jakiro
Interviewed by: Andre Cronje, translation: loxia.eth
This is an exhaustive interview about Andre Cronje, totaling 1 h 20 min.
It contains Andre Cronje’s summary of his past career, as well as a lot of experience, guiding advice, and personal ideological judgments.
As OG in many fields in the industry, Andre Cronje is the most recent and in-depth interview, worth reading and reference.
The article has a total of about 20,000 words, and is divided into 9 parts.
Andre Cronje believes it is stupid to regard any blockchain project as an “Ethereum killer”.Even if the total value lock (TVL) of all blockchain networks is added together, including Bitcoin and Ethereum, they are only a trivial part of the entire financial world.If you think a blockchain project can solve global financial problems, it’s crazy.
1.Introduction
Host 1:Hello everyone, welcome to the “Bright Speech” program!Today we have the honor to invite Andre Cronje, the founder of Yearn Finance, Phantom and Keeper Network and one of the key contributors to many DeFi projects.Andre, welcome to our show!
Andre Cronje:Thanks.Well, that introduction is a bit exaggerated, I’m just a person who likes to write code.
Host 1:When I listened to the 2020 “Extraordinary Core” show, they said you are a builder.I am not a developer myself, but I think you have a low opinion of yourself.You have a very interesting story and we can learn a lot from it.Maybe we can start in 2017, when you first entered this field, which happened to be the ICO era.So it’s great to hear how you got into this field and hope to explain to those who weren’t there at the time how crazy the era was.
2. ICO era: Andre’s cryptocurrency journey
Andre Cronje:Yes, I mean, before I got into cryptocurrency, I was a very traditional cryptocurrency skeptic.I come from a traditional finance background and used to be an architect and chief technology officer at a small financial company.We did some high throughput things, and we were using Kafka and Scala at the time.This is my background on high-throughput financial solutions.
That 2017 era was very similar to the present, and it was like this in many ways, because there was too much noise and there were many teams claiming to solve some industry-wide problems, and traditional finance and traditional distributed systems had struggled for decades.But these 18-20-year-olds, without any work experience, released an ICO and raised 20M or 40M, claiming they solved distributed systems or something.
So I initially entered this field just to test my suspicion and make sure I didn’t miss anything, you know, creating a disruptive technology and replacing the previous ones, this kind of thing is not the first time it’s going to happenof.My concern is that the blockchain field lacks a lot of research evidence, and while there is a lot of strong evidence, many people claim to have done something.So I entered this field and started reading the white paper. The white paper once again proved that a lot of proofs seemed reasonable, but now another problem, today there is, there are indeed many proofs that sound good, you would say, “There is,”The truth can be done.” But after it is implemented, when you practice it, there will be some hard restrictions that will not allow it to operate the way you expect it to be.
Even if the theory is correct, or even if the concept is correct, emmmmm (probably not necessarily feasible), so after reading a lot of white papers, I started looking at a lot of code and I started doing my original code review.I’m not doing these code reviews from any value creation perspective or any due diligence perspective, it’s totally like I read this white paper and it says it solves X problem, and then I go to the code and I’m wondering if it’s going to be fine if it’s going to beSolved the X problem, it is more like a record of oneself.
So you know, when I wrote down these processes on Medium, I just wrote it down and said, OK, this code doesn’t match what they said here, this code base has nothing to do with their statement.I made these public for some reason and in the ICO era they became very popular because there weren’t too many swearers and not too many people saying, “This doesn’t work because yoursThe code proves that you don’t have what you said.”There was a problem at this time, and it was very important that the reason why I finally stopped performing code reviews was that people began to use them as investment signals, rather than any code basis (research) because I shared it with others to learn and go onThe same learning journey I’ve tried to go through.
So, I did my own public review and then ended up working with a company called Crypt Briefing, working with Hana and John and those guys, they were still great, I still stayed in touch with them today and started doing some reviews for them, but then it got a little bit more, mixed with a shift that I didn’t like, that’s that I like to review public code, so you know, if it’s on GitHub, I can see that code while everyoneAll of them can see the code so that people can also verify if what I’m saying is true or tell me if there is anything wrong.
But as that influence grows, more and more teams want us to review their private code and then post review results, which makes me uncomfortable because it’s just a sign of investment, but anyway, that’sA parallel (another thing), we (look for another time) can go into it in depth, but after all this, you know that 99.9% is rubbish, but there is still 1% of real value, and this noise ratio is obviouslyIt is extremely high, but that 1% has always bothered me and attracted me.
So looking back at the time, my focus went from trying to understand what was going on to catch up with the industry, I think I did it in about two years, I think about around 2019, maybe earlierOne thing, maybe at the end of 2018, I think I managed to catch up.It’s very difficult to catch up in this field, I mean, there are a lot of new things coming up every day, you have to read 98% of the other published things to know what actually happened, but the actual amount of real things actually happened is very small, only taking up only the1% to 2%.
At that time, one thing I started to focus on was that the POW (Proof of Work) at the time was obviously a bottleneck.Looking at the blockchain system, you will feel that, well, the speed is obviously limited.The longest chain rule standard for Bitcoin at that time was that transactions took between 10 and 30 minutes.Until then, I was fascinated by cross-border payments, cross-border settlements and instant online payments.
I am South African, and South Africa doesn’t even join SWIFT or IBAN (the organization that provides global financial services) etc. We are restricted by foreign exchange controls and also by online consumption.Our banking system is very limited and this has been a challenge.It really appealed to me to see this freedom not under a single entity.
Therefore, I began to focus on consensus research.During that time, the research and code reviews I did began to guide me through Fantom and the team there and began to get involved more deeply.The market was fanatical when they were raising funds, and they managed to raise about $40 million in funding, through ETH!It is worth mentioning that they kept these ETHs, even during the bear market, I remember that they ended up selling when the Ether price reached about $300.However, they made a lot of promises that sounded good but they couldn’t actually do it.They seemed to realize this, but did not choose to quit actively, such as spending money or doing something to consume the money.Eventually they asked me if I could use the research I started publishing and I was thinking about whether to start my own chain, which fits well because I have no experience in interacting with venture capitalists and no experience raising funds or any other related experience.This is not my specialty, it is a skill, and I don’t have this skill.
