This interview delves into the investment trends, market landscape, and future opportunities within AI Agent projects. Analyst defioasis shares his investment experience in on-chain assets, memecoins, and AI-related projects, including the rise of key projects such as GOAT, WorldCoin, Turbo, and Pippin, along with the market logic behind them. Additionally, the discussion covers investment methodologies, position management strategies, signs of market bubble cooling, as well as the evolution of industry trends and the importance of project narratives.

Note: The views expressed by the guest do not represent those of WuBlockchain, and WuBlockchain does not endorse any products or tokens. Readers are advised to strictly comply with local laws and regulations.

Audio transcription by GPT, errors may exist. Please listen to the full podcast:

YouTube:

Spotify: https://creators.spotify.com/pod/show/7qfkmlvhrl8/episodes/EP-54--AI-e2ub8vo

Please briefly introduce yourself and review when you started researching AI Agents, when you officially started investing, and what your overall return rate is now

defioasis: Hello everyone, I’m defioasis. I’m generally interested in on-chain data, on-chain assets, and the various plays derived from them. Since last year, my main focus has been on exploring on-chain projects.

Actually, my research into AI Agents also stems from my work with on-chain mining. I’ve been monitoring on-chain assets since last year, but I’m not particularly skilled with pure Memecoins. I mostly just observe, without feeling convinced enough to deploy significant capital.

The turning point came in late October to early November of last year, when GOAT’s market cap surged to hundreds of millions in a very short period, and Binance launched GOAT futures. This made me rethink the sector. But before discussing that, let me briefly mention a few AI-related events before GOAT.

Before GOAT, two major events happened. The first was WorldCoin, which was co-founded by OpenAI’s Sam, and WLD was considered OpenAI’s meme in the crypto space, with its FDV briefly surpassing $100 billion. The second was Turbo, a so-called AI meme created by GPT, which also saw a 200–300x increase on CEX last year. From these events, it was clear that there was a lot of interest in AI+Crypto as a theme, and it was worth speculating on.

Now, back to late October last year. After GOAT launched on Binance futures, I immediately thought of WorldCoin and Turbo, and Binance’s ability to lead the sector. At that time, there were four assets that caught my attention, two of which were AI-related: ai16z and ACT. The other two, not related to AI, were LUCE and BAN. In hindsight, ai16z seems to have outperformed ACT by a wide margin, but at that time, ai16z’s founder, Shaw, was still unknown, and ai16z had some issues with its DAO and token contract, plus the pool was quite small and highly volatile.

I initially entered ai16z at a market cap of $25 million, investing around $3,000. However, due to the small pool and the FUD, it briefly fell below a $10 million market cap, and I didn’t dare to add more. So, I decided to focus on ACT, which was more stable with a larger holder base. ACT was my first major AI-related investment. On November 5th, I bought my first $3,000 worth of ACT, and later I found it on Bitget. I continued buying at an average price of $0.022, totaling over $10,000 at a market cap of $22 million. At that time, I hadn’t fully filled my position, and unexpectedly, it got listed on Binance. I remember I was attending a conference in Bangkok, and I was quite shocked. After it was listed, the market cap peaked at $700–800 million, but I sold most of my position the next morning when the market cap was around $500 million. The rest I still haven’t sold.

Later, I observed that AI Agents was the only sector that transitioned from general-purpose pumps to a more vertical and scaled industry. So, I deepened my involvement in this sector. After ACT, I also found some other promising assets, such as Pippin, where I took a heavy position and achieved over 10x returns. My overall investment in AI has grown 7–8 times since November, though it has pulled back recently, still around 5x.

What is your method for researching AI Agent projects? Can you provide an example of an AI Agent project that you’ve researched in depth, to help others understand your methodology?

defioasis: The current AI Agent or AI-related assets are quite different from past AI projects. Most of them are launched as fair launch assets, often based on Pump Fun mechanisms. This means that the founders or project teams might not hold much of the assets — sometimes even less than some snipers or large investors. This makes the founder’s character, philosophy, and background incredibly important, because without these, there’s a risk that the project could be abandoned for a new one. So, my methodology for researching AI Agents starts with the person behind it. Do they really want to build something? Do they have the capability to do it? And how successful can it be?

I can use the Pippin project as an example to explain my approach. I first learned about Pippin on December 11th through the judges of the Solana AI Hackathon, which was organized by ai16z and Solana. When I looked at the list of judges, I noticed a Japanese name: Yohei Nakajima, who is the founder of Pippin. I found it interesting that he was working on a child-oriented AI Agent, as I hadn’t seen similar projects targeting that niche. Additionally, being a judge rather than a participant obviously made him stand out, as judges are generally seen as having higher credibility.

