P talks about AI token investment methodology: narrative judgment, buying and selling points

Editor | Wu Talks about Blockchain

This interview explored in depth the investment trends, market status and future opportunities of AI Agent projects. Analyst defioasis shared his investment experience in on-chain assets, memecoin and AI-related projects, including the rise of key projects (such as GOAT, WorldCoin, Turbo, Pippin, etc.) and the market logic behind them. In addition, investment methodology, position management strategies, signs of cooling of market bubbles, and the importance of industry trend changes and project narratives were analyzed.

Please briefly introduce yourself and review when you started researching AI Agents, when did you start investing in it, and what is your overall return rate so far?

defioasis: Hello everyone, I am defioasis. I am usually interested in on-chain data, on-chain assets and the derived gameplay. Since last year, my main focus has been on playing with things on the chain.

In fact, the research on AI Agents also originated from the mining on the chain. I have been observing the assets on the chain for the past year, but I am not so good at pure Memecoin. I mainly watched it more, and I couldn’t convince myself to play with it with big funds.

The turning point was mainly in late October and early November last year, when GOAT soared to a market value of several hundred million US dollars in a short period of time and Binance launched GOAT futures, which made me rethink this track. Of course, before talking about it, I still want to talk about some AI-related things before GOAT. Before GOAT, there were two relatively large events. One was WorldCoin. Because the co-founder was Sam, the founder of OpenAI, WLD was hyped as OpenAI's meme in Crypto, and FDV once exceeded 100 billion US dollars; the second was Turbo, which claimed to be an AI meme created by GPT. It was also a target that rose 200-300 times in CEX last year. So these two things show that everyone is still very interested in the theme of AI+Crypto and it is worth hyping.

Back to the end of October last year, after GOAT was listed on Binance futures, I quickly thought of WorldCoin and Turbo, and Binance had the ability to lead the track. At that time, there were 4 assets that I thought were very eye-catching, two of which were AI-related, one was ai16z, and the other was ACT; the other two non-AI-related were LUCE and BAN. Now, ai16z seems to be much more powerful than ACT, but the situation at that time was that the founder of ai16z, Shaw, only had nobody, and ai16z was launched with daosfun, and there were some problems with the minting of token contracts, and the volatility was particularly large at that time because the pool was very small. I built a position in ai16z at a market value of 25 million US dollars, which cost about 3,000 US dollars, but the pool was very small at that time. It was once below the market value of 10 million US dollars due to fud, and I didn't dare to increase my position. So I finally chose ACT, which is more stable and has a larger number of holders. ACT can be regarded as my first operation in AI investment. The first transaction was to buy the first 3,000 US dollars of ACT on November 5th. Later, I found that it was available on Bitget, and I also bought more than 10,000 US dollars of ACT at a market value of 22 million US dollars and an average price of 0.022 US dollars. But at that time, I didn’t expect that the position was not enough after buying it one after another, and then it was listed on Binance unexpectedly. I remember that I was still attending a conference in Bangkok at that time, and I was actually quite shocked. After listing on Binance, the highest market value reached 700-800 million US dollars, but I actually sold most of it at 500 million US dollars the next morning after the Binance news, and the rest has not been sold until now.

Later, I also observed that AI Agents is the only track that has gone from a general PumpFun to a vertical track and formed a scale, so I went deeper into this track to invest. After ACT, I found some good targets, such as Pippin, which is also heavily invested and has achieved a return of more than ten times. The overall position used for investment in AI has actually reached 7-8 times since November. It has retreated a bit recently, but it is still about 5 times.

What is your approach to researching AI Agent projects? Can you give an example of an AI Agent project that you have researched the most deeply, so that more people can understand your research methodology?

defioasis: The current AI Agent or AI-related assets are very different from previous AI projects. Now, basically, the assets are fairly launched based on PumpFun-like launches, so the project owner or founder may not have many assets, which may be less than some snipers or big investors. This means that the founder's character, ideas and background are very important. Otherwise, if it is immoral or not real-name, it can be abandoned at any time to open a new one. So my methodology for AI Agent starts with people, whether he really wants to do it, whether he has the ability to do it, and what he can do.

