X Space Review: A Discussion on Trends and Opportunities to Watch in 2025

With BTC surpassing $100,000 in 2024, the crypto market closed the year perfectly. Looking ahead to 2025, market forecasts are generally more optimistic. AI Agent, DeSci, RWA, Depin, and others are seen as the key tracks to watch in 2025. At the beginning of the new year, BlockBeats invited crypto KOLs to discuss the trends and opportunities worth paying attention to in 2025.

BlockBeats: Let's start with some self-introductions.

Anthony Z: Hello, everyone. I'm Anthony, responsible for the Asia-Pacific region at The Graph Foundation. I've been working at The Graph for about 5 years, entering the industry in 2018, and have around 7 years of experience, making me a "seasoned" participant. I mainly work on non-technical aspects of Web3 data infrastructure and also engage with the technical community. We rarely discuss the market with friends in the industry in the past, but this year, I'd like to give it a try. I also have some AI-related content I'd like to share with everyone.

Sanzhi: I'm Sanzhi from AmberAC, AmberAC being the accelerator of Amber Group. We mainly focus on projects in the AI and Crypto tracks. If Anthony calls himself a "seasoned participant," then I must be the "newbie." Previously, I mainly engaged with developer communities and have been closer to developers. I started exploring market trading in March this year and have recently been studying AI Agent and meme coins.

CM: I'm Chen Mo, mainly engaging in DeFi-related research and project investments. Lately, I've been focusing on the AI agent direction and I'm in the learning phase. I'm delighted to accept the invitation to exchange ideas with everyone.

EO: Hello, everyone. I'm the investment manager at Future Money Group. We mainly invest in the DeFi and AI tracks. We began investing in DeFi in 2020, and this year, we entered a few AI projects. We are currently bullish on the AI agent track, but due to AI computing power and energy issues, the infrastructure is not yet complete, resulting in slower AI agent development. Nevertheless, we remain optimistic about this track and believe that many excellent projects will emerge in the future.

BlockBeats: Teachers, could you please share your best trade in 2024 or your most profound lesson learned?

Anthony Z: I don't trade much. This year, when the price of Bitcoin rose to over $60,000, I increased my position, and when the price of Ethereum was below $3,000, I also bought some. I mainly focus on holding. Originally, some of my friends were quite resistant to cryptocurrency and not very interested, mainly trading stocks and the like. But this year, some newly onboarded friends said they started trading meme coins. I had always been cautious about them, but seeing them indeed rising, I thought this may be because these newcomers have become the new pricers. Our old-timer understanding may no longer apply because we are a bit outdated.

Sanzhi: I remember that in January or February, I didn't quite understand tokens. Back then, I heard that Pepe was good, and a friend of mine also said Pepe was good. Although I didn't quite understand, because both recommended it, I bought Pepe. I didn't buy much, maybe a few hundred dollars, but later the profit tripled. At that time, I only dared to buy 100 dollars worth of any token.

Recently, because I needed to research market trends and hot topics, I started to trade on my own. I previously bought SOLANA, and after that, I directly placed an order for a meme coin on Pump.fun. Although I'm not familiar with these quick in-and-out trades, if I believe in a project, I will do some research. The meme coin I bought has also roughly doubled or tripled. At its lowest point, a few days ago, it was only $0.018, and now it's about $0.1, almost a 10x increase. When it was at $0.02, I enthusiastically recommended to my friends that I got in at $0.05, but later got stuck. I find it quite amusing.

Throughout the entire trading process, I didn't do very well. I tend to lean towards holding tokens long-term because I don't have the time to watch the candlesticks constantly. If the token drops after I bought it, I'll buy more; if it goes up, I'll hold. However, I have a mental expectation; for example, if I think it will increase by 10x, I won't sell if the token doesn't meet my expectations. I can accept it going to zero. I got in at $0.05, but when it dropped to $0.02, I only bought an additional 20 dollars, not fully following my strategy. However, despite the continuous price drop, I didn't sell. Now that it has rebounded, higher than before by one or two times, I still haven't sold and continue to hold long-term. I lean towards being conservative, always needing to validate that my strategy is correct. After conducting extensive research, I will buy, set a trading strategy, and execute it, maintaining a mindset where going to zero doesn't matter.