You know, that’s why I’ve never started anything, whether it’s Yearn, Keeper or anything I’ve started, there’s no investment in VC or whatever, and a lot of people think it’s something I’m doingThe statement about work ethics, it’s not the case, I’m just not good at this, so I came up with ways to circumvent it, and that’s all.
So in the end, they had the funds, they had a branded team, so in the end I pushed my research into it, and the first thing was consensus, the original consensus was ABFT (Asynchronous Byzantine Fault Tolerance), which they called Lachesis, but it is actually based on a paper from the early 1990s, Common Concurrent Knowledge, which is really just an ABFT point-to-point communication system.We initially launched in late 2019 or early 2020.Consensus itself is great, you know, it was one of the first ABFT solutions, and it jumped from the transaction speed of the maximum 7 TPS at that time, when we didn’t have the virtual machine yet, we were just doing the originalTransaction connection, because that’s just a pure payment network, we can easily touch pure payment frequency between 30,000 and 50,000 depending on the validator’s connectivity or participation network.
But we wanted to allow virtual machines because smart contracts were powerful, and at the time we chose EVM, that was our only really viable option, we had considered using WM, we had considered using risk-based compilers, etc.At that time and even now, you know that to make a blockchain really feasible and even adaptable to use, you need a lot of service providers on top of you, and it’s difficult for people to do anything on the base chain because everyoneIt’s said, OK, we’re just doing EVM, people are just forking EVM, so we’re just sticking with EVM, and then we connect our consensus as the base layer, because consensus is just a sorting system, that’s all it is,It accepts transactions, sorts them, and then these transactions are handed over to the virtual machine and executed as state.
Then we noticed that our TPS would drop to between 180 and 200, which was purely the limit of EVM, and then for the next three years we were purely working on improving EVM, and we made some progress,But I have to say that if I could go back and change that decision, I would definitely do it.
I think we chose the easiest way back then, and chose the EVM route, so we knew it would be easier to integrate with all these third-party vendors, and that was a positive option because we didn’t have the ability to build our own wallets, build ourselvesRPC node providers, their own instant deployments, etc., but anyway, this is also a topic that we can discuss in depth later.
3. Create Yearn Finance
Andre Cronje:In the previously mentioned topic, they raised $40 million and left the funds all on ETH.However, when it was finally converted to USD, there was only about $2.5 million left.I want to talk about this, and that’s the operating capital of our entire team.To manage this fund, I began to look at many lending agreements that existed at that time, such as Compound, BZX, Full Crim, and so on.All other protocols except Compound have disappeared.I check these protocols every day, remembering that Ethereum’s handling fee was only three to six cents at the time, so it was possible to operate every day.I go to check these sites every morning to see which offers the highest annualized rate of return (APY) and then manually transfer funds between these protocols.Over time, I realized it was annoying to check these sites every day, they should have smart contracts on the chain that show interest rates, I can collect all the data and show it out.
The first smart contract I wrote and deployed on Ethereum is just an annualized rate of return (APY) aggregator.It can get data from all these different places and display it.The reason I did this is because at the time I couldn’t figure out the RPC infrastructure, like Web3 JS or anything else related, in order to get data from the node and perform operations.So, for me, the easier way is to deploy on the chain and read from there.
So I started my Solidity development journey.With this smart contract, at least I can check which interest rate is the highest every morning and then transfer funds.Then I realized, hey, I could actually write a smart contract to do this on my behalf.This is where Yearn originated.Later, it became more intelligent, and its current state is simply rocket science compared to the code I wrote.But, that’s the basis.What I want is to automate what I do manually every day until it can manage the funds I manage.I ended up opening it up so that others can use the same system.I no longer need to click the button every morning to redistribute funds between different protocols because whenever someone interacts with it, whether it is a deposit or withdrawal, it redistributes funds.This ultimately automates the whole process, and that’s where Yearn origins.
However, with Yearn’s development, the launch of tokens did not go as planned.The issuance of tokens is not fair.I’m just sarcasticating these worthless tokens.As long as you provide liquidity, I will send out the garbage for nothing.This seems to be the stupidest thing in my mind, but obviously I was wrong.However, this attracted a lot of attention, people started to join in, things started to become more complicated, involving strategic investments, infrastructure and so on.
As the strategy deepens, we spend a lot of effort to harvest.We are like any agreement to dump tokens.This has also become a thing.I used to manually operate these scripts to do this.So I said there is definitely a way to do this in a public space where anyone can call it and they will have the motivation to call it.This is the emergence of missions and keepers.Eventually, this evolved into a keeper network, which worked well for Yearn.So we decided to open this up so that anyone can connect to a task and then have keepers to perform.I don’t know who these keepers are, but they will do the job.The first task I launched on the chain was really fascinating because we didn’t advertise, didn’t post anything, we just activated the task and the bot started calling it.It’s really messy to see these things happening on the chain, which is probably what used to be called the Dark Forest, now I guess it’s just Mev Forest.
4. Errors and testing in production
Andre Cronje:Then, there are a lot of mistakes…the lack of better words to describe.Before Yearn, someone paid attention to me in this field, but I had no public reputation, reputation or eyeball attention, so I developed a lot of bad development habits.For example, I often do testing in production, which means I’ll put some experimental stuff into the system that actually runs, and you know I do.Another example is that intention and direction are completely disconnected.Because hybrid testing and production are places where I think there are great risks.It’s like telling someone, “Hey, I’m doing testing in production, and if you interact with this, it’s not what you should do because the chances of something going wrong are very high.” I’m saying this to warn you if you’re inWhen interacting here, you need to understand that there are extremely high risks here.
