With the recent rise of the AI Meme sector, AI+Trading has become a hot topic. So how will AI change the trading logic of cryptocurrencies?
Moderator: Anderson Sima, Executive Editor, Foresight News
Guests: WOO COO Willy Chuang, Kaito founder Yu Hu, LingX dry goods store owner Da Chengzi
Host: Please ask the three guests to introduce themselves and their projects.
Willy Chuang: I am Willy, and I am currently the COO of WOO. Let me introduce WOO to you. We are a centralized exchange, and we also have a decentralized exchange called WOOFi, so WOO X and WOOFi are our core products. We started about three years ago, focusing on professional traders, because most of the team members come from traditional finance, and have a deep understanding of trading, especially high-frequency trading, and market making, and have many years of experience. Most of them were born in Wall Street, so we have our own insights into the market situation, and we have relatively good trading and technical capabilities, as well as complete infrastructure. So at that time we wanted to make a product that was particularly friendly to professional traders, and WOO X was born. After that, we began to make the product more comprehensive, and made many improvements and adjustments to the needs of various users.
At present, the daily trading volume of our products is about 1 billion US dollars, and the main products are contracts. Most of the users of our products are actually from Europe. This year we started to work hard to enter the APEC market. We have just launched the Social Trading function recently, and it has been very successful. We are also the first to launch the reverse order function, and the market response after the initial release was very good. This time we are fortunate to cooperate with Kaito and launch George AI automatic trading AI robot to operate our Social Trading function, and form trading strategies through Kaito's social signals. This is what we have always hoped to launch innovative products, rather than just doing a centralized exchange step by step. We hope to make our platform fun. At the same time, solve user pain points and become more friendly.
Let me briefly talk about the decentralized exchange. Its core product is cross-chain transactions. It now supports 12 chains, including Solana to EVM. Therefore, we have not been directly facing a lot of trading volume sources like 1inch in DeFi. Instead, other aggregators give them to us, and we help them settle and complete their transactions.
Of course, for WOOFi, we also have a staking program for our WOO token. WOO token was released a few years ago and is listed on major exchanges, including Binance, OKX, Bybit, etc. Various decentralized exchanges have contract spot. WOO token is currently ranked around 150th in market value. We also pay attention to the establishment of the WOO token ecosystem. There are also many investments from the perspective of WOO token. We are also actively building the integrity of the ecosystem. This is the general situation. At the same time, we were also a project invested by Binance in the early days, and I think it has been developing very well.
Yu Hu: Hello everyone, I am the founder of Kaito AI. I have been working in the traditional financial field for about 10 years, mainly in trading. Before founding Kaito, I worked in a hedge fund for about 5 years, mainly trading stocks. Since 2017, I have developed a strong interest in cryptocurrencies and have also traded some cryptocurrencies in my spare time.
Later, I gradually discovered that the useful financial tools I used in traditional finance did not exist in the cryptocurrency market. In particular, information was very fragmented, and many things could not be found on traditional search engines, such as Google or other search engines. At the same time, it was also very difficult to find information on Twitter. Therefore, I thought of creating such a search engine-based product myself, which is actually a very high-tech product. So, I went to find my former friends who left Facebook and Amazon, and we co-founded this search engine-based financial product.
About two and a half years ago, we have been focusing on building a search engine. Until the beginning of this year, we released the product for the first time. At the beginning, it was mainly used by hardcore research institutions such as the Ethereum Foundation, Pantera, and Grayscale, who were also our early customers. After that, we gradually developed trading indicators based on the search engine, which caused a great response in the market.
Since the beginning of this year, we have released some new features step by step, such as the attention indicator mentioned by the host, our public opinion indicator, and how to mine project indicators through tracking information of important KOLs on Twitter. These features have opened up a huge market for us. From the beginning of the year to now, we have grown about 70 times. Now there are more than 500 institutions using our products, and of course there are also some hardcore individual traders using them.
There are several main ways to use our products. One is to do basic research, and use search engines to see who is discussing what. This helps you search very quickly from an information perspective. We have actually collected information from Twitter, Discord, Telegram, and all intellectual platforms, blogs, including Twitter Space, etc. Moreover, we use AI technology to convert this information into useful indicators.
