Charles Hoskinson Entered Crypto Expecting Prison

Lex Fridman opens with a joke: if we live in a simulation, what language is it written in? Hoskinson points to a Microsoft Research paper on an autodidactic universe and to Stephen Wolfram's simple-rules program, then shrugs. Stuck inside a formal system, he says, you cannot fully describe the meta-language outside it, any more than a redstone computer in Minecraft can audit the Java that hosts the world.

His favorite philosophers are Bertrand Russell and Saul Kripke, though the cluster also includes Wittgenstein and Tarski. What they shared, he argues, was a drive to raise the precision of analysis for both formal languages and ordinary speech. Russell and Alfred North Whitehead spent a thousand pages of *Principia Mathematica* proving that 1+1=2, then Gödel, Turing, and Church dismantled the dream of a complete, decidable mathematics. Russell reinvented himself as an ethicist and prose stylist; Kripke, whom Hoskinson calls the Galois of philosophy, wrote logic papers in high school that prompted Harvard to invite him to teach graduate courses before he had finished secondary school.

That work had nothing to do with computers at the time. Today humans live in natural language and machines live in formal language, and the bridge is still incomplete. There is even a spoken language, Lojban, equivalent to first-order predicate calculus, with no ambiguity and, Hoskinson notes, people who write poetry in it. The engineering question is related: theoretical computer science will not make perfect programs, but it can put up guardrails, proving termination, ruling out buffer overflows, or showing that some catastrophic event cannot occur.

The art is deciding what cannot fail. In cryptocurrency that list includes funds, privacy, and, if the systems are adopted widely enough, lives. A missed block, a latency spike, or a node that has to restart is an inconvenience; chasing those ghosts, he says, costs a decade while the world moves on. The industry's incentives are inverted. Builders often get paid up front, tokens go liquid, and a latent DeFi bug surfaces six months later. Users eat the loss. "That's why you pay the guy who makes the brake software for your train last," Hoskinson says, "and you make sure he rides the train every day."

Asked what languages students should learn, Hoskinson splits pedagogy from production. Start with Python or JavaScript: the libraries are rich, the tutorials are endless, and even MIT swapped Lisp for Python in its intro course. The point is not to fall in love with a language. It is to fall in love with having a dialogue with a computer.

Cardano's stack is functional because the team had written more than a hundred papers and had to translate them into running code. The semantic gap between a C++ implementation and a LaTeX proof is huge; the compiler will not tell you that you implemented the wrong protocol. Haskell sits closer to mathematics, so the translation loss is smaller. The cost is real: weaker Windows support, thinner libraries, and slower average performance than C. Version one, he says, optimized for correctness and a high-fidelity rendering of Ouroboros, not for developer popularity. Once that blueprint exists, a Rust or Java port is possible. The other direction is pointless.

Concision is part of the bet. IOHK's Scala Ethereum client, Mantis, is about 14,000 to 15,000 lines; Bitcoin Core in C++ is on the order of 120,000 to 150,000. Less code is easier to read, document, and put under static analysis and property tests. For a career language he would now point students at Scala 3: it can look like Java, script like Python with a REPL, or grow dependent object types for proofs, and Netflix's microservices are written in it. He calls it a gateway drug to Haskell. C and C++, he thinks, have been replaced by Rust and Go. Google built Go because it hated C that much. Hate, he adds, can be a great motivator.

Plutus is a domain-specific language on Haskell, aimed at small-to-mid-size on-chain programs with deterministic resource costs. Run a contract locally and the fee you see is the fee you pay; account-model systems with global mutable state do not give you that. Bitcoin Script was too tight, and Satoshi's attempts to loosen it produced disasters such as the 2010 value-overflow incident that briefly created billions of bitcoin. Ethereum's mutable accounts make sharding hard. Plutus is the Goldilocks zone in between, designed to sit cleanly against off-chain infrastructure rather than to be the whole application.