CM: In recent years, I have been paying close attention to the DeFi track, especially in the last few months of 2024, where DeFi has regained some market attention. Some top assets, such as AAVE and LINK, have also performed well, which are long-term holdings of mine. In the mid to late stages of the bull market, they have shown good performance, which is related to some policy factors, such as Trump's team or projects indirectly holding some DeFi-related tokens.

I have done more research on the DeFi field compared to other tracks. I have always believed that DeFi is one of the few tracks in the cryptocurrency space that can currently change the market structure. Before the emergence of DEXs or lending platforms, users did not have many choices and could only trade and lend on centralized exchanges. Now, many excellent projects have appeared on-chain, providing more options. For example, the recent hype around Hyperliquid in the derivatives track has also received a lot of attention, which is still a trend, so I will continue to focus on this track in the long term. Accordingly, I have not seized many opportunities in hot tracks like Meme. Part of the reason is path dependence, and my energy is also limited. For the AI track, I have bought some top assets, but I have not caught onto other small projects. Tokens such as AI16Z, VIRTUAL, and AIXBT are in my portfolio. I still rely mainly on some past experience.

EO: After experiencing two market fluctuations in 2024, I have some painful experiences to share with everyone. In the fluctuations in April and December, my key takeaway is to always choose strong coins for investment. Although weak coins may be cheap, they are actually not suitable for positioning. These coins may be strong for a day or two, but they will quickly weaken again, and their overall performance will not be good, with significant candlestick fluctuations. Especially when the market is bad, these coins will experience even greater declines. During the market fluctuation in November, I went long on AAVE and DYDX. At that time, I could clearly feel that AAVE had a very strong bottom, and it would not look back with each rise; whereas DYDX was relatively weak. Although I have a soft spot for DYDX, looking at the cycle chart, the performance of weak coins is indeed not very good.

Additionally, some coins that were once strong also need attention, such as PNUTS. When it was rising, I missed the opportunity and tried to bottom fish when it dropped to around $1.2, but I was slapped in the face twice and eventually gave up on this coin. Recently, the price of PNUTS has rebounded to $0.6. Weak coins in the near term need to be vigilant. From my experience, if there is a lot of accumulation at the bottom of a coin, its price will rise faster. So, I still suggest that everyone buy strong coins during a bull market. If a coin is not strong even during a bull market, there is basically no hope.

Another experience is that the market usually gives us some buffer time during a downturn. For example, starting from the flash crash on December 9th, the market's subsequent performance was not strong, and many coins did not perform particularly well. Many people have started to sell off and exit, liquidating their holdings, while large institutions or whales usually do not buy in large quantities at this point because it is not the stage of the main uptrend. During this stage, everyone is advised to be vigilant of risks, set stop-losses to mitigate some risks, otherwise, it will be difficult to stop losses promptly during a crash.

Anthony Z: I want to ask Teacher Chen Mo some questions. After the bull market returns, the DeFi track can be said to be the most mature. Hyperliquid is quite unique, gradually transitioning from behind a DEX to developing Layer1. What is your view on this project?

CM: My personal view is that Hyperliquid actually seized the opportune moment. Many people believe that its project mechanism is not much different from DYDX. DYDX had good data during the early airdrop, but the final result was not good. The derivatives track has always been regarded as gradually being debunked, and many believe that those who trade contracts on centralized exchanges will not come to the chain. Firstly, the learning cost on-chain is high, and secondly, contract traders and on-chain players are two completely different categories of people. It's similar to people who played NFTs in the previous cycle not knowing what tokens are.

I also do not think that Hyperliquid can attract users away from centralized exchanges, but some users may come over. However, the main force is still existing on-chain users, who are already used to interacting on-chain, and Hyperliquid just provides them with a new venue.

Hyperliquid has indeed done better in early market operations and community development than many projects. Although there is FUD in the market, good products always face controversy. Currently, many people doubt whether Hyperliquid can sustain its development because the project's tokens have not been fully distributed, so people may trade based on token expectations. In the future, a clear indicator for the derivatives track is that if the volume does not drop after breaking free from token incentive and the user count does not decrease, then the project is likely to continue.