Testing in production ended up becoming a somewhat hasty, casual approach to investing money, even though it wasn’t my intention.Anyway, I’m still using my past development practices and I’m still building Eminence.I was very upset with the NFT culture at the time and I think it has improved now, but people were very stupid about how NFT was used at the time.They turned a painting into an NFT and hung it for sale for $100k.I like the philosophy of NFT because I am an avid gamer.I think they are the perfect application cases for NFT.So I got an IP license from Eminence, which is from another gaming company.We plan to build some stupid games to show how NFT works.I think NFT’s IP will always be problematic because it can’t exist in just one game.The entire plan is to build a series of different games, all using the same base layer.
But anyway, I deployed a bunch of tests, people interacted with, and had serious vulnerabilities and lost about $60M.I took a big step back because then I realized how dangerous this area is actually, how easy things are to go wrong very quickly, without the right safeguards and so on.Meanwhile, I was also facing considerable pressure from many regulators at the time due to Yearn, who characterize it as a financial instrument, which I think is fair, but I also want to keep a little distance from that.I ended up coming back firmly because one thing has bothered me for a long time, and that is how to improve the AMM curve.At that time, you know there was only one standard stable trading curve, that was Curve Finance, founded by Mitch, and was definitely a talented developer, founder, architect, and maybe I still think he was in the field I knew.One of the smartest people.But I was obsessed with it and I wanted to make something as simple as Uniswap, like X Y K.As a result, I finally designed the entire X to 3 power y plus y to 3 power x, which works very well, you can define this curve, it is very simple.
Meanwhile, I also added a bunch of them when you had TWAP (i.e. time-weighted price) and I also added RWAP (reserve weighted price).Because I don’t even need to explain about how these XY pools work, you just need to know that for TWAP it’s a fixed time price point completely ignoring the amount of liquidity, it’s saying, hey, you can sell a billion of this stuff at this fixed price, which is a big problem for me.
Note: Time-weighted average price (TWAP) and reserve-weighted average price (RWAP) algorithms use different methods to calculate asset prices, which are part of almost all DeFi fundamentals
Because many clearing robots, clearing engines, lending, and even fully decentralized stablecoins, etc. need to understand that slippage is part of the calculation.Let’s take the liquidation robot as an example, which works simply: it requires checking if I can pay someone’s debt, get his collateral for a million ETH, and then sell it in a Uniswap pool, and still be able to make a profit.If I use TWAP, my bot will say, no problem, the profit is good and can be executed.But if I actually slipped a lot after selling, I would suffer losses.So what I need is a way to take liquidity into account so that I can check it out realistically, and it is a specific time weighting so you know there is no large amount of lightning loans put into liquidity right now.I can sell it, but at the same time, it’s also a chance to snatch my robot.
So I need to go back in time to check, everything is there, and then build that method.There was also some confusion when it was launched on Fantom, as I left a week or two later.But apart from Fantom, I always feel that this is what the founders of decentralized protocols should do.If your protocol is completely immutable, without any updates, without any changes, you need to leave because you can’t be the top figure associated with that thing.I think Yearn and keeper do a good job because they manage them in a very decentralized way.You can’t really be sure who it is for both agreements, although it’s certainly a huge mess on Fantom.It has become one of the main AMMs of many new VM exchanges, like Velodrome Aerodrome and so on, and there are many more I don’t know about.
So, it achieves what I want, although not the iteration I did.After that, I decided that my development day was over, my smart contract day was over, and I didn’t have the necessary infrastructure, so I returned to Fantom full time.Sorry, this is a very long history and I have been delaying here for a while.
5. Fantom L1: Make the software as efficient as possible
Andre Cronje:I think the database is definitely useful, I think FVM is the best standard right now, and I don’t think there is a better virtual machine than this at the moment.From a data structure perspective, I think this is the case, because with Karma and the new database, we went through some regular processes.We initially used Badger, and then we did a lot of research on various different databases and switched to Pble, which gave us a good improvement in throughput, but not much change.There is a problem with all these existing databases, that is, they are designed for general-purpose data and can store anything in any way.Meanwhile, if you use the structured query language SQL at the top level, it means a lot of things will happen in the background.They are building their own indexes, creating their own P-tree, etc., which add a lot of extra overhead.
So even when you switch to key index storage, you might think that for EVM or virtual machine-based data, or we can say smart contract X data, then you’ll think that this will increase throughput, and in fact, it does, but again, there is a lot of work to support their query languages, such as GraphQL, SQL, or other possible languages.Actually, when we switched from the initial standard database structure to Pble and then to key-value storage, our throughput increased quite a bit, and now we are simply using a flat file on disk without anySomething complicated.
Our search mode is very simple because like any smart contract, it is an address, which is the first part of your index, and the rest is what address slot the data you are looking for is stored in, which is really just one or twoThree four five six, so if I want to look up, for example the first one is the name of the smart contract, it is actually address one, I have it, I don’t need more, I don’t need some complicated language.This is actually fair to say now that the limitations of EVM have been addressed, because you know that the MPT data structures used with existing index builds and other things are very data intensive.
I mean you have your actual data structure leaf nodes, then you have this complex, then there is a complex until you finally get to your top-level leaf node, so there is a lot of heavy work to do in that data structure, and most of them are hashes.This also means that it will be very dense every time you do a read and write.I mean, for our virtual machine, if we only look at Carmen, i.e. data storage, its capacity peak increases by 8.2 times, and only this increases throughput by 820, and there are many other incremental changes, but youKnow, I think that is a major event in itself.
One thing I’ve been promoting is that I think a lot of blockchain, a lot of development teams have accepted the limitations of the current situation, like according to some fixed physical rule, if you ask the people of Bitcoin, they’ll say POW isThe fastest consensus mechanism in the world.Even though I’ve interrupted you, sorry, I’ll shut up.
Host 2:No, that actually fits well with Solana’s view of the world, and if you look at Kevin Bowers, who is the new client head for Fire Dancer, all his work is to make the software as efficient as possible.Because as you said, there are a lot of weird abstract concepts that lead to a lot of weird performance problems, and then these problems keep getting worse.I mean, they even did something like a speedup hash algorithm.I think according to you, there is indeed a lot of benefits that can be obtained by first scaling vertically and optimizing the software to take advantage of physics.Then add all this complexity.I’m not going to go on, I’ll be talking about it, but I think, so what I want to do is actually for people who are not familiar with Fantom, we have a lot of developers, investors, researchers, or maybeFor those who are more familiar with Solana and Ethereum itself, please briefly explain how Fantom works at a high level and what makes it different?Then what are some obvious points?