Strip away the token price, Hoskinson says, and a blockchain is a ledger with three properties: it is timestamped, immutable, and auditable, globally or locally. Land titles, mineral rights, and water rights have always needed a record of how a claim moved from Alice to Bob to Charlie, with amounts and dates attached. Those ledgers are usually run by governments or regulated firms, which makes them efficient and also brittle to politics and war. When Syria collapsed, one of the first things ISIS did was rewrite who owned the land. When peace returns, how do you unwind that?

A blockchain offers a transnational record that persists even when it is inconvenient to a powerful actor. Hoskinson calls it an asymmetry society has not had before: synthetic laws of physics, sitting above kings and empires. Medical records are the same pattern. He had traveled to 52 countries in five years. If he were hit by a car in Zimbabwe, unconscious, a local doctor calling Colorado would be told that consent is required and the patient cannot give it. A broker that can move records across jurisdictions is the sort of middle layer a ledger can provide.

Cryptocurrency is just an application on that ledger. If you can issue property, you can issue tokens of value, then attach a monetary policy that inflates, deflates, or decays. The same machinery that protects a title or a medical file can protect ownership of a coin, fully transparent as in Bitcoin or as opaque as Zcash tries to be. Updating that ledger when many parties disagree is the job of consensus: pick who is in charge for a moment, let them assemble a block, then let the network check it. Proof of work, proof of stake, Algorand, Paxos, and Raft all share that three-step skeleton.

Proof of work is a lottery. Specialized machines buy tickets every second; whoever hits the magic number orders the transactions. The first step is enormously expensive. Bitcoin, he says, uses more electricity than Switzerland, which leaves less budget for block production and validation and creates economies of scale. Not everyone can put a mine on a volcano in El Salvador, stand up a five-gigawatt plant in Mongolia, or buy patented ASICs. Historically that concentrates into a handful of operations, a mining oligarchy.

Proof of stake replaces the physical resource with a synthetic one. If 25 percent of hash power would have won 25 percent of blocks over time, introduce randomness and let 25 percent of stake win 25 percent of the time. Designed well, Hoskinson argues, the system decentralizes as value rises, the way Bill Gates's Microsoft stake drifted from 64 percent to under 5 percent. The resource is also portable. If China bans mining, you cannot put a warehouse in a suitcase; a virtual stake can be redeployed with a click. The Goldfinger problem is the other asymmetry. A miner with enough hash to 51-percent two similar proof-of-work coins can destroy one, short the asset, and keep mining the other. In proof of stake, participation requires ownership, so destroying the chain is a net loss.

Cardano did not try to solve the whole problem at once. Hoskinson contrasts that with Ethereum's habit of bright people charging the wall. The path ran through peer-reviewed papers: GKL in 2015 at Eurocrypt, which modeled what a blockchain even is and now has more than a thousand citations; Ouroboros Classic in 2017 at Crypto, a synchronous protocol that assumed always-on nodes and was useless in real life; Praos at Eurocrypt, which added adaptive security so block leaders are not known in advance and cannot be DDoSed; Genesis at CCS in 2018, which let a joining node pick the true chain by density rather than a trusted checkpoint. Six years, about twelve papers, and still some polishing. The leap from Classic to Praos, he says, is what made a public network possible. That is what Cardano runs today.

Stake pools are the specialization that followed. Operators register on chain; wallets show more than 3,000 of them; a holder clicks delegate and can revoke later. A parameter K sets how many saturated pools the system should tend toward, with S-shaped diminishing returns past a threshold so not everyone piles onto one operator. His own pool is called Rats, king of the rats, which he accepts as the kindest title the industry has given him. The curves were not guessed. The team hired Oxford algorithmic game theorist Elias Koutsoupias and spent a year and a half on the models. Every solved step, Hoskinson says, invites ten more interesting problems, which is why IOHK keeps opening labs in Edinburgh, Tokyo, Wyoming, and Athens.

