P generation's approach to AI token investment: narrative assessment, entry and exit points
TechFlow Selected TechFlow Selected
P generation's approach to AI token investment: narrative assessment, entry and exit points
In-depth exploration of investment trends, market status, and future opportunities in AI Agent projects.
Navigating Web3 tides with focused insightsAuthor: Wu Shuo Blockchain
This interview delves into investment trends, market conditions, and future opportunities in AI Agent projects. Analyst defioasis shares his experience investing in on-chain assets, memecoins, and AI-related projects, including the rise of key projects (such as GOAT, WorldCoin, Turbo, Pippin) and the underlying market logic. It also analyzes investment methodologies, position management strategies, signs of cooling market bubbles, industry trend shifts, and the importance of project narratives.
Please briefly introduce yourself, and review when you started researching AI Agents, when you officially began investing, and your overall return rate so far?
defioasis:
Hi everyone, I'm defioasis. I'm generally interested in on-chain data, on-chain assets, and derivative玩法. Since last year, my main focus has been on on-chain activities.
Actually, my research into AI Agents originated from on-chain exploration. I've been observing on-chain assets since last year, but pure memecoins weren't really my strength—I mostly just watched without being convinced enough to deploy large capital.
The turning point came at the end of October and beginning of November last year, when GOAT surged to a market cap of several hundred million dollars in a very short time, and Binance listed GOAT futures. This made me rethink the sector. But before that, I'd like to mention some earlier AI-related events. Prior to GOAT, two major events stood out. One was WorldCoin—because its co-founder is OpenAI's Sam Altman, WLD was treated as OpenAI’s meme in crypto and一度 reached an FDV over $100 billion. The second was Turbo, which claimed to be an AI meme created by GPT and rose 200–300x on CEXs last year. These two cases clearly showed strong interest and炒作 potential around the AI+Crypto theme.
Back to late October last year, right after GOAT launched on Binance Futures, I immediately thought of WorldCoin and Turbo, recognizing Binance’s ability to lead sectors. At that time, there were four assets that stood out to me—two AI-related: ai16z and ACT; and two non-AI: LUCE and BAN. Looking back, ai16z turned out much stronger than ACT, but at the time, ai16z founder Shaw was still unknown, the token was launched via daosfun with a mintable contract issue, and extreme volatility due to a tiny pool made it risky. I initially entered ai16z at a $25M market cap with about $3,000, but FUD caused the market cap to drop below $10M, so I didn’t add more. Instead, I went all-in on ACT, which appeared more stable with a larger holder base. ACT was my first serious position in AI investing. My first purchase was $3,000 worth of ACT on November 5th. Later, I noticed it on Bitget and gradually bought over $10,000 worth at a $22M market cap, averaging $0.022 per token. Honestly, I didn’t expect it to happen so fast—before I could finish buying, it unexpectedly got listed on Binance. I remember I was attending a conference in Bangkok and was quite shocked. After listing, it peaked at $700M–$800M market cap, but I sold most of my position the next morning at around $500M, keeping the rest until today.
Later, I observed that AI Agents was the only sector to evolve from generic PumpFun launches into a vertically specialized, scalable space, so I dove deeper into this area. After ACT, I caught other solid opportunities like Pippin, which I heavily invested in and returned over 10x. My total portfolio dedicated to AI investments has seen roughly 7–8x growth since November, though recently pulled back to about 5x.
What is your methodology for researching AI Agent projects? Can you walk through a case study of one project you deeply researched, to help others understand your approach?
defioasis:
Current AI or AI-related assets differ significantly from earlier AI projects. Most are now fairly launched via PumpFun-like mechanisms, meaning project teams or founders may hold little supply—sometimes even less than snipers or whales. This makes the founder’s integrity, vision, and background critically important. If they’re anonymous or lack ethics, they could easily rug and launch a new project. So my methodology starts with the person: Are they genuinely committed? Do they have the capability? What can they realistically achieve?
Take Pippin as an example. I first learned about Yohei Nakajima (Nakajima Yohei) on December 11th through the Solana AI Hackathon judging panel. The hackathon was co-hosted by ai16z and Solana. Scanning the judges, I noticed a Japanese name—Yohei Nakajima, Pippin’s founder. His work on child-oriented AI agents seemed unique—nothing else like it existed. Also, being a judge rather than a participant signaled credibility. Further research revealed he’s the founder of BabyAGI, which has over 20,000 stars on GitHub. A quick search showed BabyAGI was widely cited in media and academic papers as an AGI concept product, proving technical legitimacy. Additionally, Yohei is a partner at Untapped Capital, whose Web3 investments include Pixel—an asset later listed on Binance. Overall, Pippin’s founder has exceptional technical and capital resources. As a real-name figure with reputation, the chance of him pulling a rug is minimal.
Pippin was around $20M market cap at the time, not widely noticed. That range fits my preferred entry point—I usually buy in the $10M–$30M market cap zone. I accumulated 0.2% of the total supply—the maximum single holding I allow myself—spending about $40,000. It later dropped below $10M, but I kept accumulating. A founder’s background and tech don’t change with price swings. After finishing accumulation, I left it alone.
Later, Pippin announced a pivot to become a framework—from a single AI agent to an AI framework—causing valuation to spike. Even though the framework hasn’t been built yet, market confidence in the founder’s ability pushed Pippin to $300M. Building a framework opens the possibility of becoming a “splitter” chain. Frameworks are currently the highest-valued segment in the AI agents space, and those capable of spawning ecosystems have potential to exceed $1B market cap.
Which AI Agent projects do you like? What do they do, and why do you believe in them?
defioasis:
I like many, and still hold Pippin—but since its market cap is high now, I won’t discuss it further. I typically look around the $20M mark, but few projects get heavy allocation.