DYDX's sustainability worsened later on. Once token rewards are reduced, users find it unprofitable, and its sustainability is weakened. You can observe whether the project continues after the token's expected end and if the trading volume and user base remain steady with market development. Many people still prefer on-chain trading, even though there are some mainstream coins on centralized exchanges and lower gas fees. Hyperliquid has potential, but it's not a sure thing that it will succeed. Once it succeeds, it will also be a phenomenal project. Many say Hyperliquid is not decentralized enough, but decentralization is gradually achieved, which is not a significant issue. In general, Hyperliquid is a project worth paying attention to, as it has performed well in seizing market opportunities and timing, but it is still premature to say it is the endpoint.

BlockBeats: AI Agent is a direction that everyone has been aiming for. I would like everyone to discuss how they view the current development of AI and the core focus areas for future development.

Anthony Z: I believe the earliest AI narrative can be traced back to OpenAI's release of ChatGPT. After that, many AI-related crypto projects emerged. At that time, those projects mainly discussed grand narratives, such as decentralized computing power and computing power networks, which were relatively distant from reality. Recently, due to the overall development of the AI industry, the integration of AI and Web3 is inevitable. Taking AIXBT as an example, it combines some macro-level cryptocurrency trading thinking with data from CoinMarketCap and discussions on Twitter to formulate strategies. However, they have not yet integrated specific on-chain data and transaction information, and I hope this is the next development direction. At The Graph, we mainly focus on the infrastructure construction in this regard. In general, most current AI projects are showcasing some practical use cases, and the next step should be a more refined development.

EO: We actually value the AI track very much, and we hope that the future of AI+Crypto will move towards automation. However, currently, AI still requires a lot of time and financial support in terms of infrastructure. Infrastructure is unlikely to see significant improvement in the short term. There are currently some AI agent projects such as automatic Twitter posting, which fall far short of everyone's expectations and imagination of an AI agent. These functions can also be achieved through scripts, which do not qualify as AI agents in my opinion. I think it will be difficult to see a true AI agent in the short term, as infrastructure cooperation is still needed.

Recently, there are two tracks that I am more optimistic about. For example, the AI and DeSci tracks, which are usually financed through token issuance, and after the funds are in place, research and product iterations are carried out. These tracks may currently be in a bubble stage, where everyone is enjoying the dividends brought by new technology, but it is uncertain how long the bubble stage can last. In the future, I will focus more on the underlying technology and infrastructure of DeFi, such as privacy technology. On the application layer side, it is also necessary to see if projects in the Web2 track can develop excellent AI agent projects.

Sanzhi: I recently joined AmberAC, so I will share some insights from a different perspective. Earlier this year, I created an AI Crypto community. We divided the AI track into several levels: from bottom to top, including infrastructure, which consists of computing power and inference training. Then data collection, including processing, labeling, cleaning, and so on. Then the model layer, and finally the application layer. Currently, many AI Agents actually belong to the application layer category. In the AI track, there are several projects worth paying attention to, such as Bittensor, which has innovated in terms of mechanism and has a clever tokenomics design. Additionally, Sahara AI raised about $43 million in the first half of the year. MyShell AI belongs to the application layer, acting as an Agent marketplace platform. Vana is in the data layer, focusing on data privacy protection, involving data and computation. In March-April of last year, AI blew into the Crypto field. In March-April of this year, AI began to develop from various levels such as data and applications. But the real showstopper should be from the AI Meme.

Whether it's infrastructure, models, or data, to attract a large number of users at once, AI is still limited by technology. In that case, it's better to extract the part that can be tokenized directly. Therefore, the AI Agent quickly became popular with the help of Memes. I think this is currently a quite good direction. During a bull market, while the technological aspect is important, what people focus on more is the marketing aspect. People are more prone to FOMO, and at this time, it's difficult to deeply attract their attention with technical content. If marketing is cleverly done, such as using Meme format to attract attention, doing more marketing during a bull market, and focusing more on technology when the market is quiet, it might be more effective. I speculate that the AI Agent will continue to gain popularity. Currently, it is still only in the textual stage, but in the future, it may expand into multimodal.

OnChain Data also has great potential. For example, if an AI Agent interacts with us on Twitter, and we tag an address, it can immediately tell us the on-chain activity of that address. Another example is DataChain, where you can feed the AI Agent some raw data. When we interact extensively with these Agents, a large amount of data is generated. This data is relatively real because it is the result of our interactions. This data is a valuable resource for training AI in the future. Therefore, I believe Data Chain also has a tipping point.