Hoskinson's one-sentence Cardano is a financial operating system, a phrase he knows people mock. Legacy rails such as SWIFT work well in rich countries and badly elsewhere. The aim is a universal protocol that moves value, represents identity, and gives a person in Senegal the same access Bill Gates has. Cardano is not that protocol yet. It is a first approximation that is supposed to evolve for decades, paid for by a treasury the system prints itself and steered by voting, the way Peter Suber's game Nomic lets players amend the rules from inside the game.

He teaches the industry as three generations. Bitcoin solved one problem: move value without a trusted third party. It is a one-trick pony, and that is why it is beautiful. By 2013 it had a billion-dollar market cap, Silicon Valley startups, and real exchanges; the idea was legitimate, so people immediately wanted what it did not have. Ethereum was the JavaScript moment, programmability bolted onto a ledger, after which ICOs, DeFi, STOs, and NFTs arrived and users started complaining that it was slow, expensive, isolated, and too big to manage with pizza and conference handshakes. Vint Cerf could once call the person on the other end of the early internet. Nobody is in charge of the internet now, so it evolves slowly.

The third generation has to be faster and cheaper as the population grows, talk to other chains, and govern itself when millions or billions of people with ice-pick rivalries share a system. Tezos leaned into self-amendment, Polkadot and Aion into interoperability, others into raw throughput. Cardano tried to hit scalability, interoperability, and sustainability at once, which is why it was slower out of the gate. Hoskinson does not believe in a god protocol. Successful DeFi, he says, will be chain-agnostic within five years, chasing liquidity, cost, and user experience the way no sane company swears lifelong loyalty to one cloud host.

His complaint about Bitcoin is not that digital gold is a bad idea. It is that the culture will not pick a purpose and commit. If the coin is a slow, expensive settlement commodity, why the years of delay on Schnorr signatures and on NIPoPoWs, a 2016 result that would give light clients and sidechains "for free"? Success, at the most attended Bitcoin conference in history, is still measured by price, he says, not by payments or businesses. He cannot tell what Bitcoin is for, and "just buy and hold" is not a philosophy.

The 2021 Miami Bitcoin conference made Hoskinson sad. He remembers San Jose in 2013 and Moe Levin's North American Bitcoin conference in Miami in January 2014, when nobody had money and you could strike up a conversation with anyone. There was a help-desk booth run by Dima and other OGs; core developers sat there and answered whatever a stranger asked. Miami had six-hour lines, superstars, and booths selling half-million-dollar diamond watches. Someone threw a roll of toilet paper labeled shitcoin. There was no help booth. He had started in the space with a free Udemy class, 80,000 students, and 5,000 emails before he stopped answering. Mike Hearn, who had emailed Satoshi and written a Java client so Bitcoin could reach phones, left in 2013 over the block-size fight and was treated, Hoskinson says, as subhuman.

The culture he still likes lives in Ergo, a project by Alex Chepurnoy, a former employee he describes as a brilliant Russian programmer with negative ego. Ergo is still proof of work and still UTXO, with Sigma-protocol contracts and an attempt at non-outsourceable puzzles so mining pools cannot form, closer to Satoshi's one-CPU-one-vote. Cardano used it as a prototype bench, the way Microsoft once bought Macs to build the Xbox 360 while Apple left PowerPC. Stablecoins and oracles were beta-tested on Ergo because it was the nearest extended-UTXO neighbor. Chepurnoy's Scorex framework, a modular teaching stack that separates consensus, networking, and transactions, also spawned Waves and a few other chains.

That accounting choice is the technical heart of Cardano. UTXO is cash-register accounting: you do not tear a twenty-dollar bill to pay $17.50, you spend the whole note and take change, every output atomic. Bitcoin never gave those outputs rich predicates, so you cannot say the coins move only if the lawn is mowed on Tuesday or the Broncos win. Ethereum's account model is a bank ledger, mutable and global, which makes expressive contracts natural and sharding painful. Extended UTXO keeps the register, then attaches data and a script. Settlement cares about its own register, not the whole bank. Local Plutus execution matches the chain. IOHK's *Chimeric Ledgers* paper argues the models are isomorphic: anything one can express can be mapped to the other with properties preserved. The code just looks different, as Java and Haskell do.

