与 Anatoly Yakovenko 讨论 Solana 的终局

I honestly think that as soon as a stablecoin bill passes Congress, you will see a massive, dramatic shift within 2 to 5 years, with us converting from current financial systems to crypto-based ones. Everyone knows this technology is better; it’s easier. All the APIs are easier, and we can start actually using these technologies. I think it’ll explode.

We’ve got a very special guest for everybody today: one of the founders of Solana. Why do we have a guest this week? Because Jonah managed to make every single holder of Solana extremely irate a few days ago by drunk-tweeting, in his words, why Solana is completely fucked. We wanted to have a conversation to figure out whether that’s true. Is Solana completely fucked?

But first, I want to figure out, Jonah, where did that tweet come from?

First of all, Toly, thank you for being such a mature, good steward of your ecosystem and responding to me in technical, thoughtful ways. Thank you for coming on the podcast to discuss this.

Let me give you a little bit of apologetic background. I was on vacation with my wife in Rome. The kids were somewhere else with their grandparents. I had one of those nights that you really don’t get a lot of when you’re a parent. I had a nice bottle of wine and was feeling pretty happy about my 24 hours of freedom.

My wife is pretty sick of listening to me talk about crypto, so after she passed out, I thought, “You know what? I don’t know if Solana is going to perform as well as ETH will in the next year. Let me talk to the internet about that.” No one cares what you say on the internet if you’re not a little bit extreme, so I could have said, “ETH is going to underperform a little bit if these things happen.” Instead, I went for it. I went aggressive, fell asleep, and woke up to discover that literally everyone who still cares about crypto wanted to fucking kill me.

Man, I’m a dad myself, and I wouldn’t spend 24 hours free from the kids without thinking about crypto. I was thinking about something else, yeah.

I appreciate that you had time to think about Solana. First of all, I hope everyone listening can be mature about these things. It’s just the internet. People should post.

If you ever actually meet those folks—even the worst shitposters in Crypto Twitter—they’re all nice people. You would normally be friends with them and grab beers or whatever, if that’s your thing. I hope people can step back whenever they get too emotional in these discussions. This is literally just people arguing about software.

I’m not even a guy who has any business arguing about software. I can write some crappy Python, and I think the most technical thing I’ve ever done was write the electronic music for this podcast.

Let me ask you a very simple question for the nontechnical folks out there about Solana. In a world where there’s probably a little too much block space relative to the demand for crypto use cases right now, what’s the vision behind creating the world’s fastest, cheapest blockchain? What use cases are you targeting?

This is a really good question because we don’t know at what point the market is saturated with block space. Are there use cases unlocked by sub-cent transactions that are worthwhile for anyone? There might be, and there might not.

There’s some indication of that if you look at IoT projects like Helium and Hivemapper. It would be really, really tough to do those on any other environment unless it was completely centralized and running on a database. You kind of need something like Solana for those. But could the world create those services in a completely centralized manner and be fine? Could nobody care, and could life just go on? That’s another question.

I think there is a general belief, similar to what motivated people building open-source software in the 1990s, that if we build these systems in a decentralized way, they’ll create more value in the long term for everyone else. You could argue that Windows has better features and is as good as Linux, and so on, but for some reason, in the 1990s I was motivated to spend my weekends working on open-source software because I thought it would be a major benefit to myself as an engineer, to the general public, and to all the other engineers I worked with.

There’s something to that. Similarly, if we can build decentralized systems, the hope is that they’re less likely to get co-opted and turned into the shitty version of the web that we have today. If Facebook were decentralized, if Google were decentralized, if all these things were decentralized from the start, they might not suck as much. You might not have your data stolen and sold all the time.

There are a bunch of benefits that come from that: diffusion of power, robustness, reliability. That’s the hope. Every person in crypto who’s building things has the option to go work at Facebook, Google, Netflix, or Microsoft, earn close to $1 million a year, and just not give a shit. For some reason, they’re choosing to do this.

Some may choose it for the upside, but a lot of people choose it because it’s more fun. You actually feel like you’re fighting the David-versus-Goliath battle. You get to push the hard technology forward. As a life and career goal, it’s more enjoyable to do that than to work in a giant Java database optimizing ads for Google, even though you’d probably make more money, have better work-life balance, and spend more time with your kids.