Andre Cronje:The first priority we focused on was consensus.At the time, Proof of Work (PoW) was the mainstream system, and while this was just an analogy, I didn’t like people calling blockchains like Fantom PoS.PoS is just an anti-fraud mechanism, not a consensus mechanism, while proof of work is a combination of consensus and anti-fraud mechanism.The core concept of consensus is shared synchronous knowledge, that is, all participants agree on an event and know that others know this too.
For example, suppose I’m wearing headphones now, I tell you I’m wearing headphones, and I’ll prove to you that I’m wearing headphones.So now you know I’m wearing headphones, and I know I’m wearing headphones.And then you tell Garrett, and say, “Hey, Andre wears headphones.” Andre confirms to me that he wears headphones, and now Garrett knows I know he wears headphones, you know, Garrett knows I know.Although I didn’t know that Garrett knew that I knew, through third-party confirmation, we reached a consensus and the participants of the entire network knew about the incident.This is Fantom’s consensus mechanism, because we want all validators in the network to communicate with each other, they send ping packets to check if each other is online, and they send transactions they know but others don’t know to keep in sync.Therefore, communication is continuous and we try to use this peer-to-peer communication to achieve consensus by sharing messages and sharing knowledge of messages.
In our network, messages spread as quickly as viruses.It starts slow from one node to two nodes, but over time, exponential growth means it spreads rapidly throughout the network.We use a DAG (directed acyclic graph) structure that does not produce blocks like traditional blockchains, but rather reaches consensus based solely on communication.We divide time into so-called Epochs, and when 2/3 nodes in the network reach a consensus, a new Epoch will begin.This does mean we are very dependent on a peer-to-peer communication network and we are working to improve and optimize it to achieve faster information dissemination and consensus reach.
In blockchain, the spread of information is the key.Just like you tell me a message, I tell others, they tell more people.As more and more people know this message, just like on the Fantom blockchain, this message will eventually be known to enough people to form a chain.We have introduced a concept called Epochs for smart contracts (EVMs) on blockchain.Whenever 2/3 of the people on the network know something, we call it an Epoch.While this may sound a bit strange, it actually just means that more than half of the people have reached a consensus.Then, this Epoch is like a block, and we hand it over to the EVM to handle.Technically speaking, we don’t even have real blocks, just that communication is constantly going on and consensus is constantly forming.
This means we rely heavily on the topology of the network.We have identified many areas in the P2P layer, which is our next focus.Communication plays a key role in this process, and even today, communication latency is not a big problem, even on a global scale.So especially if you use a broadcast protocol, it means not a one-to-one communication, but I send the message to everyone I know, so it will be a little slower to process the information, but in such a network you can stillShare information very quickly.
Now, let’s talk about the consensus layer.Initially, we just did a simple transfer, from one wallet to another, without too many fancy stuff, without virtual machines.But then we wanted to introduce virtual machines, so we had Epochs.These two separate components work together, and the virtual machine handles and updates the state.Despite our optimizations, the throughput is still limited.Our maximum throughput in the original transaction network is about 50k to 180k, depending on the hardware and network limitations.Our goal at the time was to test the real limit, but we could further expand throughput throughput through hardware and technical means.
Our research is now focused on the virtual machine field, especially in consensus, we have received some peer reviews and have received a lot of help from the University of Sydney to build a close relationship with Professor Bernard Schultz.He is a truly great person, and while some people may make you feel a little bit less clever enough, it is not intentional.They are just on different levels, which can be frustrating at times, but it is also a great opportunity to learn, and I have learned a lot from it.He is one of the pioneers in programming languages and virtual machines, bringing many wonderful ideas and a complete team.
Kman and TSA are his masterpieces and I know I can keep up with them, but I don’t have ownership of these things.So let me briefly explain that TSA is a new virtual machine.Because we have DApp developers and an ecosystem, we really have to take them into consideration.The choice we face is either to start over, which means to give up everything before, including developers and the community, and rewrite everything; or to find a compromise that meets their needs.We chose to preserve the compatibility at the bytecode level.Simply put, the previous code can still run in the new system.This way, even if we do a network fork, the previously deployed contracts will still work properly.While this is still an internal discussion, we believe that at some point decisions need to be made to reduce the cost of technology.Currently, the TSA virtual machine is compatible with the bytecode of the EVM, which means you don’t need to recompile content like Solidity contracts, but of course you can.Recompiling can bring some optimizations, as we have new interpreters from high-level languages such as Solidity and Viper.Although there is no practical application yet, we need to migrate the system to a new system, and these are just some work ongoing.
In blockchain, the opcodes we use (a computer instruction) may affect the performance of our system.For example, the Ethereum Virtual Machine (EVM) uses an 8-bit opcode, while we use a 16-bit opcode.You might think it’s no big deal, but it takes about 40 hours to execute the first to the fifth million transaction, and an interpreter with an 8-bit opcode needs it, while an interpreter with a 16-bit opcode requires27 hours.This means a 30% improvement in performance.Of course, there are many other factors that affect the operation of the system, but this is indeed an important aspect.
In the field of blockchain, there are many researches on distributed systems, virtual machines, etc.However, for some reason, many people choose to ignore these research results, and they may think they can solve problems independently and do not need to rely on those that have spent decades of research.But we think we can apply proven and effective technologies to improve our systems.This is also one of the reasons why we decided to switch to 16-bit opcodes.
Let’s explain the opcode a little bit.In traditional EVM, when performing basic mathematical operations like a plus b and multiplying by c, it needs to be carried out in steps.First, add a to b, and then multiply by c.However, by introducing super instructions, we can find that this kind of operation of a plus b plus c accounts for the majority (more than 95%).So why don’t we combine them into a super operation?This can reduce the number of operations performed and improve the efficiency of the system.
