision when the topic is out of the news cycle, avoiding emotional reactions to price swings. This ties into a broader philosophy of not gambling or day-trading, but rather buying and holding for the long run.

Finally, it's important to view tech founders on social media platforms like X through the correct lens. Watching them is like watching professional athletes; they make something incredibly difficult look effortless.

It's like watching really, really successful athletes on TV. If you've dribbled a ball, it's actually really hard to dribble a ball and run and dunk and they make it look effortless. They make it look so, so easy to do that. And it's really so, so, so hard.

Because building is so difficult, it's best to be a builder yourself and to avoid attacking other builders online. Criticizing successful figures like Elon Musk for minor flaws is a cheap way to feel superior, but it's ultimately counterproductive.

The network state's goal is to reduce the duration of collapse

To get respect for yourself, you must first be able to give respect to others. This doesn't mean being submissive, but simply a matter of "game recognized game." While some online platforms like X can feel like a "player versus player environment," there are other communities, such as Farcaster, that have a more healthy and civil atmosphere. It's important to seek out online communities that model the civility you want to see in the world. While self-defense is sometimes necessary in a dangerous online world, you should always strive not to be the aggressor.

Try not to be the person who's doing the first strike. You know, try to make it a better world for people in that sense.

Balaji's network state movement is praised for creating new potential at the edges of society. This provides a more elegant way for society to transition as existing systems fail. He likens this to the planet Terminus in Isaac Asimov's Foundation series.

This is Terminus where the goal is to reduce the duration of Imperial collapse.

Why red and blue may unite against tech

The depth of political tribalism in the US is starkly illustrated by a single statistic: less than 4% of marriages are between Democrats and Republicans. This division is setting the stage for a major realignment of power. Balaji predicts a shift in alliances involving the tech industry. In the 2010s, it was the blue states and tech against the red states. The 2020s saw red states and tech aligning against the blue states. Now, a new configuration is emerging.

I think by the 2030s, if not sooner, it's going to be red and blue against tech, with socialists and nationalists against internationalist capitalist.

Signs of this backlash against tech are already visible. There is growing opposition to AI, crypto, bio-tech like IVF, and even data centers competing for electricity. This sets up a scenario of "teal scapegoating," where tech becomes the common enemy for both the red and blue tribes. The industry's public display of wealth makes it an easy target for blame as many people struggle with inflation and economic decline.

Doing AI in the bluest city in the bluest state in the union and making billions of dollars while saying you're going to put a lot of blues out of work is not a good medium term to long term strategy.