What this covers
Balaji Srinivasan, then CTO of Coinbase, speaks with Tyler Cowen about how blockchain and decentralized networks will disrupt the institutions that currently gate access to capital, talent, and education. The conversation ranges across technology, finance, and politics, grounded in the claim that software and tokens will let individuals and small teams bypass traditional intermediaries—venture capitalists, banks, nation-states—by dramatically lowering the cost of starting currencies, markets, and financing mechanisms. Central to his argument is the idea of routing around gatekeepers: where Sand Hill Road once held a monopoly on startup funding, ICOs and Ethereum now allow worldwide participation; where geographic clustering once confined engineers to the Bay Area, the hub-and-spoke model and future telepresence systems will let workers compete globally.
The conversation turns contentious on several fronts. Srinivasan contends that the network is eroding the state's control over education—children now self-educate online rather than absorb state-directed acculturation, a shift with geopolitical consequences. He defends blockchain's energy consumption by reframing it as infrastructure for property rights, not just payments. He proposes that tokens will become a third pillar of capital structure alongside debt and equity, and that prediction markets could bring accountability to opinion-writers the way venture returns constrain investors. On healthcare, he draws distinctions: software will disrupt diagnostics and IT, but gene therapy and vaccines will resist startup-style innovation for years. He also sketches how telepresence robots within five to ten years might make migration itself a software problem, with workers comparison-shopping jurisdictions via remote labor. Throughout, he emphasizes quantifying decentralization—measuring how many entities you'd need to compromise to control a blockchain—as a way to rank coins honestly rather than by rhetoric.
Srinivasan argues that the blockchain, decentralized financing, and global connectivity will route around legacy gatekeepers (VCs, banks, the state's education system, geography itself), unlocking talent and capital worldwide, while small software-savvy nations and emerging technologies like telepresence reshape work, identity, and power.
- Blockchain dramatically lowers the threshold to start a currency, a marketplace, and financing, breaking VC geographic clustering and financial-sector fee extraction
- The network is capturing children's education hours from the state, eroding state acculturation
- Telepresence and drones will turn even migrant labor into software, letting workers comparison-shop jurisdictions
Blockchain removes centralized intermediaries by making it trivial to launch competing currencies, markets, and financial instruments.
- Financial fees will structurally decline because you can now set up a new payment rail and contract-execution system in your dorm room, providing fundamental competition at the lowest level—analogous to how the internet let you route around the Postal Service, NBC, and record companies—even though these new rails are volatile, you can root-and-branch exit the system in a way previously impossible.
“you can set up a new payment rail and contract execution system in your dorm room. That’s new because it provides fundamental competition at the lowest level.”
- The old era requiring startups to come to Sand Hill Road to get funded and then go to Wall Street to exit is over; Ethereum and ICOs have decentralized funding—the last centralized piece—worldwide, though the regulatory environment will be worked out differently in different countries.
“the old era where you had to come to Sand Hill to get your company funded and then go to Wall Street to exit is over.”
- A token differs from a gift certificate in two ways: you have total custody (owning the private keys lets you transfer to anyone without third-party intervention, unlike Macy's transfer restrictions), and programmability—you can have millions of them, programs can consume them, they can be bought and traded programmatically and used as inputs for API calls—so the agility of the value dramatically increases.
“you have, in theory at least, total custody that say you own the private keys. You can move it to somebody else without any third party’s intervention.”
- Blockchain has dramatically dropped the threshold for starting a new market—not just making it easy to start your own currency, but also to start your own marketplace—which may drive innovation in prediction markets in the years to come.
“the blockchain has not just made it much easier to start your own currency in your drawing room; it has also made it much more easy to start your own marketplace. So the threshold for starting a new market has dropped dramatically.”
- New blockchain business models won't have the same fee structure as old ones—they'll still make money but will tax what is scarce, which may not be the wire transfer itself (as shown by Germany repatriating tons of gold over years at millions of dollars when Bitcoin could move the value for a fraction) but rather money-management fees; the key result is dramatically increased choice because blockchain makes it easy to start a competing financial outlet.