Let me give you an example, let's take the experience of the Pippin project as an example. I first learned about Yohei Nakajima (Nakajima Yohei) from the judges of Solana AI Hakathon on December 11. Solana AI Hakathoh is a hackathon event held by ai16z and Solana. When I looked at the list of judges, I found a Japanese Yohei Nakajima, the founder of Pippin. It was quite interesting to see that he was working on a child-oriented AI Agent, because I had never seen agents with the same positioning. Another aspect is that as a judge, it is obviously more valuable than having to participate in the competition to prove yourself. Later, I further studied Yohei Nakajima and found that he was the founder of BabyAGI, and BabyAGI has more than 20,000 stars on Github. At that time, I searched for some information and found that BabyAGI was quite powerful. As an AGI concept product, it has been cited by many media papers, which also proves its strength. In addition, Yohei Nakajima is a partner of Untapped Capital, and the main Web3 projects he invested in include Pixel, and Pixle is listed on Binance. Overall, Yohei Nakajima, the founder of Pippin, is very powerful in both technology and capital resources. As a person with a real name and a certain prestige, the probability of him casually rugging is much smaller.

At that time, Pippin's market value was about 20 million US dollars, and not many people paid attention to it. The market value of 20 million US dollars also meets my own requirements for purchasing asset targets, because I prefer to buy in the market value range of 10-20-30M. I bought 0.02% of the total amount, which is the maximum single holding limit defined by me, and it cost about 40,000 US dollars. Later, it actually fell below 10 million US dollars, but I bought it with chips at that time. The background and technology of the founder will not change with the fluctuation of prices. So I didn't care after I bought it.

Later, Pippin announced that it would transform into a framework, from a single AI agent to an AI framework, and the valuation increased sharply. After this news, even though the framework has not yet been made, everyone's recognition of the founder's technology and capabilities also made Pippin soar to $300 million. Building a framework means that it may become a split plate. The market pricing for frameworks is the highest in the current AI agents track, and the framework or ecosystem that can form a split plate model has the opportunity to exceed $1 billion in market value.

Which AI Agents projects are promising? What do these projects do? Why are they promising?

defioasis: I am optimistic about many stocks, such as Pippin, which I still hold, but I will not talk about the stocks with higher market value. I usually choose stocks around 20 million US dollars, but there are actually relatively few stocks that I will hold heavily.

There are two main directions at present. One is to pan for gold in the Solana AI hackathon. It has now ended. There are quite a few award-winning projects in it, and they are being screened recently. The other is to bring Virtuals to Solana and cooperate with Jupiter. There will probably be many interesting projects in the future, because Virtuals has proven its success on Base. Of course, I am still watching this.

Here I will mainly talk about some projects that came out of the Solana hackathon. I will share one here, but this is not an investment advice. A project I have been observing recently is AgentiPy, which is actually also an open source framework. It uses Python to connect AI agents to Solana's on-chain applications. According to the roadmap, it may launch an autonomous narrative trading robot in Q1 and launchpad in Q2; the most important thing is that it is mentioned that its token APY will participate as a flywheel. After studying the APY token economics, it is also well designed. Although it is also based on a fair launch similar to PumpFun, it has staked 40% of the chips itself and put them all in Streamflow for two-year linear unlocking, which shows that the team is still determined. The co-founder and CTO of AgentiPy have both been followed by Solana's official Twitter account. Coming out of the Solana hackathon is at least a kind of endorsement. Of course, it is still relatively early overall, and there is a lot of uncertainty. Later, I will also pay more attention to the projects launched on Virtuals after the cross-chain to Solana.

Overall, I think AI is slowly reaching the stage of AI Application. In addition to the framework, I will also keep an eye on AI+ applications, especially AI+DeFi, which is a combination of AI and Crypto native narratives as well as DeFi assets and flywheels. There may be some good opportunities, but it is still in a very early stage and I haven't seen any good targets. I am still observing and haven't bought any new assets recently.

What is your opinion on the current AI Agents track and market status? Do you think the popularity of AI Agents will last for a long time or do you think the bubble has peaked?

defioasis: It has been quite cold recently, but I don’t think it will end. AI is still a relatively sensible thing. AI outside the circle is still iterating and developing rapidly, and is on the rise of technology and capital. This is a more important basic foundation. In fact, many AI targets are actually driven by the outside world, whether it is narrative or talent. Wasn’t Shaw also a marginal person in web2 back then? Now the ai16z he created has become a leader in Crypto AI. Influenced by Shaw, I think there will be more technical talents from traditional industries coming in to do AI. Crypto AI itself is significantly behind the outside world, so any disturbance outside the circle and any major updates will be transmitted to Crypto to form a new narrative and a new sub-track.