We’re here because we’re trying to do something interesting. In that sense, there’s a lot of faith in the idea that if we build the system—and what am I good at? I’m good at making things hyperoptimized and efficient—then if I apply all my talents to this and make a really hyper-efficient, low-latency, cheapest-cost blockchain, I think it’ll open up new use cases and could be very disruptive. That’s cool.

I don’t know what it’s going to be worth. If Solana is worth one-tenth of what it is today, we’ll just have smaller validators and features will take longer to ship because there will be fewer core developers. But there will still be people saying, “Holy shit, this is fun to work on. I’m going to keep going at it.” That’s the only difference. The rest is kind of up to the gods.

PMF is really, really hard to find and really hard to predict, so what happens in the future is tough to know.

That’s actually something I wanted to ask about. Is it important to you to build in ways that generate value for Solana? With ETH, they made the decision to introduce EIP-1559 specifically so that ETH could derive more value from activity on the chain. Is that something you view as important?

All proof-of-stake networks work the same way, no matter how the tokens move inside the black box. They all derive value in exactly the same way. There’s a resource that people want, and if people want it enough within a short enough period of time, they’ll try to access it.

Whether you have a burn or not doesn’t really matter. Even without the burn, people bid up the fees, validators earn more rewards, and owning ETH to get more stake and become a block producer more often earns higher rewards. It’s the same loop. It doesn’t matter whether it’s burned or not.

The deflationary meme is nice, and that’s cool, but EIP-1559 exists for a more technical reason: to prevent spam attacks in the ledger. The value accruing to the Ethereum network is exactly the same. It comes from people demanding access to block space and bidding for it to such heights that it becomes significant.

That’s harder to occur on Solana because we have localized fee markets. It’s slowly happening in Ethereum because L2s are separating state and each L2 has its own fee markets. You can think of an L2 as a generic blob that has its own fees. Those fees don’t really accrue to Ethereum outside of the L2 data fees, so you’re already seeing this happen in Ethereum.

The question everyone should have in mind at some point—I don’t know when—is whether Ethereum will also fix its block-space problem. Its block space is going to be much, much higher, probably to the point where the fees to access the data itself, just submitting bytes to the network, are commoditized.

There should be no more than maybe 2 or 3 times the cost of the hardware—maybe 5 times, but a small multiple of the cost of the hardware. Hardware is really, really cheap.

I remember in the early days of Solana, the hardware to run a validator was super expensive. What did it cost to run one box?

When we started, it was about $1,100 or $1,200 a month. Then it dropped to $800. Now it’s about $350. My guess is that within a year it’ll probably be below $200 a month.

Let’s say Firedancer ships and it’s a big success. What will it cost to run a validator?

Firedancer is far more efficient than the Labs client, so we’ll see how it actually performs. The problem is that when you get a box at a data center, you’re paying for the machine, the cores, the memory, the bandwidth, and the electricity.

The cost of the cores and that kind of thing—where code efficiency has the highest impact—is maybe one-quarter of the overall cost. Firedancer could be 10 times more efficient, but the price of that box might only drop by 20%. That’s it. You get some benefits, but not all the benefits.

It’s an Amdahl’s law kind of problem. When you have a bunch of things to optimize, you make one of them cost zero, and the other things become the dominant factors. You don’t see the same improvement across the overall system.

It would be awesome to see what kind of performance Firedancer gets on the same hardware once it’s out. I suspect it’s going to be much, much better, but we’ll see.

What I’m excited about with Firedancer coming out is obviously safety. We have two clients. Most of the top validators with a lot of stake have insignificant hardware costs, so they should be running both clients at the same time—one primary and one secondary. If there’s a bug in one, they can fail over to the other. That’s a massive improvement in reliability.

If they diverge on state, they halt. That means there’s a system-wide bug. Hopefully that never happens, but I’ll be able to sleep at night because we go from catastrophic safety bugs being a potential death of the network to merely being a liveness issue. Those still suck, but they’re recoverable.

Hopefully that never happens, but it’s a massive improvement. Once you have two clients, it’s easier to build a third and a fourth. There’s already a team working on a third one.

Firedancer is being built by a trading company. It’s being built by Jump. I’ve never worked at Jump, but I’ve worked at Goldman, I’ve worked at Vall, and I’ve worked at DRW. I know that trading companies are not altruistic entities, and I know how they operate. They’re pretty Machiavellian.