Later, I dug deeper into Yohei Nakajima and discovered that he is also the founder of BabyAGI. BabyAGI has over 20,000 stars on its GitHub, and I found that BabyAGI is quite a remarkable AGI concept product, often cited by media and academic papers, proving its strength. On top of that, Yohei Nakajima is a partner at Untapped Capital, which has invested in several prominent Web3 projects, including Pixel, which later listed on Binance. So, when you look at Pippin’s founder, Yohei Nakajima, he has both technical expertise and significant capital resources, and as a real-name figure with a solid reputation, the likelihood of him rug-pulling is much lower.

At that time, Pippin had a market cap of around $20 million, and it wasn’t really on many people’s radar. This market cap fit my usual investment strategy, as I tend to buy assets in the $10–30 million range. I gradually bought about 0.2% of the total supply, which is the upper limit I define for any single position, spending around $40,000. The price did drop below $10 million later. The founder’s background and technical capabilities don’t change with market fluctuations. After I completed my position, I didn’t pay too much attention to the day-to-day price movements.

Later, Pippin announced its pivot from a single AI Agent to an AI framework, which caused its valuation to skyrocket. Even though the framework isn’t fully developed yet, the recognition of the founder’s technical abilities and resources pushed Pippin’s market cap to $300 million. Transitioning to a framework could mean it becomes a platform with splitting opportunities, and the market tends to assign the highest valuation to AI frameworks in the current AI Agent space. A framework or ecosystem that can split into different parts has the potential to break the $1 billion market cap threshold.

Which AI Agent projects are you optimistic about? What do these projects do, and why are you optimistic about them?

defioasis: There are quite a few projects I’m optimistic about, like Pippin, which I still hold, but its market cap has gotten pretty high, so I won’t focus on that. Generally, I tend to make purchases around the $20 million market cap range, but I don’t usually go heavy on many assets.

Currently, I’m focusing on two main directions. One is gold mining from the Solana AI hackathon, which has already ended. There were quite a few prize-winning projects, and I’m currently reviewing them. The other is Virtuals on Solana, in collaboration with Jupiter. I expect there will be some interesting projects coming out of that because Virtuals has already proven its success on Base. I’m still watching this space closely.

Here, I’ll mainly talk about a few projects that came out of the Solana hackathon. One I’m currently watching is AgentiPy, but this is not investment advice. AgentiPy is working on an open-source framework that connects AI agents to Solana-based applications using Python. According to its roadmap, it plans to launch an autonomous narrative trading bot in Q1, and a launchpad in Q2. What’s really interesting is that the project’s token, APY, will serve as a flywheel within the ecosystem. I’ve looked into the tokenomics of APY, and the design is quite solid. Even though it’s based on a pump-fun fair launch model, the team has locked 40% of the tokens and placed them in Streamflow for a two-year linear unlock, which shows some level of commitment. AgentiPy’s co-founders and CTO have also been featured in Solana’s official promotions. Coming out of the Solana hackathon, at least there’s some backing there. Of course, it’s still early stage, and there’s a lot of uncertainty. I’ll continue keeping an eye on it, along with projects that will be launched on Solana through Virtuals.

If I look at the broader picture, I think AI is slowly moving into the AI application stage. Beyond frameworks, I’ll also be watching AI+ applications, especially AI+DeFi. This is where AI and Crypto narratives, along with DeFi assets and flywheels, can come together. There might be some good opportunities here, but it’s still early, and I haven’t found any solid projects yet. For now, I’m just observing and haven’t picked up any new assets recently.

What’s your view on the current AI Agents sector and market status? Do you think the hype around AI Agents will last, or do you believe the bubble has peaked?

defioasis: It’s definitely cooled off recently, but I don’t think it’s over yet. AI still makes a lot of sense, and AI outside of crypto is still rapidly iterating and developing, with strong momentum in both technology and capital. That’s the key foundation. In fact, many AI projects are driven by external forces, whether it’s narrative or talent. Remember, Shaw wasn’t a prominent figure in Web2, but now his creation, ai16z, is one of the leaders in Crypto AI. I believe we’ll see more traditional industry tech talent moving into AI because of people like Shaw. Crypto AI itself is significantly lagging behind the broader industry, so any big updates or shifts outside the crypto space will inevitably influence the crypto world, creating new narratives and sub-sectors.

