Interviewer: Wu Yue, Xian Ran
Interviewee: DefiOasis
This article is a text record of our interview with DefiOasis, an AI Agent blue chip hunter on Twitter, on January 26. DefiOasis' main business is the author/analyst of Wu Blockchain. He usually focuses on on-chain data and investment research analysis. He has sniped multiple blue chip targets during the hot period of the AI Agent sector, and grasped ai16z, ACT, PIPPIN, TURBO and other targets with 10 times or even dozens of times. As of the day of the interview, he has achieved an overall return of 7~8 times.
In this interview, DefiOasis briefly introduced its investment experience, investment research methodology, and position building methods in the AI Agent track, and systematically told everyone about its valuable experience and insights. It has a strong reference value for investors and even practitioners. Everyone is welcome to read it carefully.
1. Wu Yue: Can you briefly review when you started investing in AI Agent-related tokens?
Defioasis: I am usually interested in this kind of on-chain data analysis and on-chain asset-related gameplay. Since last year, my main focus has actually been on the on-chain. The motivation for studying AI Agent actually comes from my long-term exploration of on-chain assets. I have been observing on-chain assets since last year, but I usually watch more and invest less real money, mainly because I feel that on-chain assets, especially memecoin, cannot convince me to invest a lot of money.
But the turning point was in late October and early November last year, when a phenomenal project Goat created a narrative of AI coin issuance. $GOAT soared to a market value of several hundred million US dollars in a short period of time, and then Binance launched its spot and futures. This incident made me rethink the AI-related track. Of course, there were other things that influenced me before that. One was Worldcoin, whose issuer was Sam Altman, co-founder of OpenAI. Under the narrative of OpenAI's Memecoin, Worldcoin's FDV quickly reached as much as 100 billion US dollars.
The second one is $TURBO, a memecoin created by AI, which has increased by 200~300 times in CEX. From these two things, I feel the strong interest of everyone in the theme of AI+Crypto, and it also makes me sure that the targets related to AI Agent are worth ambush.
So after $GOAT was listed on Binance, I quickly saw the potential of the AI Agent track and felt that Binance had the ability to lead this track, so I began to spend time looking for some AI-related targets. At that time, several targets caught my eye, one was ai16z and the other was ACT.
From now on, ai16z seems to be much more powerful than ACT, but at that time, Shawn, the founder of ai16z, was not very famous, and the way he issued tokens, including the token contract, also had problems. Since $ai16z was more volatile and less popular, I also opened a position in ai16z at that time, but the position was not large. In addition, $ai16z once fell below the market value of 10 million US dollars, so I didn’t dare to buy it later. In the end, I invested heavily in ACT, which is more stable and has a larger number of holders.
The first time I bought $ACT was probably in early October last year. It was my first official position in the AI+Crypto track. I bought about $3,000. Later, I bought ACT several times after the market value reached 20 million US dollars. The average price was about 0.022 US dollars, and I bought more than 10,000 US dollars. What happened afterwards was very unexpected. ACT was listed on Binance. I remember that day I was still in Bangkok for a conference and was quite shocked when I learned the news. After $ACT was listed on Binance, the FDV reached as high as 700 to 800 million US dollars, and I sold a large part of the tokens on the second day of ACT listing on Binance, and I still hold the remaining ones.
Later, I did observe that AI Agent can be said to be the only track that has developed from a general-purpose Pump.fun to a vertical scale, and then I began to delve into this track to do some investment research. After ACT, I also gradually grasped some good targets, such as PIPPIN, which I once held a large position in and achieved a return of more than 10 times. In the end, my overall investment return in the AI Agent track was about 7~8 times from October to a few days ago. It has retreated after the recent market crash, but it is still about 5 times the return.
2. Wu Yue: So what is your methodology for studying AI Agents? Can you give some specific examples to help everyone understand your investment research methodology?
defioasis: Many of today's AI-related assets are very different from some previous AI projects. Recently, people have been paying more attention to assets based on fair launches such as Pump.fun. The assets held by the project parties or founders may not be as many as those held by some big investors. In this case, it is necessary to pay attention to whether the project parties have long-termism and whether their background conditions can support the project. If the founder has no morals, the project may be abandoned or a new one may be opened.
So my methodology for AI Agent is to first examine the project from the perspective of people, to see if the founder really wants to make the project a success and whether he has the ability to do it well. I can give an example in this regard. Take PIPPIN just now. In December, Solana and ai16z held an AI hackathon together. I noticed a Japanese man named Nakajima Yohei among the judges of the hackathon. I learned that this person was working on an AI Agent project, which is PIPPIN. It is a child-oriented AI Agent project. I think it is quite interesting because I have never seen an Agent project with the same positioning. On the other hand, Nakajima Yohei is also a judge of this hackathon. I think that as a judge, he has to participate in the competition and prove himself. Obviously, this judge has higher gold content.