Let’s say Firedancer is a huge technical success. It decreases the cost of block space, everything you mentioned works smoothly, and suddenly machines all over the world are running Firedancer. It becomes the go-to validator client for Solana. How should users feel about the fact that the go-to software for validating blocks is controlled or invented by a company that clearly has an incentive to build this? They’re not doing it for free.

This is what’s cool: trading companies are actually mutually aligned with the idea of a giant, transparent state machine that competes with centralized exchanges.

When you connect to Nasdaq for high-performance access, you have to pay them an arm and a leg for every trade. Everything you submit—you can keep track of your own data, but they keep all the data, and then they charge you to get the data back. You pay them an arm and a leg to connect, and you pay them an arm and a leg to read the data. It’s available only to the top traders, and being part of that network is very expensive.

It’s opaque. It’s transparent to them, but this is an open system run by volunteers. Anyone in the world who pays for a hunk of metal can connect and be on the same level playing field. That’s great for Jump. They get more competition, but they get to compete in an open environment. Their connectivity is free. It’s just the hunk of metal, and the data is available to everyone.

They get to grind out a profit based on their algorithms, know-how, and so on without this additional man in the middle—Nasdaq or naisy—taking a rake. That’s a huge win for every trading company.

In a lot of ways, Jump is doing this because the vision is aligned with removing an extractive middleman in finance: the centralized entities that run markets. It would be better if they were all decentralized markets, because that’s an open space where trading companies can compete without paying astronomical fees.

What value does an exchange provide? It’s just a bunch of boxes somewhere that you connect to. Why do they get to make so much money for running a market? You have to ask yourself that.

Exchanges are very valuable businesses. They don’t take any risk, and Jump actually takes capital and risks it. They try to synthesize all the world’s information and predict the future. That’s really, really hard.

An exchange doesn’t do that. It just runs a market and says, “I’m going to take a trade from you and a trade from you, and I’m going to take a cut in the middle. You guys do all the work, and I get paid.”

If this works—if we can get a decentralized system to be bound by physics, remove all the bottlenecks, and have Solana globally distributed with information propagating at the speed of light—there would be no arbitrage between the state on Solana and the state on Nasdaq. That’s a really, really cool outcome.

That’s aligned with Jump because it will literally cut costs for them over the long term. It’s like Google investing in Linux because it sees Microsoft making a shitload of money on Windows and thinks, “If we invest in Linux, it’s a free operating system. It might take a chunk out of this big competitor that’s extracting a lot of value from us.”

I know this was a big narrative in 2021: Solana as the basis for a lot of financial transactions and for the places where an exchange could go build DeFi. DeFi might get really, really huge, and then you’d get the financial system rebuilt on Solana.

Realistically, after the FTX implosion—because FTX was a big supporter of that narrative—that sort of died a bit. On Solana, you started to see activity in NFTs and gaming, and people started paying attention. Maybe it wasn’t going to be used for finance, but it was going to be used for NFTs, which were blowing up at the time.

My question for you is: on the financial side, what has happened over the last year or 18 months that people might not be paying attention to, which is proving that you’re getting closer to everything you just described? Obviously, there were a lot of roadblocks.

The hardest part has been grinding for TVL and volume. Those are really, really tough businesses to run. We’re heads-down optimizing the network, but there are startups like Phoenix, MarginFi, Drift, Zeta, and others that are running systems, creating all sorts of trading environments, running markets, and grinding for PMF. They’re competing across the board with everyone else in the bear market for traders, and they’re slowly growing.

Look at Phoenix and Ellipsis Labs, the central limit order book that was built by a couple of engineers over the last 6 months. Its volume-to-TVL ratio is 30 times higher than Uniswap’s. It’s just insane how efficient they are.

That’s because it’s a highly optimized central limit order book. The compute costs are much lower, even compared with Serum and the old-school central limit order book. It’s easier for developers to use, and they’ve done a meticulous job optimizing it. They’ve been doing an amazing job.

The volumes are still small—we’re talking about $1 million to $5 million per day—but they’re not zero. They’re growing.

It’s really tough to compete with Ethereum there because of the way Ethereum DeFi works. People deposit a bunch of tokens that they don’t want to sell and that they have a very long position on, and then they collect yield on them. They just sit there. They’re not very capital-efficient; they’re optimizing for long-term preservation of exposure to that coin or asset.

If you’re a big ETH holder, you don’t really care how much it costs or how much your capital could be gaining somewhere else, because you’re not going to sell your tokens. You’re just going to keep them, deposit them in that protocol, and get whatever yield you can because you’re maximally aligned with that asset.

