Multicoin Capital: How Technical Scalability Creates Social Scalability

Original Title: "Technical Scalability Creates Social Scalability", Author: Kyle Samani, Co-founder of Multicoin Capital, Compiled by: 0x33, Rhythm BlockBeats

Editor’s Note: This article assumes you are familiar with Nick Szabo's article "Social Scalability," Vitalik Buterin's paper "Weak Subjectivity," and Haseeb Qureshi's paper "Why Decentralization is Not as Important as You Think."

This article is not a rebuttal to Szabo's paper.

At the end of his paper, Szabo defines social scalability as "sacrificing computational efficiency and scalability—consuming cheaper computational resources—in order to reduce and better utilize the vast human resources expenditure required to maintain relationships among strangers involved in modern institutions such as markets, large corporations, and governments."

Bitcoin's triumph over BCH and BSV undoubtedly supports this theory.

However, since Szabo published his groundbreaking paper in February 2017, the crypto ecosystem has come a long way. Although Szabo conceived the idea of smart contracts about 20 years ago, it is only in the last 24 months that the world has begun to appreciate the most useful applications of smart contracts: DeFi.

DeFi changes everything.

This article assumes that the most important function of blockchain itself is not sovereign money, but rather DeFi.

Shifting the primary function of the system from strict censorship resistance and sovereign money to a powerful and programmable development environment for financial applications should impact every layer of the technology stack at the network layer (e.g., gossip vs. Turbine) through the execution environment (e.g., EVM vs. SeaLevel). (For a brief overview of the differences between Solana and Ethereum, see the appendix of this article.)

Therefore, blockchains should primarily be designed and managed as DeFi development platforms.

Predictability is Power

"I often get asked the question: 'What will change in the next ten years?' It’s a very interesting question; it’s also a very common question. However, I can’t remember the last time I was asked the question: 'What won’t change in the next ten years?' Here’s what I want to say: the second question is actually the more important of the two because you can build a business strategy around things that are stable over time. … In our retail business, we know customers want low prices, and I know they will want that 10 years from now. They want fast delivery; they want a lot of choices. It’s impossible to imagine that 10 years from now a customer will come out and say, 'Jeff, I love Amazon; I just wish the prices were a little higher,' or 'I love Amazon; I just wish you would ship a little slower.' That’s impossible. So we focus on these things, break them down, and we know that the energy we put into these things today will still pay off for our customers ten years from now. When you know something is true, you can afford to put a lot of energy into it, even in the long term." --- Jeff Bezos, Founder and CEO of Amazon

Coinbase has approximately 50 million registered users, and Robinhood does as well. Most major banks in the U.S. are similar.

We assume that Coinbase's strategic imperative is to migrate its 100% of users to DeFi as quickly as possible, and that the regulatory environment supports this. How can Coinbase achieve this on Ethereum today?

Currently, this is an unanswerable question. This is not to say that it is technically impossible; it is possible. However, there is no one or organization in the world today that can answer this question. Why?

Because no one really knows how Ethereum will scale. For example, Vitalik has stated that he believes optimistic rollups may be the best scaling solution in the near to medium term, while zk rollups will dominate in the long term. When and how will this transition occur? What infrastructure needs to be (re)built? How will the flow of funds between these different rollups work, and what will be the implications for smart contract developers, wallets, users, liquidity providers, cryptocurrencies, etc.?

Moreover, it seems that no matter how important the scaling solution ultimately becomes, it is unlikely to have a single holistic instantiation (e.g., meaning a single Optimism rollup). The future of Ethereum scaling will be heterogeneous.

In the long run, this may be a good thing for Ethereum. Each of the various scaling solutions has its own trade-offs, and it is unclear which trade-off set is optimal or what the best way to combine these scaling solutions is. Therefore, the best approach for the Ethereum ecosystem in the long run is to experiment with multiple scaling solutions, figure out which solution works best for which application, and then determine the bridges and other interoperability solutions as well as latency issues.

Additionally, all teams building scaling solutions are well-funded and have begun to go live and serve users. Therefore, none of the teams will give up quickly.

So how does Coinbase onboard over 50 million users to DeFi?

At such a scale, the most important thing to optimize is determinism. Companies of any size will demand determinism now and in the future.

Large organizations simply cannot bet on the wrong tech stack. The opportunity cost of getting it wrong and the explicit costs of later migration/bridging are enormous.

I believe the only blockchain protocol that can answer this question or will answer it in the next 24 months is Solana.