CM: I approach this more from the perspective of Crypto users and market demand. Currently, I think the AI Agent is still in the conceptual hype stage. Everyone knows that the current technology is not yet mature. Compared to traditional AI, the existing AI Agent has not achieved much substantive results. In the cryptocurrency community, imagination is most important. Early DeFi was also driven by imagination. Although DeFi is now one of the few tracks with better implementation, in the early days, it was also driven by imagination. People went through various stages of mining in the early days, fell into numerous pits, but also had many opportunities to get rich quick. This is a very tortuous process, but the premise is that this track requires enough imagination. NFT is also full of imagination, and it has a strong driving force, which can generate high attention.

The AI Agent also has a high ceiling in the traditional AI field. Even though the current technology is not mature, you can imagine the future development potential. Even in the Crypto field, we realize the limitations of current technology, but once traditional AI technology makes a breakthrough, the prospects for the AI Agent will be very broad. It is unlikely to rely on various means to accelerate AI evolution in the Crypto field. The main battlefield for AI still remains in Web2.

The Crypto space has a significant advantage in that it can amplify any effect, especially some positive effects or effects that can make money. For some independent developers, they are fully capable of migrating mature AI technology from the traditional AI field to the blockchain, and the cost of doing so is relatively low. I believe that in the future, there will be many bottom-up opportunities in the AI Agent field, not necessarily all VC-led formal projects.

At this stage, there will be many opportunities for developers and entrepreneurs to enter this wave, which is very important. There was a stage before where everyone was FUDding VC coins because many ordinary entrepreneurs found it hard to participate as the market had already become so intense. However, for AI Agents, the future is promising. Therefore, I believe its impact in the user community will be greater than in other fields, which is my somewhat superficial view. As for how far the technology can go or whether it can change things, from my perspective, it can only be observed and waited upon.

BlockBeats: Existing AI large models, whether in Web2 or Web3, have already trained on all freely available high-quality content on the Internet. So, if we want to continue developing high-quality models, we may need some paid data support. What do you think the competitive landscape will be in the future for AI data acquisition and service protocols?

Anthony Z: This year, we have made many attempts. The Graph has been laying out AI for a long time, and the training of large AI language models requires a large amount of data, which is exactly what The Graph does in terms of data infrastructure. In March of this year, we released a whitepaper introducing our progress in AI. We have a dedicated team working on AI. Two years ago, I met this team, and they proposed leveraging The Graph data to build a tool for querying on-chain data using natural language. For example, you could ask it about Vitalik's account transactions in the past year and a half, and then it would convert this question into The Graph's query language, retrieve real-time data from the chain, and provide answers and data lists. In March last year, the team successfully developed this tool.

To help everyone understand The Graph better, let me give a brief introduction. Many people ask me about the difference between The Graph and Chainlink. Chainlink brings off-chain data onto the chain, such as providing token prices from exchanges to on-chain DEX. The Graph, on the other hand, brings on-chain data off-chain for transaction data analysis or front-end transaction history display, etc. It is a developer tool.

In the traditional Internet, Google is responsible for indexing and filtering Internet data and providing data services to other companies or independent developers. On the other hand, The Graph has over 100 Graph nodes that provide indexing services for on-chain data. Projects like Aave, Uniswap, SushiSwap, among others, utilize The Graph's API. When you make a trade on Uniswap, the displayed trade pair price and trade history data are retrieved through The Graph. When you make a transaction, you are also calling The Graph's API. When these API calls are made, the data is actually sent to The Graph's network. Projects need to pay query fees when accessing data. Therefore, The Graph itself acts as a decentralized data network infrastructure and data marketplace, embodying a decentralized data network architecture.

Running and training AI models only requires adding another set of systems. We have been developing this system over the past year and have released a product called Agentc. Agentc's frontend is similar to ChatGPT; you can chat with it, and it can provide feedback by accessing data. The previously mentioned AIXBT doesn't actually connect to on-chain data, but if combined with The Graph, it could retrieve real-time transaction data for analysis. For example, you could monitor the activities of large wallet addresses or track transactions of meme coins. Through this data analysis, the trading recommendations provided by AIXBT may be more reliable.