Ethereum was first and led DeFi Summer, generating so much market cap across the board. It’s really, really tough to compete with that by saying, “With one click, you should move your TVL from an A to an equivalent on Solana.”

I always wondered why more people don’t go down that path and try to make it really, really easy to port over activity.

Jupiter is working on some of that. It’s working on bringing in bridges and instant, on-demand bridge liquidity, so you don’t have to worry about where it’s coming from. You just click one button.

The pieces are there. People are starting to realize that this is a way to attract users and grow their businesses, and they’re adding those user-experience improvements. Everyone is fighting for this—all the L2s and so on. Everyone is trying to find a product that can attract traders.

The number of traders is kind of fixed right now. Maybe we’re not in the bear market anymore, but we’re not really in a full market either. The winds might be shifting, but this is a zero-sum game where everyone is player-versus-player competing for that activity.

Solana gets a bad reputation, Toly. No one wants to be like us.

I hear you on attracting TVL and attracting flow, but I go back and forth on this. My geopolitical view on crypto is that, in its current form, crypto is a fantastic solution for emerging markets. It provides dollarized checking accounts and banking. You can earn yield on your dollars if you live in a place like Argentina, Iran, or Venezuela. You can effectively have a U.S. dollar bank account.

However, I think a lot of developer activity in the West, at the cutting edge, is focused on applications that aren’t necessarily needed in the West. It would be an interesting thought exercise for people building financial applications on a chain like Solana to think about how they could create basic DeFi primitives that are beautiful and easy to use for people who actually need them.

I’m not sure the market is ready for a central limit order book for options trading.

I honestly think that as soon as a stablecoin bill passes Congress, you’ll see a massive, dramatic shift within 2 to 5 years, with us converting from current financial systems to crypto-based ones. Everyone knows this technology is better. It’s easier, and all the APIs are easier.

I don’t know if you’ve ever been a developer and tried to connect to PayPal or Stripe, but it’s a pound of flesh and a firstborn child to deal with the mess of setting up merchant payments. It’s much easier to generate an address with a private key and be good to go. That’s it.

Once we can actually start using these technologies, I think it’ll explode. You’ll have bank accounts taking USDC deposits across a bunch of networks, just like exchanges do, and people will forget about wires and all this other stuff.

From a technical perspective, is that a technical issue, or is it that they don’t want to make it easy?

The way the U.S. financial system works is built around compliance silos, AML, and things like that. Credit cards are the way you interact with users, but credit cards themselves are bizarre.

When I make a credit-card transaction, it’s equivalent to sharing my private key with the merchant. Then there’s this massive process to figure out whether the merchant stole the private key in the process and made an invalid transaction. It’s bizarre.

If you step back for a second, you have to ask, “Who built this?” There are a lot of problems with how this stuff works in the United States, and it’s all based on KYC and AML silos.

If and when we get a stablecoin bill and can start actually using these technologies, I think it’ll explode. You’ll have bank accounts taking USDC deposits across a bunch of networks, just like exchanges do, and people will forget about wires and all the other outdated systems.

I think Visa also sees the writing on the wall for that business model.

Providing credit and underwriting risk is real value. You have to do work and figure out how to manage that. That’s awesome.

I hope Visa thrives, creates massive success for everyone, and has an awesome way for me to get credit in USDC when I need it, in a very simple way, out of my wallet. That would be awesome.

We need this stuff to actually go mainstream, and it’s really, really hard for that to happen in the United States until there’s actual stablecoin legislation. It’s coming. They’ll do everything else before they do the right thing, but eventually they’ll do the right thing.

You’re playing the long game here, and I think I heard on another podcast that you’re talking to Visa a little bit.

The foundation has worked with them. Visa did a bunch of research across a bunch of different blockchains, and they like Solana’s performance and costs because they’re comparable to their infrastructure costs.

They’re exploring launching a cross-border remittance program using USDC on Solana. Remittance is an obvious win. You’re competing with SWIFT fees and a whole bunch of other nonsense involving 50 different banks to hop between any two points. With USDC, it’s such a direct win that it’s obvious everything should run on crypto for cross-border activity.

From their perspective, it’s a much better technology that can help them gain market share and prove this out as a win for the company. That’s really, really cool.