On the other hand, many people also use our attention products or indicators, such as comparing the market value and attention ratio of a certain Meme Token, or tracking which tokens are being discussed more now, so as to enter in the early stage. I will continue to share more about these with you later.
Of course, many people also use our tracking data to discover some early projects, or VCs discover some early investment targets. So now we have a variety of use cases. If we look back, Kaito will make more community-oriented products. We now have a very cohesive institutional community, but we hope to have a stronger retail-based community in the future. Therefore, many of our future products will bring our trading indicators to all ordinary retail groups.
This time, we have partnered with WOO to launch George AI, allowing everyone to indirectly access all of Kaito's trading signals and gain market insights based on this.
Big Orange: Hello everyone, I am Big Orange. I entered the Crypto field early. Specifically, I entered the crypto in 2013 through the XRP arbitrage transaction, which was a move from overseas exchanges to domestic exchanges. Since then, I have been constantly learning in the crypto, and now I can barely be considered a semi-professional trader. However, I mainly focus on arbitrage transactions, and have less involvement in areas such as CTA with one-way direction. It is worth mentioning that even some of the gameplay of DeFi is essentially inseparable from arbitrage transactions.
Today, I actually came here with a learning mentality, because this topic and sector is still relatively new in our circle. Although there may be some mature products, I hope to understand it more deeply, and I also have some questions of my own.
I think AI is something we must learn in the future. AI tools will undoubtedly become an indispensable assistant for every trader in assisting transactions. Recently, Ethereum has performed relatively strongly. Without some tools to assist in analysis, we may find it difficult to understand the reasons behind it. But if there are AI tools that can help us sort out and analyze public opinion, we will find that it was because of the change of the SEC chairman after Trump took office that the market believed that the Ethereum ETF might support Staking and distribute the proceeds to users. This is just one example, but it is enough to illustrate the potential of AI in trading.
I think AI can help us make price predictions in CTA and assist us in determining the subjective trading direction. In addition, AI can also play an important role in currency exchange. Functions such as public opinion analysis and information collection are indispensable components of AI in future transactions. Therefore, I think AI is a very good future tool that every trader must master in the future. At the same time, I also look forward to using AI to optimize our own trading models and iterate and upgrade them to better adapt to market changes.
Host: How to combine AI technology and crypto trading?
Willy Chuang: I would like to share with you a story about our cooperation with Kaito. This happened around August of this year. At that time, we were examining all AI projects from the perspective of ecological investment, and were working on the sorting and planning of the entire AI project. Kaito is undoubtedly a striking presence, especially considering their strong investor background. Therefore, we took the initiative to contact Kaito's team and started in-depth exchanges. However, during the conversation, we learned that they had completed a large amount of financing and were not our direct investment target. But this did not stop us from thinking about how to cooperate with Kaito.
When considering using Kaito's products, one point that particularly attracted us was the management of our own WOO Token. The management of WOO Token is actually a very complex task, and we have put a lot of effort into it. One of the tasks is to study the attention economy effect of WOO Token. I think that people's cognitive focus is limited, very limited. For example, when asked how many luxury brands they can name, most people may only be able to list 10 to 15. Similarly, in the field of cryptocurrency, the number of tokens and projects that most people will actually pay attention to every day is also limited. In addition, various factors must be considered, such as sector rotation, the different situations of traditional finance and cryptocurrency markets. On social media such as Twitter and Discord, the information is even more complex and dazzling.
At the time, we were thinking about how to improve the competitiveness of our tokens in the attention economy. Kaito provided us with a lot of valuable data, and now Kaito's data has become an important reference for our internal PR and brand departments to conduct evaluations. At the Token 2049 conference in Singapore, I met with Yu Hu. At the time, we didn't have too many ideas, but we felt that Kaito's signals were very good, and they also did some backtesting themselves. So, we decided to put this idea into practice, develop an AI trading robot, and combine it with our Social Trading. After two months of hard work, this product was finally born.