From another perspective, within the crypto space, AI Agents are the only sector that has transitioned from general-purpose pump fun narratives to becoming a vertical market with real scale. DeSci might be half of that, but it’s cooling off as well. Beyond that, there aren’t any other sectors that have made that shift from general to vertical. This indicates there’s a strong demand for AI narratives. Right now, AI Agents are cooling down due to the previous overhype. There’s an oversaturation of individual agents, which is pushing everyone to move towards frameworks, causing fatigue with this kind of thing.

That being said, if AI+ applications, especially AI+Crypto native narratives, can come together, I believe they’ll spark a new wave of excitement and open up new opportunities.

What are your insights on investing and trading? Do you have any tips for position building and exit strategies?

defioasis: We’ve already talked about picking assets, but I actually think position management is much more important. When choosing assets, it’s all about the technology, resources, and background. Even though many of these projects are based on pump fun fair launches, the research approach has become similar to how VCs approach investments — looking at technology, resources, background, the team, and endorsements, as well as analyzing tokenomics, ownership structure, whale addresses, etc.

Now, I’ll focus on position management. A decent asset typically goes through three stages: PvP, the second stage, and Top CEX listing. However, most assets don’t make it past the PvP stage. I usually focus on the second stage, targeting assets in the $10–30 million market cap range, which I refer to as the “sweet spot” of on-chain investments. I’ve found that many good assets tend to stabilize and consolidate in this range after their initial price surge and pullback. I usually only allocate in this price range. I pay attention to assets that have experienced a 70%+ retracement at least once, and have since stabilized in the $10–30 million market cap range.

When I spot a good asset, I add it to my watchlist and categorize it into S-tier, A-tier, or others. S-tier assets are those that can form a structured ecosystem, like a parent coin benefiting from the constant issuance of child tokens. For example, coins like SOL from the Pump Fun narrative or the parent token of Virtuals follow this type of flywheel effect. These S-tier projects are usually frameworks with the potential to build ecosystems or generate continuous demand. Pippin, for example, rose because the market had high expectations for its framework development.

A-tier assets are judged based on narrative, background, technology, resources, and team. These don’t form ecosystems but have strong fundamentals. The founder’s background is key here. For instance, whether they come from the Solana Foundation, have a high-profile GitHub, have worked on notable products, or have a compelling narrative that could lead to a niche sub-sector.

Typically, I build positions in both S-tier and A-tier assets. For each position, I’ll usually invest around 2–3k USD. However, I’m more strict when it comes to deciding whether to add to a position or make it a heavier allocation. I observe the project for some time to decide if it’s worth adding more, mainly focusing on community activity and what developers are working on.

I also set strict rules for the maximum holding size of any single asset. My limit is 0.2% of the total supply, meaning I typically buy around $40k worth of an asset when the market cap is around $20 million. Every time I consider adding to a position, I reassess whether the asset still aligns with my original investment thesis. If it deviates from that logic, I will stop adding to the position and leave the remaining holdings as-is. If the price drops significantly below the established market cap range, I will also stop adding to the position but hold the current allocation. As a rule, I’ll only consider selling when an asset has reached around 10x returns, typically when its market cap is around $200–300 million.

The goal of building positions is often just to ensure you’re on the ride, but when it comes to adding more, I’m cautious and prefer to scale in gradually, with a hard cap on any single asset to avoid getting too overconfident and getting stuck in one asset. Of course, if I do decide to add more, I believe in my initial judgment about the project.

What I’ve described above is mostly for second-stage plays. I feel like the market — whether users, devs, or token factories — has gotten pretty familiar with second-stage strategies, and lately, it’s been harder to play in this space. So, now I’m also experimenting with “lottery flow” plays, or PvP. At this point, I’m not limiting myself to AI anymore and have prepared a few SOL for small buy-ins, around 0.1 SOL each. I’m monitoring the market, tracking high-probability addresses, and making small, high-leverage bets. I’ve found that, given the current market conditions, lottery-style plays are a more profitable approach. It’s more about probabilities — there are so many projects out there, and as long as you keep your bets small and diversified, and monitor high-win-rate addresses, hitting even one big win can recover all previous losses. In the current market with little big action, this could be a great strategy, and it also helps me prepare for a potential market uptrend. Keeping a consistent profit mindset is key.

Follow us
Twitter: https://twitter.com/WuBlockchain
Telegram: https://t.me/wublockchainenglish