“they’re going to tax what is scarce, and that may not be necessarily the wire transfer itself.”
- Tokens will become an integral part of every suitable company's capital structure over the next 10 years as a third financing instrument: debt is a claim on future cash flows, equity is a claim on future liquidations, and network tokens are a claim on a company's future digital network value—and just as not every company suits debt or equity, token financing suits only companies with a network effect, with debt tokens and equity tokens also emerging.
“What is debt? When a company issues debt, it’s a claim on the future cash flows of the company. What is equity? ... it’s a claim on future liquidations of the company. What are tokens ... ? They’re a claim on the company’s future digital asset value or digital network.”
Decentralization's viability depends on whether underlying systems can actually resist compromise across miners, developers, and exchanges.
- The most persuasive argument against crypto's future viability is that the decentralization model doesn't actually work and the underlying blockchains can be compromised; decentralization can be quantified by the minimum number of entities (miners, developers, or exchanges) you'd need to compromise to control a coin, allowing coins to be rank-ordered by purported decentralization.
“the minimum number of entities that you needed to compromise was the degree of decentralization of the system.”
- Peter Thiel's contrast between blockchain's decentralizing aspects and AI's centralizing aspects is complicated by recent developments allowing machine-learning coefficients to be calculated locally on phones via a MapReduce-style approach where data stays local and the central server is just a coordinator that doesn't store the data, as in Google's published research.
“there’s recent breakthroughs or developments in AI which allow for calculations of the coefficients locally on your phones.”
- While it is technically possible to exit Facebook or Google (delete accounts, ban domains via local DNS, or move to China), it's too hard for ordinary people given the ubiquity of trackers; the right approach is to build exits and alternatives so people can leave and don't feel constrained, rather than the platforms being inescapable.
“the right approach is to build exits and build alternatives such that folkscanexit from those things and they don’t feel constrained by them.”
- Comparing Bitcoin's energy consumption to PayPal's servers is inapposite because Bitcoin and Ethereum are not payment systems but a system of property rights and contract law, more comparable to Delaware or US government functions; even all of Bitcoin's energy consumption is cheaper than the Brink's trucks, US military, and police required to protect property rights in the American context.
“what Bitcoin is, and then more generally Ethereum and all the blockchain stuff is, is basically a system of property rights and contract law. It’s more comparable to, let’s say, Delaware or even the US government”
- Healthcare can be segmented by gradation of risk and susceptibility to software disruption: health IT/EMR and diagnostics are within software's wheelhouse, prosthetics is tractable as software/hardware robotics, but therapeutics, gene therapy, and vaccines (under CDER/CBER) remain difficult and not amenable to startup techniques for some time.
“That’s kind of a gradation of risk and also gradation of, let’s say, susceptibility to a software-based attack.”
- Vaccines are an edge case for live-and-let-live values because, unlike choices of clothes or food, refusing vaccination when one carries a contagious disease harms others' lives and livelihoods; it is an education problem software cannot solve, and self-quarantine by anti-vaxxers would be the appropriate response.
“if they don’t believe in vaccines, and their child or themselves has a contagious disease, then it can screw up other people’s livelihoods and lives.”
Digital networks and tokens let individuals worldwide finance themselves and participate in the global economy without traditional gatekeepers.
- A worldwide search for smart kids can be solved in nontrivial part by tech—using a MOOC and taking the highest scorers (as already happens, e.g. 'The Boy Genius of Ulan Bator' who outscored MIT kids in a circuits class)—but the missing piece is plugging those high-scoring kids into the global economy so they can contribute and get their dividends, where blockchain may play a partial role.
“use aMOOCand take the highest scorers.”
- In roughly 20 years, an individual with promise could issue a personal token—selling, say, 20 percent of their future earnings for token-based financing rather than taking a student loan—and anyone worldwide (India, Japan, Kenya, Brazil) could chip in, opening up financing at a very individual level in a way never before possible, because broadcasting and financing are becoming low-threshold activities like sending a tweet.
“you might issue a personal token. That would be something where you say, “Hey, look, I’m going to sell about 20 percent of my future earnings for token-based financing.””