What we are doing now is integrating Agent construction, operation, and execution on The Graph. After the release of Agentc, we have been working on this aspect but found that the currently trained LLM model only has an accuracy of just over 63%, which is not sufficient. Therefore, we are developing a new model called Knowledge Graphs & LLMs and creating a new transaction data standard called GRC20. While ERC20 is the standard for issuing tokens on Ethereum and represents a value exchange standard, GRC20 is an information exchange standard that includes relationships between data. In summary, these efforts aim to leverage on-chain data and AI for simple integration, but for complex use cases, errors become more prevalent. We not only need to obtain on-chain data but also need to combine and label the relationships between various data. This is what we are going to do next.

EO: I think AI data is quite crucial. In theory, the advantage of Crypto is being able to obtain data through token incentives that Web2 cannot access. Many projects are also working on this. Currently, most valuable data is controlled by large companies. I believe that projects need to consider how to obtain data from end-users, and this data must be valuable. Secondly, users need to voluntarily contribute this data. Finally, it is crucial how projects use the contributed data. If any of these points are missing, you may not know how to use the data or how it can be passed downstream. If middleware can handle and package the data and pass it downstream, downstream data utilization efficiency can be greatly improved. Additionally, users are very concerned about data breaches, so encryption and privacy protection are also noteworthy areas. I am particularly interested in the field of AI plus FHE and am considering how to effectively avoid user data leakage.

Sanzhi: I believe that a project that performs well in the Crypto and Web3 fields must innovate around asset issuance to mobilize synergies. The popularity of AI agents is due to their relevance to asset issuance. Previously, we used smart contracts to generate a token, but now we can use an AI agent to bind and generate a token. Additionally, data is also crucial. In the past, we did not have enough awareness of our data sovereignty, and our data has never been directly monetized. The previous data monetization logic involved selling data to an application, which would then sell it to advertisers, going through many intermediaries without direct monetization. However, the data monetization model has changed. There are two main new ways: using data to feed models for monetization and issuing data as an asset. For example, data from a tumor patient can be packaged as an asset, offering significant potential.

When trading, we not only look at information on Twitter but also examine transaction data in GMGN. If there were an Agent to integrate this information, providing a multifaceted view and objective data, such an Agent would be very valuable. If an AI agent could help me process all on-chain data, offering intuitive insights, I would be willing to pay for such a product because it could save a lot of time. Transforming such a product into an asset issuance also holds great imaginative space.

While individuals may not pay much attention to privacy protection, large companies are very attentive to it due to privacy protection regulations. With increasingly standardized privacy protection laws, there is vast imaginative space for data sovereignty, data privacy, and data assetization. Decentralization is interrelated with ownership. If the aspects related to collaboration and ownership can be combined, there may be significant development potential. As for hardcore AI projects, further observation is needed regarding their development.

CM: My perspective aligns with previous guests and mainly focuses on data quality issues. Currently, much data is held by tech giants, while user-side data acquisition can be achieved through token incentive mechanisms. Many projects are developing this model, which I believe has great potential. Previously, people were somewhat averse to the play-to-earn model because its intrinsic value creation was relatively weak. However, for AI data labeling, it is quite valuable, and this data can be profitable when transferred to AI products. Token incentives in AI data acquisition may be effective, but each part of the process still requires sound mechanisms.

BlockBeats: Finally, let's discuss the main focuses for 2025 and the betting opportunities in the DeFi space. For 2025, what specific operational plans do you have?

Anthony Z: This year, I will pay particular attention to the RWA track because it is an area supported by policies that can bring in new money. Although RWA is an old concept, it has received support from the Trump administration and related interest groups, which will unlock a lot of liquidity. I will focus on the top projects in this track. If you are in Canada or Asia and want to invest in U.S. assets but don't have the right channels, if assets can be tokenized on-chain and have relevant policy support, this will bring many opportunities. This kind of innovation is also beneficial for the U.S. itself and will create new channels and opportunities for the international financial market.

CM: Regarding DeFi, I roughly divide it into two directions. First, from a policy trend perspective, I am focusing on RWA-related content. Many traditional institutions such as BlackRock are also issuing tokenized funds on-chain, and there are more operational and liquid scenarios for these assets than we might imagine, which may have greater scalability. So, RWA is not just about moving off-chain assets to on-chain but has a richer range of applications. We can call it RWA 2.0. The general direction is still guided by compliance and policy, especially since the Trump administration took office. Additionally, DeFi has been developing for several years, from starting with just Uniswap and SushiSwap competing to later introducing lending and various composability.