Speaking of the problems we want to solve, the product of leading orders is actually the underlying logic of our original design of this project. Thinking further, you will find that the lead traders (we call them LTs) mainly conduct subjective transactions. From a data perspective, their transactions are very volatile and have high drawdowns. When the market is volatile, if the market is good, everyone makes money; but if the market is not good, the followers will also suffer losses. Moreover, it is difficult for these subjective traders to accurately capture the rotation of sectors, such as in-depth analysis of sub-sectors such as DePIN and DeSci, or the BTC ecosystem before.
At that time, we felt that through Kaito's signals, we could clearly capture the market's attention and then convert this attention into trading strategies. This is undoubtedly an attractive option for users. The trading range we set for users is relatively conservative, trying to ensure high returns, so that everyone's income (PNL) and return on investment (RIO) are relatively high.
When we released this product, we also wanted to make it more interesting, so we let George AI compete with several real subjective traders. From the data, although some people are worried that KOLs will be replaced by AI in the future, many Agents now interact spontaneously on Twitter. Some people also think that AI's video and image generation technology may replace KOLs. But I think there is still a deep emotional connection between people.
In the end, we found that through this activity, subjective traders attracted more funds than George AI, but George AI's overall performance was better than subjective traders. This is a very interesting phenomenon. This activity may last for 30 days, and we will continue to pay attention to the results. But we ultimately hope that this AI can be the first step for our WOO to develop in the field of AI, and work hand in hand with Kaito to launch this product together.
Yu Hu: First, let me start with my personal background. Before joining Kaito, I worked at Citadel Hedge Fund for more than five years. Citadel is actually one of the largest market makers in the US stock market and is very keen on using artificial intelligence technology to develop various quantitative trading strategies. Therefore, in the traditional financial market, AI trading is already quite common, and different strategies have their own unique development directions.
However, one striking feature of the cryptocurrency market is that it is significantly different from traditional stock trading. The cryptocurrency market is a sentiment-driven market. I have long believed that if you can have the best data in this market and launch a sentiment-based trading strategy, the effect will be far better than the traditional stock market. Although emotions also play an important role in traditional stock trading, their influence is undoubtedly more significant in the sentiment-driven market of cryptocurrency.
From our perspective, the key is how to extract effective trading signals from a massive amount of information. This is undoubtedly a challenging task. Take Twitter as an example. It is full of noise, reports, and various information posted by robots, and the signal-to-noise ratio is extremely low. In order to understand each tweet semantically, we need to consider factors such as the publisher of the tweet and the influence of his historical behavior.
Therefore, this involves the construction of a huge data system, including data indexing, cleaning, understanding, and the construction of the entire data system. Only on this basis can we create useful trading indicators. Since our team itself has a Twitter background, when reviewing these trading indicators, we always use a critical eye to evaluate their effectiveness for the market.
We discovered many trading strategies early in the project. For example, when discussing with Willy, we showed him Solana’s trading indicators and the sentiment comparison between Ethereum and Solana. In the three extreme sentiment confrontations between Ethereum and Solana that occurred last year, each was accompanied by extreme values of the exchange rate between the two. These phenomena were well reflected in the market, which strengthened our confidence in developing trading signal strategies.
Next, I would like to share with you how we implemented this strategy from a technical perspective and some of the difficulties we encountered. As for why we do this, I think there are two main trends.
First of all, the relationship between humans and machines in the future will be division of labor and cooperation. There are some things that humans will do better, such as the relationship between KOLs and people mentioned by Willy. But in some things, the performance of machines will far exceed that of humans, such as what we are doing now to use machines to measure emotions. Because humans cannot read every piece of information, nor can they completely get rid of the influence of subjective emotions, it is even more difficult to conduct quantitative analysis. For most Meme transactions, many people cannot perceive the changes in the popularity of Memes in time. Machines can capture and quantify these emotional changes in real time. Therefore, in terms of transactions and interactions, the division of labor and cooperation between humans and machines will be the general trend.
