Crypto AI: Just Getting Started or Already a Bubble?
TechFlow Selected TechFlow Selected
Crypto AI: Just Getting Started or Already a Bubble?
"Compared to crypto-native narratives like DeFi and NFTs, or modified narratives like GameFi, AI is an external narrative."
Navigating Web3 tides with focused insightsHost: Alex, Research Partner at Mint Ventures
Guests: Max, host of the YouTube channel "Max's Blockchain Space"; Lydia, former researcher at Mint Ventures, currently a researcher at Particle Network
Hello everyone, welcome to WEB3 Mint To Be, brought to you by Mint Ventures. Here, we continuously question and deeply reflect—clarifying facts, uncovering realities, and seeking consensus in the Web3 world. We aim to demystify the logic behind hot topics, offer insights that go beyond surface-level events, and introduce diverse perspectives.
This episode is the first in our series “The State and Future of Web3 Sectors,” where we explore the much-discussed Crypto AI sector. In upcoming episodes, we’ll invite guests to discuss DeFi, Memes, Public Chains, DePIN, Gaming & Social, PayFi, and Web3 policy.
Disclaimer: The views expressed in this podcast do not represent those of the guests’ affiliated institutions, and any mentioned projects should not be taken as investment advice.
Alex: Today, we’re diving into the highly anticipated Crypto AI sector. We’ve invited two researchers who have been closely tracking this space. First is Max, the creator of the YouTube channel “Max’s Blockchain Space.” The second is Lydia, a former Mint Ventures researcher now working at Particle Network. Beyond Crypto AI, she has also been deeply focused on chain abstraction. Let’s start with brief introductions from both of you.
Max: Hi everyone, I’m Max. By day, I work as an aerospace engineer in Web2, but during nights and weekends, I transform into a crypto researcher. Occasionally, I publish my research on YouTube and Substack. I'm excited to be here to talk about Crypto AI—one of my most anticipated narratives in this bull market. Thanks for having me.
Lydia: Hello, I’m Lydia. I started following the AI sector around the end of last year. Alongside chain abstraction, I see AI as one of the two most important new narratives at the application layer in this cycle. Glad to be here to exchange thoughts with all of you today.
Understanding Crypto AI
Alex: I feel like this conversation comes at just the right time. On one hand, many Crypto AI projects have seen impressive price gains recently. On the other, traditional AI has had major product developments. OpenAI officially launched the Pro version of ChatGPT, priced at $200 per month. Sam Altman has announced 12 days of product updates coming soon. So let’s examine what’s happening in the Web3 world regarding Crypto AI. To start: how do you both view the Crypto AI sector? What commercial problems is it trying to solve? And how urgent are these issues?
Max: I think Crypto AI emerged to address two core problems. First, from a humanistic standpoint, centralized AI inherently carries certain issues—like censorship and other problems stemming from centralization. By integrating crypto, we gain decentralization benefits and can create systems more aligned with public desires. The second, and perhaps more interesting, aspect is the introduction of incentive mechanisms. Crypto’s main innovation is the token. With tokens, decentralized AI systems can experiment in entirely new ways through incentivization. For example, there’s a project I really like called Bittensor. It uses token incentives to support various subnets, each focusing on different research areas. This connects the open-source movement—something many want to promote. Researchers face a big challenge with open source: there’s no effective way to reward contributions. By linking AI development to crypto and tokens, we finally have a mechanism to reward open-source progress, rather than letting every company hoard their findings privately. Even OpenAI initially wanted to open up AI, but gradually shifted toward a more closed model—likely because they need a sustainable revenue stream. So fundamentally, Crypto AI leverages tokens as incentives to promote openness and decentralized development.
Alex: Got it. So through crypto-based rewards, we’re taking a completely different path compared to traditional AI development. Currently, most mainstream large models are closed-source, with few truly open ones. Recent analysis even suggests some open-source models may eventually turn closed. But within Web3, the presence of tokens allows AI projects to remain open-source while enabling diversified development and sustainable incentives. Lydia, what’s your take on this?