All rollup-based scaling solutions are plagued by the issues I outlined above. Sharding is too. Despite billions in research and development from a range of capable teams (Cosmos, Polkadot, Avalanche, etc.), no sharding system has actually achieved meaningful scale (most cannot even operate in daily output). Even if they work with PoC (proof-of-concept) consensus, there will be many new emerging issues that must be managed (e.g., cross-shard transaction failures, trading platform integration, etc.).

It is important to clarify that I am not saying that scaling through sharding and rollups cannot work. In fact, I am optimistic about both solutions ultimately. However, neither of these scaling strategies can truly work today and will create many secondary and tertiary issues that must be resolved. This makes it difficult to see a fair organization that desperately needs scalability and determinism being able to obtain the elements it requires in the next 24 months, as there are too many interwoven components to scale Ethereum.

The Costs of Fragmented Social Scalability

In addition to the uncertainties outlined above, distributing applications across various shards and rollups will create significant new social coordination costs that do not exist in a single shard system. Some examples:

Block times and computational throughput vary between Layer 1 and various Layer 1s. This directly relates to how risk should be managed across all DeFi protocols that manage risk (which includes almost all major primitives except Uniswap/Sushiswap). DeFi teams have committed to deploying contracts across multiple Layer 1s and Layer 2s. However, each execution environment will require unique risk parameters. This will increase the amount of social coordination required by each protocol community and slow down development.

Exiting from optimistic rollups (ORU) takes a long time. The market generally expects market makers to bridge liquidity between rollups and Layer 1. However, the implementation details of this operation are tricky. Should the protocol front end be provided natively? If so, should they "contract" with specific market makers (e.g., how Citadel Securities contracts with Robinhood for PFOF)? Or should the front end let users figure it out themselves? What if a user wants to move from one ORU to another… how does the user inform the application so that they can use Connext or Thorchain instead of exiting to Layer 1?

For Metamask users (who are mostly advanced users), it may be reasonable for users to manage this complexity themselves. But for curated wallets that try to abstract away this complexity (like Exodus or Argent), how much additional development time will the wallet teams spend solving these issues? How much actual new feature development will be sacrificed? What if market makers stop providing liquidity on that specific bridge/segment for some reason? What backup options are available? Developer tools must be updated to handle new data structures (unprocessed transactions of ORUs, zk outputs of ZKR). The indexing and querying layer will require significant upgrades, and application developers may need to rewrite their subgraphs to handle new data structures (e.g., it is impossible to map EVM subgraphs to Starkware's Cairo).

Developers will be forced to rewrite a significant amount of applications across various heterogeneous scaling solutions.

As the surge of sharding and rollups continues, DevEx will become more challenging. None of these issues are insurmountable, but they will slow down development and frustrate many developers who do not want to solve all these problems.

Predictable, Boring Scaling

Solana currently supports speeds of 50,000 transactions per second. The global network node count is about to exceed 600 as well. But most importantly, Solana provides a predictable path to infinite scalability. Because it can execute transactions in parallel on GPUs, it can leverage the massive gains in parallelism that GPUs provide.

Moore's Law may be the most important economic force of the past 50 years. But today, it looks more like an illusion.

About 10 to 15 years ago, Moore's Law stopped working for single-threaded computation. Because heat generation increases superlinearly with clock speed. This is why desktop and laptop devices have stagnated at around 3.5-4 GHz, while fanless devices (phones and tablets) have stagnated at 1.5-2.0 GHz. Although various optimizations over the past decade have improved single-threaded performance, single-threaded performance does not double every 18-24 months.

Over the past decade, almost all computational gains have come from chip specialization (FPGA and ASIC) and parallel architectures. Modern desktop graphics cards have over 4,000 cores. Over the past decade, the number of cores per card has been growing in a way that aligns with Moore's Law, and this trend will continue as the impact of the increased number of cores will not be the same as the impact of increasing clock speeds.

Solana is the only blockchain that runs SeaLevel, which natively supports parallelism within shards. SeaLevel executes transactions natively on GPUs. When Nvidia releases new GPUs with 8,000 cores in 12-24 months, the computational bandwidth of the Solana network will roughly double. Developers won’t know, won’t care, and won’t have to change a line of code.

This is the definition of predictability: writing code now, knowing it can be preserved indefinitely, and knowing the cost of executing that code tomorrow will be less than today.

The primary physical limitation to scaling computation is heat dissipation. Solana is literally scaling at the physical limits.

Quantifying Decentralization and Weakening Subjectivity

On the surface, many people believe the Solana protocol is not decentralized enough. They do not really quantify this claim, but they repeat it nonetheless. Let’s quantify what can be quantified.