Hoskinson's line on oracles is deliberately crude: doing smart contracts without them is like having sex with your pants on. Two problems follow, aggregation and injection. Wolfram Alpha is an odd aggregator, a computable archive you can ask how many Florida shipwrecks between 1950 and 2000 lost more than a billion dollars of cargo and killed at least one person. The weakness is centralization; Stephen Wolfram likes to run the thing himself. Injection can be as naive as a signed public key or as involved as weighted feeds and a veracity score. Chainlink, born partly in academia around people such as Ari Juels, is one philosophy of how to federate that process over time.

He reaches for *The Simpsons*: Mr. Burns and Smithers thread vault doors, dogs, and robots to kill Springfield's power, then find a stray mutt and a wicker door to the outside. Infrastructure is as strong as its weakest link. Decentralize every computation and let someone tamper with the feed, and you will "perfectly execute the wrong answer." That is why some Layer 2 oracle stacks try to pull more of the contract off chain: if you already trust them, they will charge for the rest. Native stake pools have a built-in counter. They are already online, already capitalized, already reputationally exposed, and can sell data feeds, state channels, and randomness without an extra oracle token. If you make bagels, he says, you can probably make donuts.

Smart contracts themselves were, at the time of the taping, about ninety days away. Alonzo, named for Alonzo Church, was rolling through colored testnets, blue then white then purple, until a hard-fork combinator event would make them appear overnight, like waking up with a Twitter checkmark. Mary already allowed native assets; Alonzo adds programmability. What keeps him up is not only the base protocol but the apps. Blaming Vitalik Buterin for the DAO hack, he says, was like blaming Bill Gates because Photoshop crashed. A copy-paste contract from someone who has written three lines of code can still define the platform in public. First-to-market teams, he adds, will piss in the pool for everyone else.

Decentralized exchanges are the application he most wants built without a Coinbase-shaped custodian. Mt. Gox in 2013 showed the cost of that trust. A DEX still has to live with global latency instead of colocated nanoseconds, and regulation is the existential issue: value moves among anonymous accounts with no KYC. His candidate for the next generation is regulated DEXs that keep openness and add decentralized identifiers, so compliance does not require a single gatekeeper, at least if you still want an exit to dollars.

Hoskinson made a video for Jack Dorsey to say that proof of work is a resource, not a religion. The mechanics of electing a block producer, building, and checking are the same; the trade-offs differ. "Do you want mercenaries guarding Rome or do you want Roman citizens guarding Rome?" Romans, he says, learned that citizens are usually better. He would rather talk about *n* resources: Permacoin-style storage, useful computation, relays, and stake, balanced the way factions inside a government argue. A tweet that proof of work is "more secure" is not a model. Cardano published papers, formalized specs in Isabelle in some cases, and was running in production with a million users and a $50 billion market cap. Maximalists, in his telling, still answer with the word of Satoshi.

The case he would make to Tesla is energy and programmability. Proof of stake treats joules as a cost to minimize; proof of work turns every efficiency gain into more ASICs. If Tesla is an alternative-energy company, a system with no built-in brake on consumption fights its own story. Bitcoin is also the least programmable major chain. Autonomous vehicles that need to negotiate bandwidth, data, and payments want tokens, DEXs, and stablecoins, which means Ethereum or Cardano, not a settlement commodity.

Dogecoin, a copy of a copy of Bitcoin via Litecoin, was meant as parody and became a religion, like the Emperor of Man in Warhammer 40,000. Hoskinson's video on how to modernize it was half serious, half a demonstration of how hard protocol work actually is. Satoshi's design was the residue of thirty years of digital-cash research and still did very little; extending it is years of science and engineering, not a meme. The sharper issue is distribution. He says more than 90 percent of supply sits with less than 1 percent of holders, a handful of wallets that are in profit at almost any sale price. At fifty cents those holders are billionaires. The buyers are people with a few hundred dollars a month.