People should still be cautiously optimistic. We need to see these big companies get at least 2% to 3% of their revenue from crypto activity. I’m not talking about a crypto-native company like Coinbase. I mean a legitimate U.S. financial company, whether it’s PayPal or Visa, getting a few percentage points of its revenue from crypto activity.

That means the shareholders will care, the boards will care, and they’ll start investing more in it. That’s the snowball that will get us to mainstream adoption. We don’t have to get to 20% right out of the gate. A few percentage points would be significant.

The road to those few percentage points is important. You, as the core engineer of Solana, and the traders and investors who are long Solana tokens have very aligned incentives. The worst-case scenario is the same for both the traders and the developers: a network outage in the middle of this period of sunlight and progress.

The thing that keeps you up at night is probably the same thing that keeps Solana token holders up at night. What’s going to happen the next time the network gets pulled over, if ever?

How do you feel about last year’s issues being in the rearview mirror? Is Firedancer part of that? What makes you confident that the future is going to look better than the past?

The past issues were all centered around basically one design decision that we didn’t think hard enough about. It was obvious in retrospect. When we launched, this had to do with hotspots and very large demand spikes.

For people who don’t know, Solana is a parallel virtual machine called the SVM. It runs things in parallel, which is very cool and very Byzantine—meaning it can do many things at once. But there are other parts of the system that don’t do everything at once.

One part of that system is how, when you get a bunch of transactions and decide what to run in parallel, there’s no mempool in Solana, so the block producer or leader receives as many connections from different clients as possible and then receives a shitload of transactions. It tries to find the best transactions to run in parallel in the SVM.

Before we had priority fees and localized fee markets, when there was a hot NFT mint, bots would create as many machines as they could around the internet. They would all submit transactions, trying to be first in that NFT mint—so many transactions that it would exceed the physical limits of what these machines could process, sometimes reaching 20 or 40 gigabits per second of data.

We had to fix that problem. The first step was switching to QUIC, which is an internet standard built by Google and designed for this kind of low-latency communication with the ability to implement flow control.

The second problem was that even if you switched to QUIC and received a bunch of transactions, you didn’t want botters to create a shitload of machines and all submit data, because you’d still be dealing with a large pile of connections.

We came up with the idea of quality of service. If you have some stake, you’re guaranteed a certain amount of traffic. If you don’t have any stake, you get best-effort access. Every machine with stake has to reserve some of its bandwidth for non-staked nodes.

If you have an entire network of 2,200 Solana validators and each one allocates 10% of its bandwidth to non-staked nodes, you would have to attack all of them at the same time to prevent any non-staked transactions from going through. That’s a pretty cool solution.

Then, once you get all this data, you have to sort it. That’s where the localized fee markets come in. The way the SVM creates parallelism is that developers have to work a little harder in Solana. They have to tell the network which parts of the state they’re going to read and which parts they’re going to write.

You can think of these states as buckets. When you start creating transactions, you take the most expensive ones and start filling buckets based on the state they touch. As soon as one bucket is full, you delay all the transactions that touch that bucket and continue pulling the highest-paying transactions as long as they don’t touch the full bucket.

How do you decide when something is full? What does “full” mean in that context?

Even if you have parallelism, when everyone wants to mint the same NFT, they’re all touching the same state, and you can’t parallelize that. There’s nothing you can do. No zero-knowledge magic or L2 can solve it. It’s called a database hotspot: everybody wants the exact same thing.

There’s no solution to that, so the only thing you can do is limit it. There’s a certain amount of transactions that can touch the same state in a block. A block can contain many multiples of these hotspots—five different hotspots, four, or whatever.

If there’s a hot NFT mint, you fill that hotspot first with the highest-paying transactions that all want to touch the NFT mint. Then you’re done. Anything else that wants that NFT mint is delayed.

I guess my question is more out of pure curiosity. Is that concept of “full” orchestrated by the chain or by the validators?

It’s defined by the block. You can think of it as part of the EVM gas limits. We have single-thread limits and account-rate limits.

If you have a single account that you’re writing to—which, in this case, is the NFT account that defines the NFT mint—you can fill it up to around 12 million compute units. Even though the block has 48 million compute units, as soon as you hit 12 million, you can’t add anything else to that account.

You still have all this other block space for everything else. It’s rare that you have an NFT mint, a liquidation, and all this other activity in the same block, but it does happen.