Super set is a collection that combines two operations. Instead of executing a to add b alone, a to add b to the default.Typically, you need to read, modify, and write to the target first, and then perform a second operation.After introducing the super instruction structure, the number of operations that need to be performed can be reduced by half.Especially in current virtual machines, such as DeFi and NFT operations, such as ERC transfers, although it seems to be a standard operation, it is actually a series of operations.First, read your balance, then check whether the balance is sufficient, then subtract the corresponding amount from the balance, then update the other party’s balance, check again whether the balance remains consistent, and finally submit the operation.
We can imagine how many times these operations will happen on a blockchain, which may require multiple different opcodes to be executed, but by introducing a mechanism called super instruction, we can simplify this process.However, it should be pointed out that this is not a new concept, but a research result from decades ago.We haven’t applied this concept in the blockchain space before, but now we’re introducing it into virtual machines, which is an improvement.
We also spent a lot of time studying parallel execution, and this concept is very intuitive.For example, if I’m sending USDC to G, then M’s purchase of NFT should not wait until my sending is done before submitting to the same state.Often, many of the interactions that occur in a block are highly correlated.During high activity, multiple interactions may occur in the same state, with a large number of activities before and after the transaction.
After many optimizations and parallelization, we discovered an improvement phenomenon called CL (Clarvoyance).Simply put, Clarvoyance refers to the process of brutally sorting all transactions in the first 50 million blocks and then trying to reorder to find the best way to write state.We ended up with the best sorting scheme, which is what we call Optimum Clarance.In this way, we achieved a 30% performance improvement, which is a good improvement.While this is great, we have performance improvements elsewhere, some even reaching 800%, 400%, making CL less important.
Next, the next big improvement we make is after the virtual machine starts to accelerate.We developed a simulation environment called Substrate that allows us to make small changes and then run the entire system very quickly.This simulation environment is like a container where we can test some small changes to understand what impact they will have.In the absence of such a tool, it is difficult to test the impact of these changes.We used to spend a long time building a system and then realized that the answer was not working, and we have experienced this several times.In fact, this is the first tool we have built.We plan to open source it because it is compatible with any Ethereum virtual machine network, so it will be useful for other blockchain development teams, who can also take this approach.
This is also a tool we have been using to test all these different theories and gradual changes.By making small changes to the code base and then running it quickly, we can complete testing of the big dataset in a few hours instead of days.Part of this is our analyzer, which shows where we spend most of our time doing these transactions.After we introduced many virtual machine-level improvements, such as these opcodes and super opcodes, we also introduced hotspot caches.For example, if the state is frequently accessed, we keep it in the cache and then always access the cache directly without needing to read it again.This is basically the foundation of web development, even 50 years ago, but for some reason this optimization is not popular.Additionally, we implemented hash cache because we don’t want to recalculate all the time.This consumes a lot of resources, and many hash values, such as state tree roots, will change frequently with transaction verification.
However, we found that the next biggest bottleneck is disk, which is the database.We tracked all of these read and write operations, a lot of them were done in the background, and you might not even realize that, like what I mentioned before, such as index builds and other operations.Then we went through the key-value store from Badger to Pebble to the final key-value store Carmen.It actually has only two main components, namely the new set of patterns.I tried to simplify it, and we can discuss it in detail if you want it, but I’m just keeping a high level introduction now.
So, here is about address and address space, and the basic smart contract data search method you use when looking for values.Another feature is real-time pruning, especially for Fantom blockchain.Because Fantom uses an ABFT (Asynchronous Byzantine Fault Tolerance) system, it does not actually require the longest chain rule.Once you get 2/3 of the confirmation, you can truncate all previous records because you only care about the final state.Of course, you can still keep historical data for archive, and for the same purpose as Proof of History.But you don’t need the longest chain, you’ve already achieved true finality.
This is also a long-standing topic of debate, as there are risks in some cases.Suppose someone has a secret lab of super quantum computers somewhere, they can crack the password hash in a very short time, then create a new Bitcoin chain and submit it.This may not be likely to happen in systems like Ethereum, but for systems like Bitcoin, it is a risk.They can create a new longest chain and declare, “Hey, this is the new chain, we have everything in hand now, thank you.”
Therefore, it is risky to adopt the assumption of probability finality.So far, there have been so many blocks and work that it is nearly impossible to change even systems like Bitcoin can do it.They can have a stage where all validators sign, we agree that this is our new state and then operate in the future, and only that part of it may be affected.But this possibility does exist, and while it sounds a bit scary, it may be unlikely to happen in the next 50 years.
Therefore, one of the purposes of real-time pruning is to keep disk usage low.State swelling is a huge problem in any longest chain system.It is important to fix it before the problem becomes serious, because if your blockchain gets a lot of economic activity or gets attention, your status will grow rapidly.Your validators will start to be affected and they have to upgrade and improve.Therefore, the state area is always an important concern for us.
With real-time pruning, even without real-time pruning, we can reduce the flat storage and address index lookups we use.For example, the storage on the Sonic chain we are currently running is down by 98% on disk.This is because now you no longer need all those background indexes and lookups, and everything else the database brings.This is very important because as activity increases, so does hardware demand.You have to continually weigh the trade-off between increasing specifications and lowering requirements, which is an ongoing process that cannot be avoided.Whether in Web2 or traditional finance, you end up having to scale these things, but the costs associated with them can become very expensive, so you have to find new ways to change your code.In the traditional world, you repeat this process every six months.
Overall, Carman’s new data storage solves the major bottlenecks we encountered before, enabling approximately 8x throughput.The two main issues we are currently dealing with are the P2P layer and the transaction master pool, which are standard optimization projects.We take the traditional approach, gradually dissect the stack, find the biggest bottlenecks, focus on solving these problems, and then run it again through the analyzer, repeating so much until such a slight optimization that you feel like you want to commit suicide.
6. The path to expansion of ETH
Host 2:You said something interesting, I want to briefly talk about it and make it a transition as a topic.You mentioned that you don’t quite agree with the concept of parallel processing because in blockchain, many things are interrelated, while on the contrary, SUI or MOVE adopts an optimistic parallel processing method.We’re not sure how this performs in practice, but I want to use it as a transition because you’re a pioneer in cryptocurrencies, especially Ethereum.What do you think of the new scaling method adopted by Ethereum?