A waitress in Longmont, Colorado, asked what he did, then said she owned some Dogecoin and nothing else, because Elon Musk tweeted. When the music stops, those holders get hurt, and the Elizabeth Warrens of the world arrive to argue that the industry cannot govern itself. Hoskinson is the one who would have to put on a suit for a Senate hearing while the person who crashed the market laughs with friends. If the circus brings serious engineers and capital, he will count it as a net positive. He would like to give Musk the benefit of the doubt. He has not seen Mars for Dogecoin yet.

Every Layer 2, Hoskinson says, is the same sentence: Layer 1 is slow and expensive, so batch the work in a faster, cheaper system and give up some base-layer guarantees. Hydra's difference is co-design. If Lightning had been built with Bitcoin, the base chain would have had hooks for entering, leaving, availability, and fraud. Hydra is that counterfactual, using Cardano's accounting model. Payments are the easy case, a five-cent tip on a video. Running a DEX or a game off chain and reconciling later is a state-channel problem. The current paper is a single head; next comes routing among heads, then recovery when a head dies so funds are not locked forever.

Rollups batch transactions and attach a proof you can check against history, the same family as recursive SNARKs and Mina. Different boats for different cargo. Channels, because they have operators and interaction, can grow into interoperability. In principle you could modify Lightning's BOLT spec to talk to Hydra and bridge Bitcoin and Cardano. Wrapped assets would be cheap in that direction. The other way is blocked, he notes, because Bitcoin still cannot issue assets. Layer 2s in general should be specialized tissue, blockchain-agnostic add-ons for identity, cheap contracts, or payments, mixed the way firms compose cloud services. Putting identity on Cardano's base layer would invite China or the United States to dictate it.

NIPoPoWs, non-interactive proofs of proof of work, are a free exhaust property of mining. Some solutions are rarer than others, red blocks among green. Bookmark the rare ones and you get a short proof of a long history, useful for light wallets and for sidechains that share proof of work, plus log-space mining so a miner need not hold the full chain. Dionysios Zindros and Aggelos Kiayias wrote it in 2016; it took three or four years of rejections to publish. Zindros, then Kiayias's student, was doing a Stanford postdoc and, Hoskinson says, has spent his career on interoperability. "Real proud of that kid."

Asked to play devil's advocate against Cardano, he names three failures and never mentions price. The system might never self-evolve and might always need custodians, though the path from a federated chain through an incentivized testnet, stake-pool pioneers, Shelley, and a declining decentralization parameter suggests otherwise. It might evolve into the wrong shape: a small group deciding who may use the network and watching them, China's social credit as the sketch. Or network effect could fail, churn beating growth, though he doubts it once people tattoo the logo. What he fears more is the Bitcoin sin: toxicity, maximalism, orthodoxy. He wants a Socratic community that entertains ideas without adopting them and drops them when they are wrong. When Shelley shipped after four years, few people congratulated him. When ADA hit a dollar, they did. That, he says, is the most disheartening metric in the work.

Polkadot, in his competitive reading, copied some of Ouroboros because Gavin Wood, a trained computer scientist, read the papers. That is useful: Cardano can study parachains while it designs its own sidechains and Mithril. He thinks Polkadot did commercial adoption more carefully than Ethereum's spray-and-pray, with more predictable costs, and that Cardano's native multi-asset ledger, faster proof-of-stake research, and Catalyst governance are the other side of the trade. Historically he thinks Ethereum could have shipped something Polkadot-like by 2018 if it had funded Cornell's Snow White instead of waiting on a more aggressive Eth2. Relicensing C++ Ethereum Apache instead of GPL, he adds, might have made it IBM's enterprise chain; they built Fabric instead.