What’s cool is that as the cost of hardware decreases every 2 years, we get twice as many cores for the same amount of money. That’s all we need to add more hotspots. If we’re using one core for the SVM hotspot, next year we can have two cores. Now we have twice as many hotspots. We go from 4 to 8 to 16, and at some point we’ll hit whatever demand exists.

Unless human demand for block space goes up exponentially—which I wish it would, because then we’d have a lot more value—I don’t think we’ll run out of block space.

I think it will. We’ve debated this on the podcast before. From the perspective of an oil trader, what happens is that oil gets discovered in a location and then you have to pipe it to refineries. Usually, when you discover oil, there are no pipes, so there’s this big race to build them.

The first person to build a pipe makes a bunch of money. Then pipes get overbuilt, and the price of pipe space drops to bargain-basement levels. All the pipeline companies go bankrupt, then more oil is discovered, and the pipeline companies resurrect themselves and make a bunch of money again.

In a sense, blockchains are like that because blockchains are the pipes. Some blockchains were built a while ago, when people didn’t really want to use them. Then suddenly 2021 happened, everybody wanted DeFi and NFTs, and there wasn’t enough block space. Now there’s too much block space.

In our opinion, it’s just going to keep oscillating back and forth. We think we’re primed for a huge cycle of block-space demand, which is why it’s interesting to talk to you. You’re building it. You’re literally laying the pipes.

That’s a hope, and we’ll see. Obviously I want that to happen too, but we need to know what those use cases are that will drive so much demand. You have things like Star Atlas, which is a game. It’s doing more transactions per day than an L1 or Polygon. It’s just one game.

The way they built it is to use the chain as much as possible for in-game events, so stuff like this could explode. But the question is: in the absence of block space, would Star Atlas still exist? Would the users care? If the answer is yes—if the users would still enjoy the game the same way—then it’s not being filled with traffic that’s critical.

It’s not like the oil is being used to drive trucks, grow food, or do something really important. We’re burning it to mine Bitcoin. There’s superfluous usage, and then there’s required usage. I don’t know whether there will be so much required usage that we run out of block space.

I view that a little differently. It’s almost like an intransigent-minority situation. As long as you have enough demand for block space, you’re going to have a lot of superfluous activity on that chain as well, because it will work better to do everything in one place.

That’s how I view the traditional systems we use. There are some things that are required to be built in a certain programming language, and then there are certain things that are done that way because other things are required to be built that way.

The goal is to increase the amount of activity that’s required in some capacity. But I think the reality is that the majority of it will be superfluous. It will be activity that isn’t necessarily required to be done on-chain but is done on-chain because you get easier access to the things that are required.

That’s what I’m hoping for and looking for. You get these things that move over eventually, and block space is created by computers. Computers get twice as cheap every 2 years, so block space is going to increase no matter what.

There are weird networks like Bitcoin where block space is fixed for religious reasons, but even Ethereum, after Danksharding, will improve to a point—maybe in 3 or 4 years—where it’s aligned with hardware costs. It’s inevitable.

We’re going into a future where we’ll have access to block space. I don’t know what we’re going to do with it. We’ll see what happens.

Let’s say Avi’s worst nightmare—or favorite dream, I’m not sure which—comes true, and you have a bunch of superfluous block space being consumed for no reason. Let’s say a bunch of people decide to spin up really basic computer games like Asteroids and Pong on Solana and arbitrarily consume a ton of block space.

Is the siloing fee-hotspot mechanism you described strong enough to withstand an onslaught of random crap like that?

Yes. We saw that live during a Helium IoT migration. They were minting 1 million NFTs. Mad Lads launched at the same time, which was a classic hotspot-based mint, and you had a bunch of random stuff and oracles. All that activity worked without a hitch.

It’s a matter of how many concurrent hotspots the network is currently configured to handle. It’s about 4 right now, which may sound like a small number, but when you’re talking about 400-millisecond intervals, having 4 things happen at the same time around the world is very unlikely.

Most people think in human time frames: “All of this happened within the same half-hour.” That’s eons for a computer. It’s not actually the same time.

We have enough capacity, and within 2 years, without doing anything, it’ll be 8 simply because hardware costs will drop and validators can add more cores. From my perspective, that’s just the relentless march of Moore’s law.

If that ever stops, we should all be building bunkers and stocking booze and guns, not worrying about crypto—or only worrying about BunkerCoin.