Lydia: When it comes to commercial problems, I find the answer less clear, especially from a crypto-native perspective. While the popular saying goes “AI improves efficiency, crypto ensures fairness,” if we look closely, the urgency of improving efficiency far outweighs that of ensuring fairness—at least for now. This reminds me of an article Alex wrote in 2022 about Web3’s foundational value. One key insight stuck with me: Web3’s core value lies in greater freedom and cheaper trust. Strong Web3 projects must identify shortcomings in existing services regarding freedom and trust, then offer superior solutions. Applying this to Crypto AI: does AI need more freedom? Technically, computing resources and data availability are limited, so AI freedom is inherently constrained. Ethically, a truly free AI is hard to imagine. Is AI’s current trust cost too high? Not necessarily. You mentioned open vs. closed source and data opacity—but these concerns mainly resonate with academics and media professionals, not average users. Moreover, using blockchain to solve these issues currently appears more costly. My points might sound a bit pessimistic, but this is from the lens of solving real-world problems and proving commercial value. Crypto AI is still very early. I recall someone quoting an a16z partner: “Many important technologies initially appear as expensive toys.” So perhaps Crypto AI’s greatest value isn’t in replacing current solutions, but in narrative power—it sparks imagination, allowing two seemingly unrelated yet cutting-edge fields, crypto and AI, to collide in people’s minds. We should give these technologies time; the problems they’re best suited to solve may belong to the future, not the present.
Alex: Understood. So Lydia’s point is that purely from the perspective of enhancing efficiency or product capabilities, current Web3 or Crypto AI projects lag behind mature Web2 AI products in performance and cost reduction. As for whether their proposed solutions address pressing commercial needs—well, maybe not yet. But they offer alternative, crypto-integrated approaches that represent frontier experiments, potentially leading to meaningful outcomes down the road. That’s where we stand, right?
Lydia: Exactly. Another way I see this sector is as a long-term exogenous narrative. Why “long-term”? Because AI—especially consumer AI like GPT—has profoundly disrupted our reality. It’s a transformative shift. Everyone talks about ChatGPT hitting 1 million users in days and 100 million in months. You don’t even need numbers—you just observe how frequently people around you use AI. When I graduated in late 2022, GPT-3.5 had just launched, and within a month, everyone in my class was using it. Now in 2024, graduation papers not only require plagiarism checks but also AI detection—and it’s expensive, costing $100–200 per check. From a capital markets perspective, OpenAI is valued in the tens of billions, Nvidia has a trillion-dollar market cap, and their product launches dominate headlines. The transformation is incredibly rapid and total. Given this, AI won’t be a passing trend—it will likely become one of the century’s most significant philosophical topics. It’s long-term, and crucially, exogenous. As I mentioned earlier, crypto and AI weren’t naturally connected after their inception—even competing for talent. During the 2022–2023 crypto bear market, AI’s appeal overshadowed crypto. Only recently have we begun telling stories of mutual empowerment. Ultimately, unlike native crypto narratives like DeFi or NFTs, or modified ones like GameFi, AI is an external narrative. We can see this in asset prices—projects like Worldcoin, Render, and Near move in lockstep with broader AI industry developments. They pump before major AI events and dump once the event starts. So my initial understanding—and still my view—is that Crypto AI is fundamentally a long-term exogenous narrative.
Alex: Lydia shared a lot. She believes much of the excitement in the crypto space stems from AI’s rapid expansion in the commercial world and its profound societal impact. This enthusiasm spilled over into crypto, fueling interest in related projects. Max, do you have anything to add?