Later, I further researched Nakajima Yohei and found that he is actually the founder of baby agi, which has more than 20,000 stars on Github. I searched for more information and found that BABY AGI is quite impressive. As a product of the AGI concept, it has been mentioned by many media and papers. I think this can actually prove the strength of this person. In addition, this person is also a partner of a VC. They have invested in some Web3 projects, some of which have been listed on top large firms.
So far, I think that overall, the founder of PIPPIN is very powerful in terms of technology, capital operation, and backing resources, and is a relatively prestigious person. As a relatively prestigious person, and a real-name person, the probability of him rugging or abandoning the market is much smaller.
At that time, I saw that the market value of PIPPN was about 20 million US dollars, and not many people around me paid attention to it. It was a target that had not yet been discovered, which was more suitable for me. I tend to buy tokens in the range of 10-20 million US dollars FDV. Later, I spent a total of more than 40,000 US dollars to buy 0.2% of the total PIPPIN. Later, PIPPIN's FDV fell below 10 million US dollars, but I did not abandon the chips because I felt that the founder of this project had a good background and technology, and the fundamentals were still solid. Short-term price fluctuations would not affect the long-term value of the project itself.
I didn't care about these holdings for a long time, but then PIPPIN took a turn, transforming from a single agent to an AI framework, which caused its valuation to rise sharply. After this news came out, even though there was no mature development framework, everyone recognized the founder and the project itself, and soon PIPPIN's FDV soared to 200-300 million US dollars.
Building an AI development framework actually means a split-up. The market price for this type of framework is the highest in the AI track, so I thought this project was good at the time and held on to it without hesitation. When the FDV reached 300 million, I sold more than half of it, and the rest has not been touched until now. I think that when this framework really comes out and forms a split model, it may reach a FDV of 1 billion US dollars or even higher.
3. Wu Yue: Which AI Agent-related projects do you think are promising? And why?
defioasis: Actually, there are still many promising projects, such as PIPPIN, which I still hold. I usually choose FDV with a value of around 20 million US dollars, and there are other aspects. I used to go to the AI hackathon of Solana ecosystem to seek gold. There are many award-winning projects in it. I have been screening and studying their fundamentals recently. Another one is that Virtual is about to join the Solana ecosystem, and some interesting projects may be developed later. Virtual has proved its success on the Base chain, so it is worth paying attention to.
Next, let me talk about some projects that came out of the Solana hacker's day. I'll share one here, but it's not an investment advice. Recently, I've observed a project called Agentipy, which is also an open source AI Agent framework. The core is to use Python to connect AI Agent to Solana's on-chain application. Its roadmap shows that it may launch an autonomous narrative trading robot in Q1 and LaunchPad in Q2. The most important thing is that Agentipy mentioned that it would participate in it by using its token APY as a flywheel. Then I also studied its token APY and thought the design was pretty good, although it was also based on fair launch.
But the team behind it has invested 40% themselves, and all of them will be linearly unlocked for two years. I think that being able to make such a move shows that the team is still determined. Agentipy's co-founder and CEO have been followed by Solana's official Twitter account. In addition, this project came out of Solana's hackathon, so at least it has endorsement and is relatively reliable. Of course, this project is still in its early stages and has a lot of uncertainty. I will also pay more attention to projects that are cross-chained to Virtual, as there are many opportunities in this area.
If we look at the whole picture, the AI+Crypto track has been slowly transitioning to the AI Application stage. After the explosion of various development frameworks, we should pay more attention to AI+Application projects. I think AI+DeFi is also good, which is the native narrative of AI+Crypto and DeFi assets and Framework. There may be some good opportunities, but this is still in a very early stage and I don’t see any good targets. I am still keeping a wait-and-see attitude and haven’t bought any new assets recently. That’s probably it.
4. Wu Yue: What do you think of the current AI Agent track and its market status? How long do you think the popularity of AI Agent will last? Has the bubble peaked?
defioasis: It is indeed quite cold recently, but I think the opportunity of AI Agent will not end, because AI-related concepts are relatively sensible, and AI products outside the Crypto circle are still iterating rapidly, and are on the rise in terms of technology and capital. I think this is the more important fundamental. In fact, many AI targets are driven by factors outside Crypto, such as Shawn, the founder of ai16z. He was probably a marginal person in Web2 at the time, and seized the opportunity in Crypto to create ai16z, becoming a leader in the AI+Crypto track. I believe that under the influence of Shawn, more talents from Web2 and traditional industries will enter Crypto to do business.