Now that I’ve been introduced to your community through Twitter and seen how much they care, I wanted to ask you about it. I hope folks listening from the Solana community will be nice to people. Kindness is super cheap.

I’m actually really impressed. I love your community. I had no idea they were this passionate and committed. How do you build a community like that?

You’re obviously passionate about the technology, but I’m not an engineer. I’m not in those rooms. How do you create this? So many L1s have tried and failed, but you’ve succeeded where basically everybody except Ethereum has failed.

This is really weird. I think I have some instincts that I learned from being an open-source developer in the 1990s and trying to find other like-minded people who were into hacking on Linux over the weekend.

You have to find your niche—your nerds. You have to invest in building something you’re genuinely interested in. I’m really interested in optimizing this thing, and I can talk your ear off about all the cool optimizations we can do.

People who are into that stuff can see that I’m genuinely interested in it. I’m not trying to sell them on it. I’m just a performance nerd. I think that’s the start of it.

You have to find your tribe. How big that tribe is going to be, who the hell knows, but you have to be genuine and passionate. Those are the key parts. I would prioritize those two things above anything else when trying to grow.

I think other projects fail when people try to do it in some artificial way. The earliest point when we really started building a community was when we needed the testnet up and running. It was a hard process.

We literally asked people to find a local data center. Not many people have done this. We asked them to get a box there, install it, set it up, and maintain it. That’s much higher friction than running a node on AWS.

The people we convinced to do that were obviously much stickier than people who ran incentivized-testnet nodes on AWS, earned their points, and then left. If you spent the trouble building a computer, geeking out over the specs, calling your local data center—which barely has a website—and figuring out that you could put a machine there, you were going to be committed.

That kind of weird friction adds stickiness and, in some cases, creates your tribe. You can’t really make it or create it artificially. There was a reason for that friction to exist: we needed cheap hardware and cheap internet connectivity, and there was no way to get that from the cloud.

People understood why they were doing it and saw the positive outcome. Our tour to Soul was very generous in terms of token distribution, and I think that created some stickiness too.

Your community has stuck with you through incredible social volatility. I don’t know if Shaquille O’Neal is into high-performance coding, but there’s been a long list of celebrities in and out of the ecosystem. It’s fascinating to watch. I’m mystified and amazed by it, frankly.

It’s a mystery to me too. I thought we were going to be dead more than once, but somehow people are sticking around and really seeing the value of what we’re doing. It’s awesome.

Speaking of building something awesome, if you had to guess what Solana’s product-market fit is going to look like in 1 to 5 years, is it the phone, DeFi, or something else? What do you think is going to be the next big smash hit?

I hope it’s something like Helium, where they have $5 data-plan IoT devices that are doing the whole cryptoeconomic loop. There’s demand here, and they give you more rewards the more boxes show up.

From a 50-foot view, it’s the perfect application of crypto. It’s creating value for consumers. Every member of Congress I talk to understands it. You tell them, “It’s a $5 data plan,” and they say, “That’s good for my constituents. That’s good for everyone.” I love it too.

Why the fuck hasn’t it caught on yet? What’s going on there? You’ve liked it for a long time. Have you used it? Is it usable in Miami?

It’s usable in Miami. You get $5 data, and it backs onto T-Mobile, so you can’t tell when it’s using the T-Mobile connection and when it’s using the Helium connection. They’re rolling it out, and I hope they can go national soon. We’ll see if they can take a bite out of a big national provider.

Hivemapper is another one, but it’s harder to think about. It’s mapping data that they’re collecting. Once you build the data set, you could build a Google Street View competitor, train cars like Tesla does, or train AI models. There are a bunch of cool products you can build on top of it, but it’s not as visceral as reducing the cost of mobile service.

I want to see more things like Helium. I don’t know what the next one is, but I really want to point to something my parents can use and tell them, “If you use this, you’re using Solana, and it’s cheaper or better than the alternative.”

With finance and banking, it’s going to take a while. It’s clearly better, but maybe within 5 years Congress passes a law, and within 5 years we have USDC-based wallet accounts with yields. They’re better than banking because you never get overdraft fees, you never get monthly fees, and you have all this transparency. That could be an easy one to point to, but that’s going to take law and finance.

My theory is that we’re going through the process of software eating the world. It’s a natural process, and it’s now starting to eat places where governments and humans have wanted control.