Another trend is that both trading and other activities will move towards a simpler direction. I remember that in August of this year, I mentioned an interesting example. In the United States, there is now a DEX called Narrative Trading, which allows people to invest in AI without studying complex functions or financial reports. In the cryptocurrency market, many people have similar ideas. They don’t know which Meme is worth investing in, so they buy a few and then stop paying attention. If they can do some simpler operations, such as copying George AI, etc., it will be easier to manage, understand and operate for most people.
This is also an important trend in the future, which is to make everything simpler. People can express their opinions more easily without spending a lot of time studying the holdings, trends, and attention of each Meme. We also firmly believe that this is a huge trend in the future. Therefore, when WOO found us, based on these two considerations, we were very willing to cooperate like this. We believe this is an important development direction for the industry in the future.
Big Orange: The tools I currently use include Nansen and Glassnode, which I have paid for, and tools like TradingView are essential for my daily life. I actually built a personal trading panel, in which I expect programmers to integrate the key data I need. For example, my current panel already includes data such as Binance U contract volume, order book depth of mainstream currencies, and options data.
It is worth mentioning that I always think that the bulk data of options often have a counter-indicative meaning. For example, when there is a large amount of short selling data, after the market maker takes over, they may hedge the risk by buying call options (call) and short in the contract market. This seems to me to be a coherent market behavior, but for individual traders, there is often a lag in analyzing these data, and each market action may trigger a chain reaction elsewhere. For example, an action on Deribit may affect Coinbase's USDC/USDT trading pair, because the flow of funds will also have an impact on the market. The complexity of the data I face is that my panel integrates up to a dozen different data sources, covering multiple aspects such as the total locked volume (TVL) on the chain and the circulation of stablecoins.
What I have always wanted to explore is whether there is an AI tool that can receive this data currently organized by programmers and analyze the correlations among them through data models. Combined with historical trends, I believe that there must be some correlation between these data, and this correlation is of great significance for subjective trading and the formulation of quantitative trading strategies (such as CTA). However, at present, I still need to manually analyze the data of each sector every day. Although I can evaluate some simple correlations, it is still a challenge for me to dig deep into the complex interrelationships between these dozen data sources and form a systematic methodology.
I hope to be able to hand over all the collected data to AI tools to help me analyze the opinions and correlations, including historical retrospective correlations. Of course, as the previous guests said, factors such as media analysis and market heat are also indispensable. Especially in the hype of Meme coins and Altcoin, communities often become the main channel for retail investors to obtain information. Now, with the help of AI robots, we can collect more information from foreign communities and Twitter, as well as announcements and keywords from various exchanges. By analyzing this information through the unit model, I believe that this will become a tool that every trader will use in the future. Therefore, I think the application of AI tools in trading is crucial. In all aspects of trading, everyone should have their own AI use strategy and direction.
Host: Is it possible for AI to mine Memes?
Willy: My personal opinion is that we divide the user's journey into three parts. The first part is Onboarding, which is how to use the power of AI to introduce more users into the Crypto field. The second part is Discovery, which, as Mr. Dachengzi mentioned, includes various transaction information, as well as Social Signals that Kaito is doing. How to integrate this information to provide better trading support for traders or teams is a question we need to think about. The third part is execution, which is actually an emerging field. Since about last year, the development of AI has gone through several stages, from ChatGPT to large language models, and then to earlier automation tools. However, those early automation tools were not intelligent. They just performed operations within a preset framework and could not self-optimize logic. It was not until the emergence of large language models and ChatGPT that AI began to have logical capabilities.
Now, we have entered the so-called Agents era. Various Agents such as Goat and Virtual have been strategically matched. They can not only expand the influence of the model, actively post on social media, interact with different accounts, but also have the ability to trade autonomously. The framework of Agents is gradually taking shape, which is exactly the third step we mentioned earlier - execution. It can implement what we want to do. But at present, the biggest problem is still that the various protocols or systems have not yet been connected, and everyone has not been able to truly open up their capabilities. But I think this day will come soon. Because in addition to executing transactions, Agents can also conduct self-rehearsal and innovation.