- There will be a large number of 'new digital blue-collar jobs'—Mechanical Turk-style tasks like labeling images (cat or dog, one cluster or two) for machine learning—performed worldwide from Nigeria, Brazil, India and similar places, which Earn.com is already enabling.
“the new digital blue-collar jobs, which are Mechanical Turk–type jobs.”
- The global English-speaking population outside the US (India, Nigeria, and other countries with large English-speaking minorities) is underrated; now that they have smartphones, it's only a matter of time before tens or hundreds of millions of them participate heavily in US internet discussions—as cryptocurrency already shows interest-based communities forming across locations.
“There’s many more English speakers outside the United States than there are within if you include India, if you include Nigeria”
- When Anybots/Double Robotics telepresence robots converge with Boston Dynamics humanoid robots and VR bodysuits (within 5-10 years), people will be able to 'immigrate' over an internet connection by animating a remote robot, making a country's ability to interdict internet connections its de facto immigration policy—and migrant labor itself could become software, letting workers comparison-shop jurisdictions for the best wages.
“someone can “immigrate” to your country over an internet connection, connect to a robot, animate it, and start walking it around. Now your ability to interdict those internet connections is your immigration policy.”
- Drones are very underrated because they are not just flying drones but also swimming and walking drones used for construction, and combined with telepresence you could Skype into a robot in drone mode that walks to a destination while you sleep—as depicted in the film Surrogates where people never leave home and rejuvenate a new body if their telepresent one is destroyed.
“There’s different kinds of drones. They’re not just flying drones. There’sswimming dronesand there’swalking dronesand so on.”
Traditional institutions—academia, journalism, schools—have perverse incentives; open-source and tutorial-based models are more transparent and verifiable.
- Journalism redesigned from scratch would involve far more part-time on-the-ground contributors with smartphones (like Korea's OhmyNews on a global, paid scale), and tutorials are a more trustworthy category than war or celebrity reporting because tutorials have an inbuilt fact-checking mechanism—you can independently verify every line by executing it yourself, unlike facts about Afghanistan or the Kardashians that you cannot independently verify.
“a tutorial on a new piece of software or how to fly a drone. Tutorials are something you can independently verify since youcanfact-check every single line by going through it yourself.”
- The 20th-century division between research and teaching will dissolve into a more organic, commercially-aligned career path where one starts as a practitioner, does angel investments (funding the equivalent of graduate students), researches, and then teaches via a MOOC—because a would-be Stanford professor can start and exit a company for millions rather than chase a few-hundred-thousand-dollar grant.
“Why should I go and spend all this effort for a few hundred thousand dollars for a grant when I can spend a comparable amount of effort, start a company, exit it, and get a few million dollars, and then fund the equivalent of graduate students via angel investments?”
- Academia's incentives are inverted relative to open source: in open source bugs are expected, everything is transparent and published first then reviewed, and reproducibility comes from executable source code, whereas in academia retractions are career-damaging blemishes and many papers are not transparent in methods or data—so academia should move toward an open-source-like culture, which can only happen when grant incentives are realigned.
“in open source, bugs are expected. Reproducibility is a function of your ability to execute the source code. Everything is fundamentally transparent because it’s all published. And you publish first, and then people review it”
- American public school is a stultifying 12-year 'mandatory minimum' (the 'Schoolag Archipelago') that slows learning; Balaji's own education effectively began only in 2001 when Google got good enough for self-education, and he would crowdfund a tutor rather than send his kids to American public school.
“Life in the United States begins with a 12-year mandatory minimum, the Schoolag Archipelago.”
Bay Area equity structures align incentives so all parties benefit from company exit rather than political maneuvering.
- The most fundamental management technique the Bay Area pioneered is equity/stock options, because it extends the two-by-two game-theory matrix to the 2^k case so that all k parties realize their highest payoff comes from the win-win quadrant (the company exiting) rather than from off-diagonal political moves that benefit them at others' expense.