Max: I largely agree with Lydia, but I’d like to push back slightly. You said AI is external—an existing Web2 thing—that suddenly merged with crypto. But from another angle, I believe Crypto AI is the first strong-demand case for crypto in AI since DeFi Summer 2020. Take GameFi: we added crypto incentives to gaming. But crypto is just a bonus—people wouldn’t play a game just for crypto rewards; they play because the game is fun. So crypto here is additive. DeFi is different—if you’re restricted from banking services in certain countries, DeFi becomes essential. Crypto is a hard requirement in DeFi, which is why it’s widely seen as a true product-market fit. I believe Crypto AI is the second such strong-demand narrative post-DeFi. As you mentioned, after ChatGPT launched in 2022–2023, LLMs entered early user adoption. As AI evolves, we’ll inevitably uncover centralization issues—we just haven’t noticed them yet. Unlike financial systems, which existed for centuries before we recognized their flaws (e.g., the 2008 crisis), AI is too new. But eventually, people will realize they *need* decentralized AI. Why is crypto a hard requirement in this narrative? Because certain innovations require built-in incentives. For instance, decentralized compute projects already exist. When comparing decentralized vs. centralized compute, once performance bottlenecks are overcome, decentralized options often become the preferred choice—due to lower costs or avoiding AWS/Azure. I firmly believe for Crypto AI to break out and sustain growth, it must be more efficient, better, and cheaper. People won’t adopt it just to “support decentralization”—it must outperform existing solutions. That’s the goal. We’re seeing early signs, but we can’t keep relying on Meta releasing free 3.5-billion-parameter LLMs. We need a sustainable model—this is what we must strive for.
Alex: I see. So Max acknowledges that current AI products are still early-stage and lag behind centralized counterparts in functionality and performance. But he offers a key insight: like finance, AI profoundly impacts human civilization and commerce. While AI is just beginning, issues that seem minor now may grow severe over time—problems that crypto-native approaches are uniquely positioned to solve.
Project Categorization in the Crypto AI Sector
Alex: Your discussion on the first question was excellent—precisely because you didn’t fully agree, offering richer perspectives. Let’s move to the second topic: sector classification. Crypto AI is broad, encompassing diverse business models and project types. Based on your understanding, how would you categorize these projects?
Lydia: A common approach is splitting into “Crypto empowering AI” and “AI empowering Crypto.” Currently, we see more of the latter—crypto projects adding AI features. Examples include integrating APIs to build Web3 chatbots that answer project questions, using AI to improve Web3 code, or involving AI in yield strategies. Recently, AI agents launching tokens fall into this category. These focus more on new narratives than on efficiency or fairness. The reverse—Crypto empowering AI—has higher potential but is harder to implement and validate, requiring more time. The holy grail here is embedding crypto deep into AI’s tech stack to enhance privacy and transparency, though this may take longer to materialize. More immediately, crypto could improve specific AI industry segments. For instance, as Max mentioned, GPU projects aim to aggregate and incentivize idle compute resources via crypto, lowering costs. Similarly, data and algorithm markets aim to find product-market fit through freedom. However, as I noted earlier, demand here isn’t clearly proven yet. Looking at IO’s GPU usage data, individual users remain a small portion—daily rental earnings from individuals are only around $1,000. A potential breakthrough might be Coinbase and Base exploring AI agents with payments. Of course, payment is an added feature—the premise is that the AI agent itself must be useful and high-quality. Those are my two classification frameworks.
Max: I divide it into three layers: architecture, resources, and applications. The architecture layer is foundational—like a base platform supporting various AI projects, similar to a Layer 1 blockchain or infrastructure. Projects like Bittensor, Near, and Sahara fall here. On top of this, the resource layer provides essential AI development components—compute power, data, models—such as Akash or Render for decentralized compute, or Vana for decentralized data. Finally, the application layer sits closest to end-users, including AI agents that assist with DeFi or other tasks. This three-layer framework resonates well with existing crypto sector structures, and since the Crypto AI narrative is still nascent, lacking consensus classifications, this seems like a practical approach.
Opportunities and Challenges in Crypto AI
Alex: Great—we’ve explored two classification methods for the Crypto AI sector. Let’s dive deeper into a critical issue touched earlier: has the demand for Crypto AI been validated? This is a core criticism many raise, comparing it to past sectors lacking PMF, like DePIN or GameFi—seen as pure narrative hype. Or, as Lydia suggested, it might simply reflect external commercial trends migrating into Web3, creating speculative opportunities. We won’t settle this debate today. But we know Crypto AI faces challenges. So: what do you see as the biggest current challenges? And looking ahead 1–2 years, what industry or narrative opportunities might arise—given that both Web3 AI and broader AI will continue advancing rapidly?