In addition, I think the AI projects in the Crypto industry lag behind the traditional industries. Many movements outside the industry, such as the big moves of AI giants, will be transmitted to the Crypto industry, forming new narratives and new sub-tracks. In addition, AI Agent is currently the only vertical track that has transformed from a general-purpose Pump.fun to a scalable one.
Apart from AI Agent, there are not many projects that can move beyond the pure memecoin. Another project that barely made it out might be Desci, but it is also very unpopular now. Apart from this, no other race track can change from general to vertical, which in itself shows that there is a real demand behind the market hype about AI Agent narratives. The overall market of AI Agent has now fallen sharply, mainly because it was a bit overheated before, and now single Agents are very rampant, which has forced everyone to roll up development frameworks and gradually become tired. The future may be the era of AI plus applications, AI plus some Crypto native narratives. If my point of view is valid, there will still be some new crazes and new opportunities.
5. Wu Yue: The last question is actually about investment and trading experience, such as your position building and exit techniques. Are you willing to share them with everyone?
defioasis: We have been talking about investment targets before, but I think position management may be more important, because the selection of targets is nothing more than looking at the technical resources and background of the project. Although it is now based on Pump.fun for fair launch, I think the research ideas are not much different from VC coins, which is nothing more than looking at the technical resource background, team endorsement, chip structure analysis, rat warehouse and blue chip address, etc. I will still focus on the issue of position management later.
Assuming there is a good asset, it will generally go through three stages: PVP, second-stage consolidation, and listing. Most projects fall into decline during the PVP stage, and may disappear in two or three days. My previous position-building style mainly targets assets that have experienced second-stage consolidation. For example, I choose targets with FDV in the range of 10 to 30 million US dollars to lay out, which I usually call the "on-chain sweet spot".
I have found some good targets before, which will fall back to this position after hitting a high in the first two days, and then undergo some consolidation and shock. I usually only make some position allocations at this stage, especially those that have experienced one or more 70% retracements, and then stabilized around 10-20 million US dollars, I will pay special attention to them.
If I see a good asset, I will first add it to my Watchlist, and then rate these projects. I will probably categorize them into S-level, A-level, or other levels. S-level tokens are those that can form a model disk. For example, as a mother coin ecosystem, it can continuously produce sub-coins to make profits. Sub-coins can continuously create wealth effects. People need to buy mother coins to trade, thereby continuously generating demand for mother coins. This is actually very similar to the mother coin flywheels of Pump.fun, Sol, and Virtual.
Therefore, most of the S-rated projects are framework-based projects, because they have the opportunity to form an ecosystem or a flywheel. Why did PIPPIN rise? It is mainly because the market has high expectations for its framework. As for the projects rated A, they are mainly judged based on indicators such as narrative background, technical resources, and team. For example, if the founding team has a good background in the past, has the background of the Solana Foundation, or the founder has made some products that have attracted everyone's attention and may develop into a new track, I may give it an A-level. Anyway, I will add it to the Watchlist for observation first.
Generally speaking, I will build a position as long as it is an S-level or A-level target. Each time I build a position is about 2,000 to 3,000 US dollars. However, I will be more cautious about whether to increase my position or increase my position. I need to spend more time observing to decide whether to increase my position. I need to go to the project community to see what the developers are doing every day.
After setting such a standard, I also have certain restrictions on the holding of each token. For example, I only hold a maximum of 0.02% of a single token. If the FDV is around 20 million US dollars, it will cost about 40,000 US dollars to buy 0.02%. In fact, every time I add a position, I think about the target. If I feel that the target is contrary to the initial purchase logic, I will decisively give up adding positions. Or when the target falls sharply below the range of 10 to 20 million US dollars, I will also decisively give up adding positions.
In principle, I will get at least 10 times the price before selling, which is about 200 to 300 million US dollars of FDV. In fact, in the final analysis, the purpose of building a position is to ensure that you are on the right track. You should be cautious when adding positions and gradually increase your position. This includes why a hard cap of 0.02% is set to avoid excessive blind confidence and increasing positions all the way, which will trap you in one target.
Of course, as long as I increase my position, I will believe in my judgment of the target. But what I said above are some ideas for building positions in the second stage of consolidation. I feel that everyone in the market is more familiar with the second stage routine, but if you want to hold, it is still difficult to play, so I am actually doing some lottery flow play, which is commonly known as PVP, but at this time it is no longer limited to AI, as long as it is an on-chain asset, it will be involved.
In the absence of a big market at the moment, I think this may be a good choice. To some extent, it is also a preparation for potential big market in the future. I feel that it is very important to always maintain this profitable mentality. This is probably my situation.
Brumaire: It sounds like you've gained a lot today. Thank you very much for joining us for this interview.