From another perspective, from the perspective of the industry, AI Agents is the only one that has developed from a general PumpFun to a vertical track with scale. DeSci may be considered half, and it has also been relatively unpopular recently. Apart from this, there is no other track that can move from a general track to a vertical track, which actually shows that there is a strong demand for AI narratives. AI Agents are relatively unpopular now, which is a cooling of the previous overheating. The current proliferation of single agents has forced everyone to build frameworks and roll, and everyone has gradually become more tired of this kind of thing. Therefore, if AI+ applications, especially AI+Crypto native narratives, can be established, I believe it will lead to a new wave of enthusiasm and new opportunities.

What are your thoughts on investment and trading? What are your tips on opening and exiting positions?

defioasis: The above actually talks about the issue of investment targets, but I think position management is more important. The selection of targets is nothing more than technology, resources and background. Although it is based on PumpFun for fair launch, the research ideas are not much different from VC coins, including technology, resources, background, team, endorsement, etc., as well as analysis of chip structure, blue chip addresses of insider warehouses, etc.

Later, I will focus on the issue of position management. Now a good asset can basically go through three stages: PvP, second stage and listing, but most of them are gone after PvP. I usually play the second stage, and focus on the $10-20-30 million market value range to deploy assets, which I also call the on-chain sweet spot. Because I found that some good assets will probably consolidate and fluctuate at this position after the birth of the highs and falls. I usually only allocate assets in this range. Those that have experienced one or more 70% retracements and are relatively stable in the $10-20-30 million market value will be paid attention to.

When I see a good asset, I will add it to my watchlist, which is divided into S-level, A-level and others. S-level is a coin that can form a model plate. For example, it is a mother coin ecosystem that can profit from the continuous output of sub-coins. For example, as a trading pair asset, sub-coins create a wealth effect, and the mother coin will be needed for transactions to generate continuous demand. This is the same as PumpFun's SOL and Virtuals' mother coin flywheel. Therefore, most of the S-level ones are frameworks, and there is a chance to form an ecosystem or model flywheel. The reason why Pippin can rise is because the market has high expectations for it to be a framework.

The A-level is mainly judged by narrative, background, technology, resources and team (no pattern can be formed). The past background of the founder and the founding team is very important, such as it has a background in the Solana Foundation, a lot of Github stars, has made products that have attracted attention, has a good narrative and has the potential to develop into a niche track, etc.

Generally speaking, both A-level and S-level will build positions, and each time the position is about 2-3000 U. However, I will be more strict in choosing whether to increase the position or increase the position. I need to observe for a period of time and decide whether to increase the position, mainly depending on what the community and developers are doing on a daily basis.

I also have certain standards for the single coin holding of each coin. The maximum single coin holding is 0.2%, which means that the total cost of purchasing around 20 million US dollars is about 40,000 U. Every time I choose to increase my position, I think about the target. If I judge that the target is contrary to the initial purchase logic, I will decisively give up increasing my position, and the remaining positions will be temporarily reserved. If it falls sharply below the box market value range, I will decisively give up increasing my position, and the remaining positions will be temporarily reserved. In principle, I will consider whether to sell at least 10 times, and I will consider it in the market value range of 200-300m.

The purpose of building a position is to ensure that the target is on the car. The purpose of adding positions cautiously and gradually increasing positions in batches and setting a hard cap of 1,200 yuan is to avoid excessive blind confidence and increasing positions all the way so that the funds are trapped in a certain target. Of course, as long as you choose to increase your position, you will believe in your judgment of the target.

What I have mentioned above are some second-stage gameplay. I feel that in the market, whether it is users, Dev or coin issuing factories, they seem to be more familiar with the routines of this second stage. I feel that the second stage is still relatively difficult to play recently. So I am also doing some lottery flow gameplay, that is, PvP. At this time, it is not limited to AI. I have prepared several SOLs, and I buy 0.1 SOL each time, watching the market, monitoring and following some selected addresses, and making a small bet for a big gain. I have felt in the past two days that in the current market, the lottery flow gameplay is easier to make a profit. This may be a probability. With so many plates, as long as the investment is small and dispersed enough, and some high-winning addresses are monitored as an auxiliary, the lottery flow can recover all the previous losses as long as it wins once. In this lack of big market, it will be a good choice. To some extent, it is also a preparation for potential big market. It is very important to always maintain the feeling and mentality of profitability.