Max: The main challenge aligns with what we discussed—Lydia’s view resonates. Crypto AI is still too early. Many projects have high market caps—Bittensor, for example, reached a $5 billion valuation—driven more by speculation than fundamentals. We still lack sufficient product-market fit or widely usable applications. Most projects are vision-driven, converting aspirations into speculative vehicles. Using my three-layer framework: the resource layer is most mature. Similar models existed in Web2, now reimagined with crypto. Decentralized compute, for instance, is a well-established concept—projects like IO, Akash, and others have been around. As Lydia noted, IO has fewer retail users, but this varies by project focus—IO targets institutions, Akash serves both. I’m optimistic about the resource layer needing just the right catalyst for broader adoption, whether in efficiency or usability. The architecture layer, however, remains more hype-driven—projects like Bittensor promise massive future growth, but this remains unproven. Bittensor uses token incentives to motivate subnet teams to optimize AI models—a virtuous cycle: higher token price → higher node rewards → better optimization. But if the token price drops, this risks a death spiral—insufficient incentives to sustain participation. That’s a key risk. As for the application layer, AI agents are trendy now. Lydia and I briefly discussed this on Twitter. But genuinely useful agents that simplify DeFi or GameFi remain rare. Most current AI agents feel meme-like—virtual characters dancing, accepting token tips to perform actions. These are playful, meme-oriented. Still, they draw attention, potentially paving the way for agents that genuinely simplify on-chain interactions.
On opportunities: we’re at a great moment. Bitcoin just crossed $100K, drawing renewed attention to crypto. U.S. regulatory sentiment is softening—new administration, pro-crypto shifts in Congress. This creates space for experimentation. We’ll find valuable innovations through trial and error, letting the market decide what survives. At least we’re not in a phase where the SEC is suing everyone left and right. So this is a favorable window. AI is also capturing mainstream attention in Web2. Can we redirect that energy into crypto, attracting more builders to create truly useful projects?
Alex: Max raised a good point—he’s likely in North America. Just today, Trump appointed David Sacks as head of crypto policy. For the first time, we have a government official tasked with supporting, not suppressing, the crypto ecosystem. Big news. Lydia, please share your thoughts on current challenges and future opportunities over the next 1–2 years.
Lydia: As we’ve said, the entire Crypto AI sector is early. To refine that: I’d say it’s near the peak of the Gartner Hype Cycle—market FOMO is high, supply is exploding, but quality is mixed. Within this, agents are relatively mature—not because they’re technically advanced, but because they’re close to end-users and leverage existing Web2 tech, making them appear more tangible. The main challenge, I think, is a mismatch between market sentiment and technical progress. Why? Because many in crypto—researchers, investors, founders—still don’t understand AI well. We’re collectively catching up. As a result, I rarely see in-depth, back-and-forth discussions about Crypto AI projects, especially critical ones. Take AI agents: the hype builds, but no one questions their actual utility or whether they deliver promised freedoms like speech. Long-term, this isn’t healthy. Look at Luna—watch her stream, and you see a crude anime character twisting around, not even singing or dancing, yet the price keeps rising. No one asks why. Other projects see this frenzy and rush to copy—since most agents offer little differentiation, mainly posting on Twitter. Just like meme coins fragmented across chains, agent projects now fragment under the same framework. Essentially, it’s a theme for countless quick token launches. The biggest challenge is this sentiment-technology gap. It’ll persist, but we must assess which force dominates at each stage.