When all this becomes possible, we will usher in a true Trading Copilot. Trading Copilot may revolutionize the risk management, security, and personalized experience of trading. In the trading process, we may no longer need to pay attention to so much complex information and deep learning. From the DeFi level, operations that originally required more than a dozen steps may now only require an intention form, which can be completed through Agents interacting with smart contracts at the bottom layer. However, there are still some problems. First, I think the API is not mature enough. Secondly, from the data we need to determine the authenticity of the information, such as whether the price of Bitcoin is real and where this price comes from. This may not be a problem for simple price queries, but at the Meme level, things become complicated. So, is there a possibility of error?
In the field of Crypto, people are trying it with real money, not just information-level exchanges. So, these problems may be very tricky. However, I personally think that this trend is unlikely to stop, and it may double in growth. In particular, the combination of Meme and AI, I think it is the best combination at present. On the one hand, there is the incentive of tokens, and on the other hand, it has the ability to be fully automated. The combination of these two, as can be seen from Kaito's board, has attracted more than 50% of the attention of the Crypto industry. Therefore, I think this trend will not stop, but many projects will be eliminated in the process. Only those projects that can really make products, solve user problems, and give users a better experience can survive. And those projects that just release a lot of tokens and pursue token price growth but have a poor overall trend may have a certain negative impact on the development of the industry. This is my personal opinion.
Yu Hu: First of all, I would like to talk about the combination of Meme and AI from a general perspective. Recently, the attention of Meme has increased significantly, from about 5-10% at the beginning of the year to about 15-20% now. At the same time, the attention of AI is as high as 30%. Therefore, the combination of AI and Meme is undoubtedly a track with great potential, occupying nearly half of the attention in the Crypto field.
At the beginning of the year, we pointed out that Meme only accounted for 2% of the Crypto market value, but today this proportion has risen to about 4%, although it is still far from its 15-20% attention. This shows that although the Meme market is small, it is favored by investors and funds flow frequently. Therefore, we are still optimistic about Meme's performance in this cycle.
Next, I would like to share some observations at the micro level. Currently, many data platforms such as Nansen are tracking Meme transactions. In the division of labor between humans and machines, we can clearly see the advantages of both. At the cultural or Meme level, human judgment is far better than that of machines. For example, when Clanker releases a Santa Meme but it unexpectedly becomes a bug, or when a meme full of memes such as PNUT appears, the machine cannot capture the cultural significance and fun behind it as quickly as humans. However, at the trading level, machines show strong advantages. When the attention of Memes such as PNUT continues to decline, machines can make more accurate trading decisions based on data signals, while humans may ignore these signals due to subjective judgment.
In addition, we also noticed that KOLs play a pivotal role in the Meme market. They lead the market trend by controlling the attention economy. For example, Murad led a wave of Meme craze after his speech at Token 2049, and recently a new group of KOLs led the trend of AI Meme. These KOLs are important because they control the lifeline of the attention economy.
Machines have also demonstrated their powerful capabilities in tracking KOL holdings and following new projects. They can monitor KOL holdings and following dynamics in real time, helping investors quickly understand the historical background, origins, and relationship networks of projects. For example, when a16z first appeared, the machine quickly captured the attention of a16z founder Marc Andreessen, providing investors with valuable information.
However, humans have more advantages in obtaining external information and dealing with black swan events. For example, events such as ACT's listing on Binance are unpredictable by machines, which requires investors to make comprehensive judgments based on market opinion at the time, Binance's leadership decisions, and wind direction. In such a process, I think the best way is to make the division of labor between humans and machines very clear, and give full play to their respective strengths. From a human perspective, we can understand some soft things very well. For example, when a Token is just released, you need to understand its narrative, its background, and cultural aspects. This is what humans do well.
But at any time, people are better at judging the market or externalities. But for example, when there are a lot of sentiment indicators or some tracking indicators, when the market value of a token becomes larger and the attention economy determines the trend of the token, machines are often better at this time.
Big Orange: I will talk about some problems or challenges that AI Trading will face in the future. AI Trading can actually be understood as artificial intelligence quantification. Generally speaking, quantification will have problems that AI may also have. For example, the first point is the quality and diversity of signal sources. Quantification actually has this problem, which is that traders are constantly required to improve and iterate the signals.