Andre Cronje:Ethereum has not made any difference in dealing with scaling, and they are already scared because past attempts have not been successful.I don’t quite agree with this, I often think about it, like Fantom.It has relatively little economic activity and low value, but every time we deploy or change, I feel nervous, even with the Sonic system that we have tested several times, because you know, an error in an opcode or a bytecode can cause huge losses.Especially after the Eminence incident, I was even more worried because even the smallest details you didn’t consider could have catastrophic consequences.
I respect Ethereum’s choice out of concerns about this financial loss.I think the transition to Proof of Work is a major historic moment, which is a huge success for Ethereum and all of the participants, and I want to give them appreciation and congratulations.But in my opinion, I have seen this when working in any company.When you are new to you, you are willing to take risks and keep breaking through barriers, but as reputation and value grow, you become no longer willing to take that risk.Ethereum’s strategy now is to take a conservative approach focusing on L2.People seem to forget that the Lightning Network is the first L2 on Bitcoin, but no one is actually doing anything there.This may be the right approach, but it shows that we are satisfied with the status quo and have little demand for improvement.Historically, this has happened whenever a new competitor emerges, capable of taking risks and surpassing the previous generation.This is what Ethereum does for Bitcoin, and I think the next generation Layer1 is doing the same for Ethereum.It’s too stupid to say that the Ethereum killer is that our economic activity, TVL and everything from Bitcoin, Ethereum, Solana and other networks are just a trivial part of the financial world.If you think this is the limit of global financial solutions, then this kind of tribalism is really crazy.
Anyway, I say that because you know, sometimes you don’t use Ethereum to buy a cup of coffee at Starbucks.It’s more about portfolios, or in some old-fashioned banks, they’ll only change things once a year.Ethereum is the perfect choice for these uses because it is built on factors such as security budgets.It’s a bit joking, but that’s why it exists.Just like if you want to pay someone, you won’t use Bitcoin because almost no one is using it.I still remember when I entered this field, my first salary was paid in bitcoin because then it was the main payment network.But now, I can’t imagine waiting for an hour to receive the bitcoin sent to me by someone else.I prefer to use Ethereum.
7. Fantom’s Marketing Plan
Host 1:Andre, I have a question here, I think it has something to do with a lot of what we are talking about, which is actually a marketing question about Fantom, but to really discuss this I want to go back to when you were at CurveWhen you were at Curve, you made this so-called fair start, and the founder didn’t take any tokens, which basically created a culture of personality cult, and someone even wrote a book calledThe e-book of Blue Pills, describes you almost as a god because you did this fair start project, I know you talked about this issue once, there was a time when you said I was just code and prod, and then the whole worshipThe group starts to form around you, which makes your operation very difficult because everything you do is highly scrutinized and focused, anyway, then you suddenly leave Yearn and then (Yearn’s) price is relatively overnightIt just collapsed, OK, and then after a while, you brought Fantom, and Fantom so it went up, I think it was in 2022 and then you left again.
At that time you actually wrote a blog where you explained why you left DeFi because of these toxic cultures and a lot of people started blaming you and saying you rug pull everyone, even though you were just one of the ones working on this projectPeople, but you rug pull us, that’s why the tokens fell 50%, and although you said in the cryptocurrency field, if the founder created a stable project and needed to leave, one of the reasons why these projects were so successfulNot just because you are the first to do this, similar to Yearn, you are the first to actually have this aggregator on Ethereum, but you establish that reputation when you go to Fantom, thisReputation follows you, which is an important reason why there is so much speculation behind the project.
My question is, now you’re in Fantom, how do you think about marketing, because all these blockchains are working to achieve this convergence to some extent to have the same kernel,But there are differences in some subtle differences, so now how do you think about this because you no longer really rely on personality cults to attract speculation and money, how do you think about differentiating Fantom from other projects when many technologies seem to beWhen converging to the same location?
Andre Cronje:Yes, so I think we’ve taken the right pace in that direction, but the first thing we need to do is determine who is our real user, you know, is with Rabby or metamask or whatever you want to useAre the people who interact with the wallet Ledger? We quickly realized that they are DAPP developers, they are our customers, they are our users, and these people are the objects we should serve.
Then we start thinking about how we can improve the lives of our DAPP developers, there are many small things, such as, just give an abft example, when you issue an RPC call to submit a transaction, when you receive your 200 OK replyWhen this means that the transaction has been confirmed, it is final, completed, you don’t have to wait in the background and poll, you know that it is confirmed, when the blocks are released, whether there are enough blocks,This is a lot of extra boilerplate code. If you are doing something about the website, you have to write as an adapter, there are many small examples, we are just trying to improve the coding experience and the overall development process, and the next thing we consider is most of the costs,gas fees and so on will give security validators, we can take 20% of them from and support the economic activity of DAPPs, so they actually have a source of income, not just a large-scale fundraising at the beginning, and then slow downSlowly drain and die because you don’t have another source of income and you don’t want to charge users because then someone will just copy your code and provide service for free until eventually it also dies, but that’s anotherOne thing.
Our gas incentive program is a very successful big project.Gas subsidy is another thing. When new users come to use someone’s DAPP, they don’t need to pay gas fees. DAPP developers can provide a little subsidy, they can get it from their gas incentives to pay for all these interactionsFees, which is like a local repeater, which means that their users will get a much simpler entry experience.
Plus, we focus more on school/university-level projects, we hold hackathons every three months and work hard with students to teach them because we don’t expect them to be the next batch of buidlers, but I hope in five to ten yearsThey were able to be the next batch of buidlers and then you know, it would be great if they were familiar with our technology stack and ecosystem.So, this is a longer-term approach because we realize that fighting for mental share in the current environment is meaningless and a total waste of time.Because in the past we actually started with the incubation and seed investment teams, and then they would usually go into a round A or B, or whatever, and then find the dominant venture capital.Fantom does not have VC support, it is just public sales, so we do not have those big-name VCs behind us.Then when these teams go to the dominant investors, their dominant investors always tell them, “No, you’re going to post on the ABC chain, not on Fantom because that’s where they invest.” SoThey can’t be blamed, but at the same time I feel a little deceived.So we basically stop interacting with those teams and things.