1. Hardware Costs:

Bitcoin runs on a $25 Raspberry Pi with a weak internet connection.

Ethereum runs on a $500 laptop (although this claim is questionable given the current Gas limits) and requires a broadband connection.

Solana runs on a server costing around $3,500, assuming a gigabit connection.

The last major consideration is state size. With 50,000 TPS and billions of users, Solana's scale will skyrocket.

This is good, why? Because

1) Assume Solana runs on servers with upgradable storage (not non-upgradable laptops),

2) NVMe SSDs linearly scale read/write performance through RAID 0, and

3) Multi-TB NVMe SSDs cost less than $300. The costs and logistics of scaling state storage are negligible.

If your understanding of this article is so deep and you understand everything I have written so far, then your likelihood of using a Macbook Pro costing more than $2,000 is greater than 50%, which is also the high-end computer preferred by around 50 million to 100 million developers worldwide. I suspect optimizing for hardware costing $500-$1,000 is not optimal. What is special about the $500-$1,000 price point?

Considering the upper limit of hardware requirements should be fair. $25,000 is certainly too high, as developers do not own such hardware. Instead of stipulating arbitrary hardware costs, a better way to consider this issue is to think about how many nodes are sufficient to achieve adequate censorship resistance. Clearly, the term "adequate" is inherently subjective, but if you assume 1M is the correct number, then the natural question is: "Are there enough $3,500 servers with gigabit connections in the world to provide a reasonable path to 1M nodes?"

Given the number of gamers, developers, and businesses with high-end hardware in the world, it is hard to see that the answer to that question is negative.

The hardware cost issue cannot be considered in isolation. It must be considered within the design goals of the system.

Earlier, I believed that blockchains should be optimized for DeFi rather than for maximum censorship resistance (which means optimizing for 100M or 1B nodes instead of 1M). There is no need to optimize for hardware costing $25 or even $500, as the vast majority of people will never run a node. So why bother optimizing for hardware costs and protocols targeting them?

Everything is Subjective

This leads to weaker subjectivity and the realization that decentralization is not as important as you think.

The world is subjective. What does that mean? Consider the following example.

When you last walked into a skyscraper, did you first check the building and interview the general contractor to ensure the building wouldn’t collapse and kill you?

Or did you apply the same concerns to an airplane, a car, or your house?

Everything in the world is based on some degree of trust. If everyone had to independently verify the structural integrity of everything they interacted with, the world would not function.

Conversely, this can be stated as: the world works because everyone knows that enough other people have already verified the assumptions of the system, thereby guaranteeing the system's safety with a very high probability.

This is the fundamental assumption of weaker subjectivity. When applied to node count, the key question is whether you can reasonably assume that there are enough other individuals and institutions running nodes that you can trust the system, even if you do not run your own node?

Today, Solana has about 600 nodes, up from about 100 a year ago. Like almost all blockchains, this number grows over time. As long as the ecosystem continues to grow, there is no reason to believe that the node count will decrease. Like all other major blockchains, the Solana network naturally decentralizes over time as more people use it and more value flows through it.

This is also why Qureshi is correct, and decentralization is not as important as you think. Decentralization is crucial for achieving censorship resistance. It is currently unclear where that threshold lies (there are not enough counterfactuals from which we can reason), but the actual threshold itself is not important. What matters is:

1) The shape of the curve is actually an inverse S-curve, and

2) Knowing that blockchains become more decentralized over time, thus increasing their resistance to censorship. As long as the chain decentralizes at a sufficient speed (and you believe it can maintain that speed of decentralization), then users are likely to maintain the censorship resistance they require.

Image Source: Haseeb Qureshi, Why Decentralization is Not as Important as You Think

Conclusion

The fundamental question at hand is: what is the dog, and what is the tail?

In the first decade since Bitcoin's inception in 2009, it was clear that censorship resistance and sovereign money were the dog, and other applications were the tail.

But that is changing. Now for everyone using cryptocurrency, DeFi will completely reshape finance. I believe the roles have reversed: DeFi is the dog, and sovereign money is the tail.

Both require some degree of censorship resistance, but in an engineering-limited context, there is a fundamental trade-off between maximizing the utility of DeFi and maximizing the system's censorship resistance.

When the fundamental design goals of the system change, the technology stack should change accordingly. To scale DeFi to billions of people, every layer of the stack must be reconsidered from first principles.

Thanks to Hasu for reviewing the draft of this article.