Looking ahead: industrially, we can revisit Max’s framework—assessing cost and demand across layers. We know AI runs on three pillars: compute, data, and algorithms. Can crypto significantly reduce the cost of accessing these? That’s the real demand driver. For agent-related projects, the key is moving from virtual to real utility. My master’s thesis was on virtual humans. Initially, Web2 digital humans exploded as short-video influencers—leading to many IP-based virtual idols, hosts, etc.—but most failed to find PMF, wasting resources. Ironically, while we hear less about them in media, functional digital humans are now everywhere—on Taobao and Meituan livestreams, hyper-realistic, often indistinguishable from real hosts. These functional avatars found PMF against real humans: they never sleep, run on electricity, and are vastly cheaper long-term. For Crypto AI, how can AI agents make this leap? A natural path is digging deeper into the first category I mentioned—AI empowering crypto—finding elegant integrations that boost efficiency. For example, imagine a Solver layer using AI to analyze and predict fund flows—say, from SOL to BASE or vice versa—enabling proactive liquidity allocation and settlement, dramatically improving capital efficiency. End users experience this as: faster, cheaper, better than alternatives. That’s industrial opportunity. Narratively, I recommend tracking non-crypto AI developments—especially those covered in mainstream news, not academic circles. This ties back to AI’s exogenous nature. Right now, crypto is hot, AI quieter, so “Crypto AI” manifests as “AgentFi.” But if market conditions shift or AgentFi hits growth limits, crypto may need to return to AI for fresh narratives. I’m watching ethics-related topics—like Deepfakes. These aren’t fully explored yet. I prioritize ethics over model or tech updates because Web3 folks struggle to grasp technical depth, whereas ethical dilemmas evoke universal emotional responses. Plus, when AI ethics come up, crypto’s strengths in transparency and openness shine—there’s narrative potential there.
Alex: OK, I’d like to add two points. We’ve discussed two topics, but why revisit AI agents? Because lately, so many friends ask me about them—especially crypto investors wondering, “What do you think about AI agents? Is this the next big wave?” Mainly because these projects have surged recently and are mostly new. Our current view aligns with what both of you shared: this is a meme-driven theme, like stock market themes—constantly shifting, not necessarily due to breakthroughs, but because the market collectively embraces the narrative. Why? As Lydia said, and Max echoed, current AI agents don’t introduce new business models. They do things already possible in traditional internet—like aggregating web info to suggest tokens, with hourly or daily updates. Coincidentally, the market’s red-hot, some tokens spiked, and people think, “Wow, it’s like a genius AI investment advisor!” But viewed as a product, it’s not magical—just doing what was always possible. The hype resembles our earlier obsession with DeSci (decentralized science) or political memecoins—attention migrates, and people speculate. It’s not about industrial breakthroughs—yet.
Another upcoming narrative: Musk and Sam Altman both suggest AGI—Artificial General Intelligence—could emerge by 2025. Per OpenAI’s disclosed roadmap, they may launch an AI agent product in 2025. By then, the public may still be unprepared for what AGI means. Today, most use GPT as a tool—for formatting, writing, image generation. It assists decisions but isn’t a human-like agent. That moment hasn’t arrived. But I believe it will—by 2025 or 2026. Then, human labor and even existential value could face massive disruption. When that global economic and commercial shock hits, society’s attention to and anxiety about AI will reach unprecedented levels. This overflow of attention could massively inflate speculative value in crypto’s AI projects. That’s why we’ve long believed Wordcoin has significant upside. It aims to solve two problems. First: distinguishing humans from AI agents. Today, this seems trivial—few AI bots running wild—so verification feels unnecessary. But post-2025, this could become a real crisis. Second: as AI agents flood the workforce, labor costs plummet, white-collar jobs vanish. Wordcoin’s focus on universal basic income—giving everyone money to secure basic survival—might then resonate deeply. So this narrative could become a major societal touchpoint in the next 1–2 years.
Promising Crypto AI Projects to Watch
Alex: Let’s get more concrete. Many in our audience are crypto investors who want to know: if you had to pick one or two AI projects to watch closely, which would you choose? Share your reasoning and the potential risks involved.