I also believe that most developers currently building decentralized applications are just copying what we see in the traditional world, and that’s nothing wrong, but the next important decentralized applications will be those who grow with decentralized networksThe people who created them wouldn’t be those 30-year-old developers who are now starting to think about new things, because we haven’t encoded this thinking into our DNA.So we also want to cultivate this new energy.I mean, this may not be what your token holders and communities want now, but I don’t think we’re going to win that game at all in the next two to four years, and I don’t think our tech stack has yetGoing to the level we should do to achieve that goal, but we are lucky in terms of spending and funding, we can take on something quite long term, which is what we are trying.
Host 1:We have a very deep community here.I saw you tweeting and someone asked you about whether rollups would come to Fantom, you said there were no rollups on the roadmap, but we were doing something about SVM.Can you give us some Alpha Insight?
8. SVM is the best virtual machine
Andre Cronje:Yes, like I mentioned before, I think SVM is the best virtual machine right now.Our consensus mechanism is separate from virtual machines, which means we can easily integrate other virtual machines (such as FVM) into our systems.Our goal is to use the consensus mechanism as a sorting system and then output and process the data from it, which is not complicated.While one might expect to see something more exciting, we are convinced that SVM is the best virtual machine technology right now and we are working to combine it with our consensus system.
9. Airdrop, regulations and bull market forecasts
Host 1:If you have any attention to Solana, you probably know that they have recently conducted JTO airdrops, which is to some extent an airdrop season. I haven’t seen much tokens before, but now there are a lot of them.You’ve been through the airdrop process and I know it’s very interesting for some people who aren’t very familiar with it.Yearn had airdropped all the tokens at once before, during the defi boom in 2020.A year later, the community actually worked together to reduce the supply of tokens by 30% to reward contributors and put a portion of the funds into the fiscal treasury.I think people have realized now that maybe we should not release all the tokens at once, nor should we ignore the establishment of the fiscal treasury.But you were too early at that time.I’m just curious, what advice or lessons do you have on this?How do you think of these airdrops in order to succeed?Will you continue to do this, or do you think it’s just a business practice?
Andre Cronje:No, I won’t do that now.The airdrop has been operated by machines and turned into a liquid farm, losing its real value. Like similar to Keeper, these tokens are not available for free, you can’t mine to obtain them. I call them “olm option” at the timeLiquidity mining, you obtain options through mining.If I invested my liquidity or something and I made 10 tokens, I wouldn’t actually get those 10 tokens, I just got the right to buy these tokens at a discount, so we did it at the time50% or something so I can buy these tokens at 50% of the current market price.
We were testing this approach back then and I think there was something nice about it, and although I don’t feel it’s adopted as wide as I expected, you know, one of the best systems I’ve ever seen,Inversion, I think that’s a good example, it’s done manually, the first person said, “Well, I like you three, you three will get some tokens.” So now let’s form one togetherCommittee, the four of us decide who is the next contributor to get the tokens, and I think that is one of the best, because the airdrops are now mercenaries, it’s hard to avoid this.Another thing is that no matter what you do, there will always be a large majority of people who are not satisfied, so the best way is if they are not satisfied no matter what, then exchange some value for it and don’t give it to you in vain.While this has attracted a lot of attention and attention, I also think it is not a good thing for a lot of teams because their initial ICO airdrop launches fairly, or whatever you call it, that’s their peak, the best event,The best of all, and then from then on, slowly, slowly, silently corrode until death.
I understand it’s because you’re in a highly competitive field and if you don’t do that, you have to put in more effort, but if you take the long-term path and decide to make the long-term choice, you won’t be thatWith more hype, you won’t have that many users, but I can assure you that those airdrops and xx and others don’t have that kind of real hype and real users, just have some financial value for the project to take advantage of, that’s itThere are many reasons why there are activities, but a year ahead they may have lower usage and value than those who don’t do that and just slowly get stuck.In this field, it is really hard to compete with a lot of lies and scams that people think are true and will defend them relentlessly, you have to fight against that kind of slow way that is really hard, but in the long run, it worksbetter.
You know, for me, a perfect example is Solana, where they could have taken the EVM approach, like “go to him, let’s integrate the EVM into Proof of History”, but then we integratedThat moment, I think it was one of our biggest mistakes because at that point in time, we made a quick trade-off.Now, that means people can come to Fantom very quickly and deploy things, but that’s why you see 99% of the stuff on the new chain is forked, people copy and paste from A to B, and then they do the mining, do liquidity mining, create tokens, and then the team leaves, and then others leave, because you can migrate back so easily, unlike if you do something new, like SVM, it does take longerTime to accumulate users, you have to do more teaching and other things, but like these people you have at hand, you know, they can’t give up quickly and then jump away.
Even if they take these measures, they realize, “Oh, it’s not actually that good here, I want to go back.” So this proves that Solana is a good case for me, and the long-term approach is usuallyBetter choices and better rewards, but you have to endure them for longer.
Host 1:This is a very insightful point of view.For everyone, Fantom is the second oldest EVM chain, second only to Ethereum, and obviously now you have what I call FVM, which is coming soon.OK, I promise this is the last question, at the end of us, you used to be on a podcast, I think it was in February, and they asked you know when the next bull market will come, you said you actuallyI really appreciate the bear market because that’s when you can focus and build, but you did say there that you think the end of the year will come, and now the prices are rising a lot, I’m just curious if you feel the same way, do you think the tide has changed, or people are just too bored, we have this temporary increase.
Andre Cronje:There are so many factors at work now that I don’t think the world is changing so quickly.Major events like covid, Ukraine, Russia, Israel, Palestine have happened so much this year that it is difficult to make any real predictions.I do think that emotions have changed, and emotions change is usually consistent with price activity, and now the mood changes are positive, people are happy again, and people are starting to enjoy it again.Most people will, I don’t know if it’s Buffett or Munger or who said, you have to wait for the tide to recede before you can see who is not wearing clothes. This has happened before, we saw who is not wearing clothes, they ran away or were lockedGet up, or whatever.Now those who still stayed have gone through the depression and everything else, like you are in the “but it can’t get worse” stage, so it has to get better stage.