“What equity does, it takes that from the two-by-two to the two-by-two-by-two-by-two case, the 2^kcase, where even if you havekparties, all of them realize that their highest payoff is from the win/win quandrant”
- Venture capital's geographic clustering in the Bay Area is driven partly by the under-appreciated fact that VCs take board seats and don't want to fly 6,000 miles for each of many board meetings, so they prefer companies within driving distance—and this preference compounds across A, B, and C rounds, constraining startups to the Bay Area.
“A VC does not want to fly 6,000 miles for every board seat if they’ve got 10 board seats and four board meetings a year per company.”
- Companies are converging on a hub-and-spoke model where Bay Area HQ hubs handle in-person executive functions, management, and whiteboard sessions, while international spokes do engineering work—because face-to-face is necessary for managers and executives but less and less necessary for engineers who primarily work with electrons.
“it is true that executive functions, management, things like that where you’re managing people, then face to face is good. Whiteboard sessions are good, et cetera. If you’re really just primarily working with electrons, you don’t need face to face as much”
- A useful meta-algorithm for attacking unstructured problems ('how do we increase sales / raise funds / do X') is list-rank-iterate: make a list (of prospective doctors, VCs), impose a ranking function (which zip code, has this VC invested in companies like ours), then iterate brutally with a preset limit (do 150; if no hits, switch to a new list/strategy).
“you can start by making a list ... Then impose a ranking function on them ... Then you just iterate brutally through this. The key thing is you put a limit.”
Small, software-savvy nations and China's AI-driven governance will outcompete large traditional democracies in the technology era.
- The countries that will win the future are small, software-savvy nations like Estonia, Singapore, and Israel whose leaders are almost like software CEOs (Estonia's Ilves a computer scientist, Lee Hsien Loong the Cambridge wrangler, Netanyahu an MIT graduate), with China as the major large-country exception that treats instruments like the Great Firewall as first-class policy.
“the countries that are going to win the future will be Estonia, Singapore, Israel — very software-savvy small countries where their leaders are basically almost like software CEOs.”
- China's use of AI on surveillance footage—pulling faces from cameras into a database to track a human across their entire life—moves surveillance toward an all-seeing, all-knowing god of myth, leveling up the threat of a surveillance camera from human recognition to automatic computer recognition; this may yield a very low-crime but dystopian society that many people might nonetheless choose to live in.
“what they’re almost moving AI towards is almost like the all-seeing, all-knowing god of myth.”
- Chinese tech workers are commonly said to work harder (12 hours a day, 6 days a week) and, more distinctively, feel they are trying to show China can make it on the world stage—a civilizational hunger to prove something—whereas for Americans it's more of an individual than a civilizational drive, as seen at the Beijing Olympics.
“the Chinese that I’ve met, at least, really feel like they are trying to show that China can make it on the world stage. They have a hunger to prove something ... For the Americans it’s more of an individual thing as opposed to a civilizational thing.”
- The Chinese internet (Jack Ma, Pony Ma) is far more innovative than the broad mass of Americans realize—the show Silicon Valley's opening titles feature only Alibaba and not the rest—while Chinese entrepreneurs are hyper-aware of US developments, making the awareness asymmetric.
“The Chinese internet is alotmore innovative than, let’s say, the broad mass of the US has realized.”
Technology creates tradeoffs: kids gain self-education and peer communities but face downsides like impaired focus and biometric identity risks.
- Technology produces 'more upside, more downside' on balance: kids face downsides like impaired eye contact, possible inability to focus at medium distance, and difficulty with deep reading due to distractions, but gain free access to the Library of Alexandria and the ability to self-educate and find like-minded peers rather than enduring one-size-fits-all schooling—net positive, though downsides need amelioration.
“That’s my one-liner for the future: more upside, more downside.”
- Biometric authentication like retina scans is good because it's unique to you but bad because there's no password reset—so if password reset is a core feature of a security system, biometrics may not work for it; future identity verification will vary by context, mixing email, phone, username, and biometric auth to fit each use case.
“retina scans and biometrics aregoodin that they’re unique to you. They’rebadin that there’s no password reset, for example.”
- The network (laptops, iPads, iPhones) is capturing the state's six hours of daily compulsory education as kids self-educate and plug into their own sub-communities during school, eroding the state's control and acculturation—which historically (since Bismarck) was designed to raise children obedient to the state—with significant medium-term consequences as kids absorb diverse ideologies younger.