Finance is very connected to government, and there’s a lot of tension there. People don’t want to give up control over things, but it’s inevitable. Within 50 years, can you imagine banks still being run by 10,000 people trying to figure all this stuff out by hand? It’s crazy. There’s no way.

They’re fighting it pretty hard right now. To your point about finance, the fact that a financial transaction settles in 2 days—for typical T+2 business days, not even calendar days—for a high-value, Wall Street-type trade is incredible to me.

In the world of zero interest rates, who cares? But if your short-term interest rate is above 5%, trillions of dollars of capital are being held up all over the world for 2 days. If you could just let that capital flow, how much value could be created?

In theory, Solana could do that if it just replaced existing financial infrastructure. But, like you said, there are some people who don’t want that.

When you talk to members of Congress and regulators, especially in the wake of what happened with SBF making everyone upset about all of this, how do you cool the jets in Washington? How do you explain it to them?

A lot of the people involved in government are super smart, even if it may not seem like it. They’re very smart, but there’s also a lot of politics. Politics is populism. It’s trying to build a narrative, and sometimes it’s kind of stupid. People try to snipe each other, and that’s the theater of it.

Under the hood, very smart people are trying their best. They have to fight counternarratives, stay relevant, raise money, and compete in a system. I’m very bullish on it as a robust system that will eventually get the right people to do the right thing. I think it’s much better than any alternatives out there, but it takes time.

The most important narrative you can show people is that this is a product their voters will actually benefit from. Helium is my go-to example.

The financial examples are harder because the Wall Street Journal will publish a totally incorrect estimate of how much terrorist activity uses crypto, and then you have to explain that the estimate was wrong. You have to bring the counterpoints and explain why using a public, immutable ledger is one of the dumbest places to commit a crime.

They need those bullet points. Then there are people like Warren. I don’t know what’s driving her internally, and I’m genuinely curious to find out why she’s so anti-crypto.

Is it because her base of power is the financial committee, and she wants appointees who can maintain human control over banks? Or does she really believe there’s no value being created for consumers? You have to figure that out, and I don’t know the answer yet.

Most people are smart. Crypto is a technology, and they understand the benefits. You can explain those benefits, and eventually the right thing will happen. It just takes time.

Speaking of the right thing winning, I want to wrap up by asking you: why is Solana going to crush Ethereum in your mind? Why is Solana the right thing to win?

“Crush” is a very strong word. That part I don’t know.

Ethereum has had incredible success. It was first to market with smart contracts, and now its momentum is driving it toward a particular part of the Pareto-efficient curve of trade-offs for what a Layer 1 blockchain should look like.

Ethereum can’t really change itself into Solana. It would be really, really hard for Ethereum to rewrite everything and become Solana. The ecosystem, the code, and everything else are moving toward this multilayer model. That’s fine.

Because of the way my brain works and the passion that kicked off the Solana project, Solana is moving toward a different spot on that Pareto-efficient curve. We’re going to build a single Layer 1 that can host as many applications as possible, to the point where it should be more expensive for you to take an app and run it on your own app chain than to use Solana.

That’s it. It should be economically irrational to do that. People will still do it for other reasons, but there should be no economic reason to do it.

Which one wins? Who the hell knows. It’s up to the PMF gods or whatever. I don’t see a world where Ethereum loses and Solana wins solely because of that. I think you’re going to see ebbs and flows, with some applications being built on Solana, some on Ethereum, and activity shifting between them.

Maybe Solana will process 90% of the world’s transactions, while 90% of the TVL is on Ethereum. Who won? I don’t know. You can’t really tell what that means.

My core belief is that people are using the chain, building apps on it, creating things consumers want, and creating value. If we get that right, we have a shot at beating Ethereum—crushing it, whatever.

I asked it in the most inflammatory way possible, but I like the sentence I can now quote-tweet and tell everybody you said: “You’re economically irrational if you use Ethereum.”

No, I said it’s irrational to use an app chain instead of running an app on Solana.

It’s going to get more traction the other way.

I love it. Do it. You can say that.

Toly, I really appreciate you spending time with us today. I sat here and listened for most of it, but I feel like I learned a ton from hearing you speak. Jonah, you had some great questions in there too.

This has been a great conversation. Thank you for sharing such a well-articulated vision with us, Toly.

I learned a lot too. I’m kind of blown away. I’ve got to go back to the drawing board now.

Thank you guys for having me. We’re all working in open-source software, and we’re all kind of on the same side. All the details don’t really matter over the long term.