Lydia: My immediate thought is Bittensor—I just saw its price nearing a new high, possibly hitting it today. I’ll highlight three aspects—won’t dive into technical architecture or tokenomics now. First: narrative mastery. This is often mentioned but underestimated. I’ve watched Bittensor’s YouTube videos and social media posts. Their team presents itself in a way that deeply appeals to developers—warm, sincere, yet ambitious. They look at you with wide-eyed innocence and say, “I want to build something great—will you support me?” It’s hard to refuse. I suspect key team members are Hayek fans—they frequently quote his views on free markets and neoliberalism, consciously applying experimental principles to token design and resource allocation. This resonates with investors interested in capitalism and market philosophy. They reinforce this image through live streams, documentaries, and offline gatherings. This cultivated identity is hard to falsify in the short term, especially during a hype cycle. The result? Bittensor has a large, high-quality fanbase—top institutional researchers, investors tweeting about it, experienced AI and Web3 developers publicly joining the ecosystem. Each offline meetup converts new followers—almost missionary-like—and this shows in the price. That’s narrative power.
Second: institutional adoption. Grayscale announced a decentralized AI fund in July, initially including TAO, FIL, NEAR, and RENDER. TAO’s allocation was small—around 3%. I found that odd, but within a month, Grayscale separately launched a TAO-only trust. Last month, Grayscale’s parent company formed a subsidiary, Yuma, dedicated to developing the Bittensor ecosystem, led by Grayscale’s founder and CEO himself as the subsidiary’s CEO. This level of institutional endorsement is unprecedented. Also, TAO is young—many don’t realize it only launched in 2023—making its positioning unique.
Third: resilience through adversity. Unlike many projects that explode overnight, Bittensor weathered major FUD. In March, Twitter flooded with attacks on its subnets, tokenomics, and team. Its price plunged—from March to August, it lost about two-thirds, bottoming near $200. But before the pre-AI-agent rally—around September—it rebounded sharply and held. This demonstrated resilience and strategic direction. Today’s subnet projects differ greatly from those in March. Researchers mapping Bittensor’s ecosystem show it now resembles a real AI network. I follow it because TAO can act as a basket token representing its entire ecosystem. Each subnet is a standalone project. Hardcore TAO fans believe all AI projects could eventually join the TAO ecosystem, unified by TAO as the intermediary token. It’s a complete, self-correcting system—with淘汰 mechanisms. As agents gain popularity, stronger agent-focused projects enter, pushing out weaker ones in token capture and emissions. It’s a constant renewal process.
Alex: Great, Max—feel free to share your thoughts, even if you also focus on Bittensor.
Max: If we were to dedicate a full episode to Bittensor, we could. I’ve written a research piece on it, and I believe Lydia has too. Bittensor is definitely my top focus. But let me add a few risk points Lydia didn’t mention. First, why I care about Bittensor: as I said, crypto’s core function in Crypto AI is incentives—enabling novel, transparent, decentralized products. Bittensor’s mission is simple: build a strong incentive mechanism. Get that right, and success follows. Evidence shows it’s on the right path—better models rise, weaker ones exit the mainnet, thanks to effective incentives. Going forward, Bittensor aims to sustain this Yuma Consensus mechanism over the next 5, 10, or 20 years, fostering a decentralized AI ecosystem. It’s pioneering this space. Its tokenomics mirror Bitcoin—capped at 21 million. As Lydia noted, the team is brilliant—nearly every YouTube video is highly technical, assuming prior knowledge. They avoid price chatter, instead focusing on problem-solving. I recall a hack on their mainnet—they identified and fixed the issue within days. Impressive.
Now, the risks. First: its Bitcoin-style tokenomics mean extremely high issuance—around 20–30% annual inflation. Daily token output is massive, continuously diluting value. But in a bullish market, price effects may be masked. Second: although aiming for decentralization, the mainnet is currently controlled by OpenTensor Foundation—the team behind Bittensor. They hold governance power, though plan to transition to PoS, distributing control to stakers. For now, despite its decentralized ambitions, Bittensor remains highly centralized. These are two major risks. Still, Bittensor is fascinating—entering its ecosystem reveals endless learning. It hosts 50–60 subnets: some work on DeepFake detection, others on LLM optimization, decentralized compute, data, etc.