But I think, we’re there now, I hope I hope it’s true, it can’t get worse, but I mean, there’s so many catastrophic events going on over the years that you never know what’s going to happen nextWhat, assuming there is no new event in a lifetime, I think the situation is moving in a positive direction.I think the pain in the traditional finance world is not over yet, I don’t know how much of this will affect, I think there will be more pain there, at least a year or two, and I do think cryptocurrencies may be with thisDecoupling, I’m optimistic that because cryptocurrencies are more interesting, so unless there is a catastrophic event, I think there is a reason to be a little bit optimistic.
Host 1:As you mentioned before, it’s hard to imagine going back to a private equity or investment banking job after getting involved in cryptocurrencies.I think if someone does this, they might turn to another frontier area, like artificial intelligence.Now people may be back in the cryptocurrency space, so it’s very interesting.I believe everyone who is listening can feel how smart Andre Cronje is and how much impact he has had on the ecosystem.Also, I didn’t know until I was preparing for this interview that you were a lawyer, or at least you had a law degree, is that true?
Andre Cronje:I can’t practice, I have never taken the bar exam.The first thing I first learned was law, probably around 2001, so definitely don’t trust me to give legal advice.
Host 1:I believe that as you enter the crypto world, it will also be beneficial to have some legal knowledge.
Andre Cronje:Yes, it was also an important driver of my entire crypto-regulatory affairs, but I also suffered severe criticism for this.The argument I’ve been trying to make is that this is why I distinguish between regulated crypto and crypto regulation because crypto regulation is impossible.You know, we’re going to see in all these countries, they’re all like, “Okay, but you can’t make ETH transfers on XYZ.” But you can’t stop this, like I can run my own node and do it, youThere is no way to really stop this, but you can provide useful legislation for regulated crypto companies like Circle, Tether, etc., like we see now like Coinbase, Binance, you know, they can do thisa little.I’ve always been an advocate of this, you know, regulating things you can and should regulating, but things on the chain are not regulating.We don’t necessarily know what’s going on, the lessons learned from FTX, Three Arrows, etc., but also admit that you can’t really touch cryptocurrencies.But this subtlety is lost in this area, just like what everyone hears is Andre Cronje wanting to regulate cryptocurrencies and pulling up the documents behind him to show to others, but that is not the case.There are subtleties here, but you know, subtleties aren’t suitable for headlines, and no one has the time to read more than a line.This is just a simple question.
Host 1:Yes, I think it’s a bit whimsical to have no rules at all, I think a lot of people, you know, they say what we want most is clarity, but that also points to L2s, like Base, run by Coinbase, a single entity, thisIt may be regulated, you know, maybe in the long run, it’s actually a sign of a bull market, you know, because maybe you’ll have more users certified by KYC, they can use Coinbase’s credentials and so on, and thenYou might have something like Fantom or Solana that doesn’t require KYC, or some applications don’t need it because now you can do these hooks, you can know that it will only be sent when running the program behind the scenes.So I don’t know what the future will look like, but regulation like that will certainly become part of a lot of cryptocurrencies.
Andre Cronje:I agree, I think, you know, like we see in Ethereum, a lot of block producers don’t allow transactions to those on the list, although the transactions will still happen because others run the allowed oneBlock producer.But you know, if I was a validator and I was running in the OFAC list with a significant U.S., I needed to comply with this because I lived in that jurisdiction, it was that simple.But if I’m in a remote and nameless country where the list doesn’t matter, I can run a validator without it, I’m not under your jurisdiction.So for me, entities should regulate cryptocurrencies where they can recognize.Like, “Hey, X people do Y activities in my jurisdiction, and that’s my jurisdiction.” But trying to stop something globally, and then you know, that becomes if the developer is like the foundation layer, they take responsibility, and even if they want to change these things and the verifiers don’t adopt them, then they are done.Then they’ll try to lock in those bad guys, which we’ve seen in some cases.
This is a phenomenon of losing opportunities.It’s like you know, and that’s what I want our entire industry to understand and we have to help educate.If we just keep burning the flags and trying to get these guys out, the more you resist, the more they can fight back, and they will make your life painful.At a certain point, you can only say powerless so many times until you start lying and saying OK, I’ll do something and then set up a backdoor or something in it, which is worse.
Host 1:Does this still make you feel scared?I know you mentioned the token launch in Yearn and so on, do you know the CIA or maybe it will come to you, and that’s still in your heart?
Andre Cronje:Yes, look, I’ve had my fair share discussion with a few three letter agencies and I admit that when I first received that letter, I was scared.But as the process progresses, you begin to understand that it is discovery and education.I think it’s actually much better than I originally thought.And you know, I thought it was fun at that time because like everyone on crypto Twitter was saying, “Ah, these guys don’t know what’s going on, it’s going to take years to understand and catch up.”And the questions these guys ask me are so low-level technical, like I think you know the field better than most people on Twitter, and there is no offense, but this is how I interact with them.Overall, honestly, it’s actually a little more positive than I originally thought.But it’s still scary, but you know, things will change soon.I may make a lot of low-level mistakes, one of which is not to raise funds, is to issue tokens, everyone comes and then launch fairly.The other is leaving at due time, and if not for these two points, I think I will have more problems now.
Host 1:Yes, well said, this is actually some of the fightback I’ve heard about the airdrop style I think you describe as an option.Some people worry that this will actually put protocol operators in the control of regulation because you are selling things, right?
Andre Cronje:Yes, yes.
Host 1:Anyway, Andre, thank you for being involved and sharing your story, it’s very engaging.I’ve been following your story and never thought I would actually be able to talk to you, so it’s really cool.So thank you for your participation.
Andre Cronje:Welcome anytime.This is my honor, thank you for your invitation.
Host 1:OK, see you next time.