For example, CTA needs to be iterated all the time. I think even AI Trading may have this problem. It’s just that for the same signal, it may have a stronger ability to analyze and collect data, but for the iteration of different signals, I think it still requires manual adjustment of the model.
The second point I think is the avoidance of some black swan events. What CTA strategy fears most is something that is not anticipated outside the model. I think AI Trading may also have this problem. We need to observe what the AI's ability to respond is by then.
The third point should be the issue of computing the required resources, because after all, it is AI Trading. Does it cost a lot to maintain AI Trading? Then is it equal to the benefits generated by your strategy?
Host: What is your outlook for future market conditions?
Willy: Regarding short-term fluctuations, I think the current environment does not give too many clear comments. However, in the long run, I am still optimistic. This is mainly based on the following reasons: First, despite the many uncertainties brought about by the trade war, the overall environment is still positive. Judging from traditional financial data, the economic conditions in Europe and the United States are still strong, and many core technology industries are still developing. In addition, the annual growth rate in the field of AI is expected to be as high as 16%, and the market size will reach 240 billion US dollars in 2025. These figures undoubtedly show the strong investment driving force of the technology industry.
In the field of cryptocurrency, I have observed that the market is currently showing a polarized trend. On the one hand, the attention economy, cultural drive and community power represented by Meme are rising and attracting a lot of attention.
On the other hand, some relatively mature projects, although they may not be so eye-catching, have performed steadily in terms of indicators such as revenue, application and TVL. I think the market will look for opportunities between these two extremes, and liquidity will also gather in these two directions. Projects in the middle may face a more embarrassing situation.
At present, those projects with fundamental advantages have not yet fully demonstrated their value in 2024. But I think that by 2025, the potential of these projects may gradually emerge. Of course, this is just my personal opinion and is for reference only.
Yu Hu: I am optimistic about the market outlook, mainly based on two rather special events in the current cycle. The first is the approval of ETFs, which triggered an early and abnormal market pull at the beginning of this year. The catalyst for this wave of market can be traced back to the global policy benefits brought about by Trump's coming to power, which injected strong momentum into the market.
Looking back at the last cycle, the Fed's massive money-printing policy was undoubtedly a major turning point. This time, the cycle has not yet fully unfolded, and the most critical and decisive factor will be the trend of risky assets. From historical experience, the market usually experiences a lag period within 6 to 12 months after the first rate cut. Therefore, we have not yet entered this critical window.
Of course, in addition to the above factors, there are other variables worth paying attention to. For example, the compensation involved in the FTX incident is expected to be gradually distributed in the first quarter of next year. Although this topic has gradually faded from the public eye, from a medium-term perspective, it will still have a positive impact on the market outlook, further strengthening my optimistic expectations for the market outlook.
Big Orange: I think that before January 28, 2025, the overall market should be relatively stable. Today's pullback is mainly because the moving average is too far away, and the market has been rising for many days in a row. The pullback meets the technical requirements. At present, the market may consolidate near the daily moving average. For those investors who have been waiting to enter the market, I suggest that you can wait until the market returns to the moving average and consider buying some large-cap currencies.
If you are currently holding a long position and facing pressure, and you intend to continue holding, then I suggest that you consider buying doomsday options every day, especially doomsday put options. The reason for doing this is that you can provide some protection for your long position in case of a possible market crash. Since doomsday put options are relatively cheap, you can increase the confidence of holding a long position without significantly increasing the cost.
In addition, regarding Meme coin, CZ expressed his opinion on Twitter today, which roughly means that he is not optimistic about the current situation of Meme coin. I think the fate of Meme coin is closely related to trading platforms such as Binance. If the listing expectations of platforms such as Binance change, the short-term prospects of Meme coin may be greatly affected. Therefore, in the current environment, I suggest that investors can pay more attention to mainstream currencies and consider currency-based resttaking or currency-based income-increasing strategies to achieve a more robust return on investment.