Currently focusing on two directions: mining projects from the Solana AI Hackathon, which has ended and produced many prize-winning entries—I’m currently filtering them. The other is Virtuals moving to Solana and partnering with Jupiter, which could spawn interesting projects, given Virtuals’ success on Base. Still evaluating this space.
Let me highlight one Solana hackathon project—not investment advice. Recently watching AgentiPy, an open-source framework connecting AI agents to Solana dApps via Python. Roadmap suggests launching autonomous narrative trading bots in Q1 and a launchpad in Q2. Crucially, its token APY will act as a flywheel. Tokenomics look solid: despite a fair launch model similar to PumpFun, the team allocated 40%, all streamed linearly over two years via Streamflow—indicating strong commitment. AgentiPy’s co-founder and CTO are followed by Solana’s official Twitter account. Coming from the Solana hackathon provides at least some endorsement. Still early, with high uncertainty. I’ll also keep an eye on projects launching on Virtuals post-migration to Solana.
Overall, I think AI is gradually entering the AI Application phase. Beyond frameworks, I’ll keep watching AI+ applications, especially AI+DeFi—where AI meets native Crypto narratives and DeFi assets/flywheels. There may be good opportunities, but it’s still very early. Haven’t seen strong candidates yet. I’m staying observant and haven’t entered any new positions recently.
What’s your view on the current state of the AI Agents sector and market? Do you think the AI Agent hype can last long, or has the bubble already peaked?
defioasis:
It’s definitely cooled down recently, but I don’t think it’s over. AI still makes sense. Off-chain AI continues rapid iteration and development, riding strong technological and capital momentum—that’s the more important foundation. Many AI projects are actually driven externally, whether by narratives or talent. Shaw himself was once a web2 outsider, yet now leads ai16z as a top AI project in crypto. Influenced by figures like him, I believe more traditional tech talent will enter the AI space. Crypto AI significantly lags behind off-chain progress, so every external ripple or major upgrade can spark new narratives and sub-sectors within crypto.
From an on-chain perspective, AI Agents is the only sector to evolve from generic PumpFun launches into a规模化 vertical track. DeSci might count as half, but it’s also cooled lately. Otherwise, no other sector has made this transition, indicating strong demand for AI narratives. The current chill in AI Agents reflects a correction after overheating. Now, single-agent saturation forces everyone to build frameworks and compete fiercely, leading to fatigue. Hence, AI+ applications—especially those tied to native AI+Crypto narratives—if viable, could reignite enthusiasm and unlock new opportunities.
Any insights on investment and trading? Tips for entry and exit strategies?
defioasis:
Earlier I discussed project selection, but I believe position management is even more critical. Picking assets involves assessing technology, resources, and background. Even with fair launches on PumpFun, the research approach now resembles VC analysis—evaluating tech, team, backing, token structure, insider allocations, blue-chip wallets, etc.
Now let me focus on position management. A promising asset typically goes through three stages: PvP, second-stage pump, and exchange listing—but most die after PvP. I mainly play the second stage, targeting assets in the $10M–$30M market cap range for positioning—what I call the “on-chain sweet spot.” I’ve found that decent assets often consolidate here after initial spikes and pullbacks. I only allocate capital within this range, focusing on those that have weathered one or more ~70% drawdowns and stabilized in this zone.
When I find a promising asset, I add it to my watchlist, categorizing into S-tier, A-tier, and others. S-tier assets can form pattern-based markets—like母币 ecosystems that profit from continuously spawning子coins. For example, as trading pair assets, newly created子coins generate wealth effects requiring the母币 for trades, creating sustained demand—similar to SOL on PumpFun or the母coin flywheel on Virtuals. Thus, S-tier usually includes framework projects with ecosystem or flywheel potential. Pippin rose largely due to high market expectations around its framework pivot.
A-tier is assessed based on narrative, background, tech, resources, and team—but cannot form self-sustaining models. Founder/team history matters: e.g., Solana Foundation affiliation, high GitHub stars, past attention-grabbing products, or potential to pioneer niche sub-sectors.
Typically, I establish initial positions in both A and S-tier projects—around $2,000–$3,000 each. But deciding to add or overweight is stricter—I observe for some time, evaluating community engagement and developer activity.
I have strict limits per asset. Maximum holding per coin is 0.2%—about $40,000 spent near a $20M market cap. Each decision to add is re-evaluated: if the project contradicts my original thesis, I stop adding; otherwise, hold existing position. If price breaks below the consolidation range significantly, I stop adding, keeping the rest. In principle, I aim to hold at least 10x, potentially exiting around $200M–$300M market cap.
Initial entries ensure I’m “on the train”; cautious, gradual additions with a hard 0.2% cap prevent overcommitting due to blind confidence and getting trapped in a single asset. Once I decide to add, I trust my judgment.
The above describes second-stage strategies. But recently, the market—whether users, devs, or launch factories—seems too familiar with this playbook, making second-stage plays harder. So I’ve started experimenting with lottery-style tactics—playing PvP rounds. This isn’t limited to AI anymore. I’ve set aside a few SOL, buying $0.1 SOL chunks each time, actively monitoring and copying high-conviction addresses—small bets for big returns. Over the past couple days, I’ve found this lottery approach more profitable in current conditions. It’s probabilistic—so many launches, if bets are small and diversified, and aided by tracking high-win-rate addresses, hitting just once can recover all prior losses. In a market lacking big trends, this works well—and prepares me mentally and financially for potential breakout moves. Maintaining a sense of consistent profitability is crucial.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
吴说区块链