Fridman asks about Vitalik. Hoskinson is tired of the question. He was at Ethereum six months, has been building Cardano for years, has been to 52 countries, and still gets introduced as an Ethereum co-founder. He barely talks to Buterin. The dark side of the industry is labels, unconfirmed books, and cascades of reposts. He would rather be defined by what he builds, including an AGI collaboration with Ben Goertzel in Rwanda after Goertzel handed him an 85-page paper and a hat he never seems to remove.

Lincoln's line about testing character with power lands on a practical fear: not cartoon evil, but losing contact with people and harming them by accident. He now has a Wyoming ranch with 400 bison and no ranching experience. The vaccine is work that does not care who you are. A donkey, he says, will defecate on a billionaire and a beggar the same way. Hay does not care that you were meeting the president of El Salvador. The second habit is giving. He entered the industry with a free class. He gave away 293,000 ether, more than $1.2 billion at the all-time high, to a secretary who was catching transitive hatred and a collapsing marriage. "All the best things in my life have come as a consequence of first giving." After Ethereum he spent two years at bedrock, reputation damaged, phones unanswered, and rebuilt with people who would not have stayed for a narcissist.

He thinks the most likely Satoshi is Adam Back: right place, time, age, and skill, Forth-style script from 1980s European teaching, Hashcash as a predecessor, Microsoft in the résumé, code that looks like it was born on Windows and later cleaned by people such as Greg Maxwell. Not Hal Finney, whose Unix taste would not have smelled like an older academic choosing secp256k1. A U.S. Army paper on code stylometry claims about 94 percent accuracy; nobody, as far as he knows, has matched the original Bitcoin tree against contemporaneous public repos. Text stylometry, he recalls, pointed at Nick Szabo. Anonymity after a double-blind conference, he adds, is career poison unless the work is a whistleblow. His company inherited Qeditas, a proof marketplace from a pseudonymous "Bill White," and he regrets the silence because the work could have been published in the open.

Cardano's bet is explicit on-chain governance, not Bitcoin's implicit culture of BIPs. Some inflation pays block producers; some fills a decentralized treasury that already held more than a billion dollars of ADA, spent through Catalyst. Two metrics matter: how many holders vote, and how deeply they debate proposals on IdeaScale rather than dropping a ballot. If absolute and meaningful participation cross a threshold, he wants votes on parameters such as minimum fees and K, then on hard forks via the combinator, inside a constitution that distinguishes a new crypto primitive from a change in monetary policy. Ranked-choice voting is his preferred ballot because it kills the South Park "turd versus douche" trap; Arrow's theorem still applies, and no system is perfect. Brexit, he notes, might not have survived a two-round vote once the costs were visible.

What enters the funnel should not be what gets funded. Ideas go through a crucible of iteration, combination, and destruction; bad ones die on the vine, good ones leave with an auditor attached because a chain is not a company and someone real has to be accountable. SIP-007, on pledge, is the example that delighted him. A community member read a hundred pages of mathematical prose in the formal spec and proposed a simpler design that would be kinder to small operators. Knowledge had left the building. Protocol research is already decentralized by peer review, more than a hundred papers and growing without IOHK's permission. The remaining job is three independent clients and a specification, not a Haskell binary, as the canonical artifact.

Those tools are meant to travel. In Ethiopia, after four years of handshakes that were supposed to take six months, Atala PRISM is issuing decentralized identifiers for credentials with the Ministry of Education, starting at about five million students and possibly twenty million within two years. Graduates struggle to prove a diploma is real; a DID can attest GPA, a degree, or being over 21, then later payments and Cardano DeFi. The prime minister is himself a cryptographer in his forties, author of Digital Ethiopia 2025, and wants national digital identity and eventually e-voting. A government that uses Cardano, Hoskinson says, gets a parameterized government-in-a-box; a DeFi protocol gets a voting system it does not have to reimplement. Wyoming primaries with ranked choice and on-chain registration were on the 2022 wish list, riding a market for distrust after the Trump years.