Now I believe that DeFi is poised for the next stage of opportunity. First is chain abstraction. If you pay attention to the OP Super Chain, it will be easier to understand. In the Ethereum ecosystem, OP's horizontal expansion almost completely surpasses Arbitrum, despite Arbitrum's different strategy. But in terms of horizontal expansion indicators, most Layer 2 and Rollup solutions are now based on OP, including recently hyped chains like Base and Worldcoin, all adopting OP's underlying technology to realize their Super Chain plans. In the future, Super Chain will form a natural state of chain abstraction. Asset transfers between Super Chains, inter-chain contract calls, and protocol communication will become very smooth. This is also a direction of DeFi evolution.

Bringing RWA assets on-chain will definitely lead to one certain thing: an increase in on-chain assets. After the asset increase, whether as underlying collateral to support other assets or as a circulating asset, the most direct impact is on the DEX and lending tracks. The revenue in these two areas is most pronounced and can also be developed in Super Chains and chain abstraction. Secondly, when discussing the development of traditional DeFi, it will certainly usher in a new stage. Trump's administration has provided policy support for crypto, and I believe there will be significant growth in DeFi by 2025.

Sanzhi: In the long term, from an application layer perspective, I personally will pay attention to Worldcoin. Previously, after contacting the Worldcoin foundation team, I learned that they are currently very focused on compliance and adopt a top-down strategy because they need to use devices to collect personal data, leading to more interaction with governments. Worldcoin's infrastructure performance is currently not very outstanding, but they have a large user base. If you download their app, you will find that it is mainly peer-to-peer interaction, without forming a connected network, and there is no chat function inside. However, they plan to release this feature in the future.

From a track perspective, I mainly focus on AI and BTCFi. The reason for this focus is that after the ETF is approved, many large institutions will have a demand for BTCFi. The third aspect is AI. I believe that the current bull market has already started but there is no concept of breaking the circle yet. I think the concept of breaking the circle may arise in IP or AI Agent because even outsiders can understand these two areas. Outsiders may not understand jargon like RWA or chain abstraction. I initially thought that IP was an important direction, but now I find that AI Agent is something outsiders can also understand.

In addition, the Ethereum ecosystem also has tremendous breakout potential, such as OP Stack and OP Superchain. Ethereum's Layer2 is not like other new public chains; Ethereum's Layer2 may develop into dedicated chains, such as payment chains, RWA chains, or specialized identity chains, which can drive the breakout of more applications. The Base ecosystem also has great potential. On one hand, this may be because of U.S. policy support, and on the other hand, the collaboration between Base and Warpcast is quite interesting.

EO: The main areas I will focus on in 2025 are DeFi, AI, and DePin tracks. Our fund mainly started with DePin, and we are very interested in this track. When many traditional funds enter this field, they often do not know what projects to invest in, but if there are specific physical assets, they will be more interested. For example, the Hivemapper project, the team has earned a lot of money by selling cameras, and we are quite satisfied with their revenue and have high expectations.

These kinds of projects can indeed attract new users and have real demand. Hivemapper's map already covers 20% to 30% of the global area, and through a peer-to-peer token reward mechanism, it encourages users to collect new maps. Therefore, I believe that these projects can attract traditional funds and have real demand. Only DePin can connect with traditional businesses. The DePin track has not fully erupted yet. If you are interested, you can pay attention to it. The future potential is quite significant, and I look forward to more projects erupting in the DePin track this year.

Another track is the DeFi track, which is a track that traditional Crypto partners tend to favor. This is because Crypto's early days relied mainly on financial innovation. In fact, I started participating in the DeFi cycle in 2020, and I have a deep affection for this track and am quite optimistic about it. It is a track that truly has the ability to generate value in the Crypto field, where both large and small projects can make some money. Another track is the AI sector. The AI sector is currently in a very hot phase. Although the returns may not be as stable as DeFi or DeFi 2.0, I believe its wealth effect should be the largest this year.

Space Link: https://x.com/i/spaces/1ZkKzRblznyKv