“the network is getting a lot of those hours and taking them away from the state because the kids are plugged into their laptops and iPads and iPhones andself-educating during the period they’re supposed to be having compulsory state education.”
Innovation requires ignoring trend signals and mining all of human civilization for problems technology has not yet solved.
- The most underemphasized success factor in the Bay Area is the degree to which you can and should set your own path rather than following TechCrunch and Twitter, because it's very hard to tweet or TechCrunch your way to innovation; the biggest innovations (Satoshi's Bitcoin, Musk's SpaceX) had no validation from the feed—so look at all of human civilization for what technology has not yet moved into, where the opportunity is underexploited.
“it’s very hard to tweet or TechCrunch your way to innovation. I really do think you have to tune out alotof what the Valley is thinking and then innovate.”
- Ramanujan is underrated because he is not yet world-famous (only famous in India), is India's Einstein, and is spiritually important as a rags-to-riches figure who, though never rich, came from destitution in the middle of nowhere and made a mark—embodying the internet's potential to find and lift up left-behind people via information and blockchain financial tools.
“I think he’s underrated because he’s not yet world famous. He’s famous in India. ... He’s India’s Einstein.”
- Singapore is Balaji's number-one place to live because it balances many factors: it's hot (relevant since many South Asians in Silicon Valley have low-level vitamin D deficiency from insufficient sun and computer time), friendly to capitalism, full of skilled immigrants, easy to start a tech company, safe, and good for raising kids—unlike alternatives such as Switzerland that get cold.
“Singapore is friendly to capitalism. It’s got a lot of skilled immigrants there. You can start a tech company really easily. It’s safe. If you’re raising kids, it’s good.”
The West should exercise modesty about intervening in other nations' governance, and cultural shifts like large families drive outcomes regardless of logic.
- As bad as a Chinese surveillance dystopia might be, there should be modesty about the West intervening, because (a) it may not be within the West's power to prevent it, and (b) the Chinese people may not want the West to prevent it—and the controlled internet, which many predicted would collapse innovation in the early 2000s, instead got very far.
“as bad as I think that future is, (a) I’m not sure it’s within our power to prevent it, and (b) I’m not sure the Chinese people wouldwantthe West to prevent it”
- China, Russia, and Eastern Europe have been 'civilizationally vaccinated' against communism by their 20th-century experience, giving them a near-term 30-year memory of why moving in that political direction was a bad idea, so they established states that—while not all free-market capitalist—are no longer hardcore communist mass-murdering regimes.
“because they were civilizationally vaccinated against the lure of communism due to the 20th century, the Chinese, the Russians, the Eastern Europeans have a civilizational memory ofwhythat was a bad idea.”
- People who have large families will 'win the argument' for large family sizes not by logic but by evolution—because they have more kids who have more kids—and this will keep the world somewhat religious on average since religious families tend to have more children; Bryan Caplan's book makes an interesting case for large families.
“the folks who have large families are going to win the argument simply because they’re going to have more kids and those kids will have more kids ... It’s an evolutionary argument. Darwin’s going to have the last word.”
Prediction markets enabled by blockchain will create accountability for opinion-writers comparable to venture capital returns.
- Private companies haven't adopted prediction markets partly because they already have real markets (stock, commodities, bonds) where players use real money to win, whereas internal toy markets lack the dollar totals to attract serious players, attract only academic nerds rather than people gaming the system, and therefore lack the predictive value of real markets.
“those real markets have real money, and all the folks playing in those markets are really playing to win. Whereas a toy market — at least that’s what we were perceived as at the beginning — we don’t have the dollar totals required to be interesting.”
- Prediction markets, newly feasible with blockchain, could bring to opinion-writers the accountability that venture capitalists already have via returns—because people who opine for a living traditionally have no record of whether their predictions were right or wrong.
“there’s no record of whether they were correct or not in their predictions. People who opine for a living don’t traditionally have a way of collecting, “Hey, I was right about this, wrong about that.” Venture capitalists do. It’s called returns.”