El Salvador making Bitcoin legal tender might be a publicity stunt, or it might be a back door onto forex rails through reciprocal recognition, a cleverer version of an old joke about buying Tuvalu. Optics are not commitment. Adopting these systems pushes power to the edges, which autocrats may not actually want. The deeper pitch is that programmable finance can bind asset, identity, metadata, contract, and regulation into one library call, and that tokenomics can pay you today for work that helps people you will never meet in 2100. Capitalism, he says, is mis-parameterized toward the short term. Judge societies by the least well off, HDI not GDP. He entered the industry expecting jail: Liberty Dollar, E-gold, Gaddafi, prison, or becoming the corrupt banker you opposed. Still being free and building is already a win.

A productive day, if he can hold one, includes a wet year and a dry year so that nothing, including alcohol, is done too much; 2020 was dry and miserable. He tries 16:8 intermittent fasting, once pushed a fast to two weeks, fainted near the end, and called his physician brother about refeeding syndrome. Three or four days is the sweet spot before muscle loss. Float tanks, naps, and a Vielight photobiomodulation headset that shines red light into the brain, with some peer-reviewed claims of better blood flow and neurogenesis, are the rest of the kit. CEO life, he says, is the five-minute decision; he misses ten-hour deep work and has not yet kept the company Friday of no meetings that his chief of staff recommended. Live streams, 40,000 to 50,000 people, still feel like a gift. An Air Canada agent in Vancouver whispered that he loved ADA. A Heathrow passport officer welcomed "Mr. Hoskinson" before the passport came out, which felt, for a second, like an arrest.

Mushrooms are science and business. Cordyceps hijacks insects; lion's mane is studied for nerves, depression, Parkinson's, even COVID. Michael Pollan's *How to Change Your Mind* and Johns Hopkins psilocybin work suggest one or two treatments can match or beat SSRIs that patients take forever, and end-of-life distress after cancer is, for him, reason enough to study a cheap fungus you can grow almost anywhere. Farmbox Foods puts labs in shipping containers that yield about 400 pounds of gourmet mushrooms a week. At a Georgia festival he met Bill Yule, who looks like a druid and has spent decades on beetles that must vibrate a mating harmony on one species of bolete. After COVID emptied supermarket shelves, he wants $40–50 million aquaponics plants beside major cities so a community has some food if the long supply chain dies. The donkey, again, will still defecate on you.

The Cardano community wants him on Joe Rogan. Do not lead with protocols, he says. Rogan hunts elk, so start with elk tags on a chain; he has Spotify intellectual property, so talk about NFT royalties that keep paying a clean-water charity after the artist is dead. Once the guest is hooked, then eat the broccoli of proof of work versus stake. He bought the entire IP of *Legends of Valor*, the unfinished 1980s RPG that inspired *The Elder Scrolls*, as a blank slate for algorithmic music, GPT-class dialogue in the spirit of 2016's *Event[0]*, and hyperbolic geometry like *Hyperbolica*. His three favorite games remain *Baldur's Gate*, *Arcanum*, and *Planescape: Torment*, a thousand pages of dialogue in which you can talk the villain into suicide or never fight at all. Games, he thinks, are one of the last media that still get undivided attention; people beside him in a pre-COVID theater were on their phones.

Advice to the young: learn how to learn, with tools such as Zettelkasten; do not undervalue emotional intelligence and active listening; give early, because Gauss and Feynman had doctoral advisors; and do not fuse livelihood with vocation or you will burn out. Asked for the meaning of it all, he tells a Japanese story of a shogun who abdicated in his late forties to hunt a perfect cherry blossom, wandered thirty years, and was found dead under a tree holding a wilted flower. The point was never the blossom. It was the days, the people, the *ikigai*, Camus's Sisyphus finding joy in the roll. You are a blip. Find meaning in the daily act, and if misfortune comes, still find a little love to give away. His white whales are Goldbach, automated theorem proving with Lean, a cloned mammoth among the bison, and catching the same scarred catfish in his sixties. Then a heart attack, and the mammoths eat him.