Beyond Bittensor, other projects show promise. Vana, for example, focuses on decentralized data. As LLMs consume more data, authentic data will grow scarce and valuable. Vana uses token incentives to encourage users to manage and contribute their own data. Future AI apps needing such data must pay Vana token holders. Arweave, building AI compute at the architecture layer, is also promising. NEAR, through its incubator, supports diverse AI applications—worth watching.
Evaluation Strategies for Crypto AI Projects
Alex: After discussing specific projects, let’s talk investment methodology. Both of you shared your favorite AI projects and reasoning. If we abstract these thought processes: what dimensions matter most when researching and selecting Crypto AI projects? What are the core factors in deciding whether to invest? If you had to list 3–5, what would they be?
Max: I evaluate projects across five dimensions: team, product, profitability, future outlook, and tokenomics. Among these, team matters most. Investing in crypto projects is like investing in startups—the founding team determines everything. Their capability shapes product-market fit, profitability, roadmap execution, and value creation. When assessing a Crypto AI project, I first investigate the founding team. Then, I expand to backers—VCs like Multicoin or other reputable firms. If known VCs invested, they likely did due diligence—I consider that a signal. Community is also key: does the community obsess over price, or genuinely discuss the project’s future, challenges, and solutions? This reveals intent. Are they in it for the long haul, contributing ideas and fixes? Or just speculating, planning to exit after a price pump? So team—founders, backers, community—is the most critical factor, not just for Crypto AI, but across DeFi and other sectors. Capable team, strong support network, positive reputation—all matter.
Lydia: I align with Max—I prioritize team too, specifically their storytelling and execution abilities. Bittensor exemplifies top-tier narrative skill. Another example: Virtuals Protocol. Its token $VIRTUAL recently entered the top 100 by market cap, quadrupling in under two weeks. I noticed them early—they began as a Southeast Asian gaming guild, pivoting to AI this cycle. My first exposure was a video showing a Mario game extended infinitely via AI—creating a perpetual gameplay loop. That referenced the book *Finite and Infinite Games*—intriguing. Two months later, they launched their agent platform. Matches my ideal team profile: visionary and executable. Such traits make projects feel organic and alive. Other examples: Ronin, Pendle—teams I see as elite in crypto. A strong team, regardless of domain—gaming, DeFi, AI—must敏锐ly capture narratives and decisively pivot to their strongest advantage, executing relentlessly.
Given investment context, I always examine what the token represents. This reflects the team’s crypto understanding—do they know how to leverage crypto for growth or internal resource allocation? Bittensor’s exploration is cutting-edge—Max mentioned its inflation, seen by many investors as risky. But consider: to sustain dozens—eventually thousands—of subnets, massive daily emissions may be necessary for proper incentives. Perhaps this reflects the team’s free-market philosophy. Virtuals is more pragmatic—its token acts like a platform currency: to engage in AI agent speculation or investment on their platform, you must hold $VIRTUAL. But both share a key trait: utility from Day 1. The Crypto AI narrative can be abstract, but the token must have concrete use.
One more flexible criterion: how cool is the project’s brand, culture, and community? No single standard, but loosely: if a project only claims to outperform others in some metric, it’s not cool. Cool projects say, “I’m fundamentally different—no one compares. If you appreciate that, you’ll naturally join us.” That kind of vibe.
Alex: Understood. I’ll add a small point: when researching Crypto AI or other emerging Web3 sectors, I apply cyclical thinking. As Lydia noted, revolutionary innovations—like DeFi—typically experience short-term over-optimism, followed by bubble bursts and long-term pessimism. We should assess where a sector stands: in over-optimism or prolonged pessimism? If its commercial value is real, it will eventually emerge from pessimism. So for long-term investors, the best entry point is during widespread, long-term pessimism. For example, buying quality DeFi or L1 projects in 2023. Though DeFi’s recent gains aren’t huge, if you recognize its deep value, you can allocate heavily and hold long-term—no need to monitor daily prices, as fundamentals change slower than price.
AI may follow a similar pattern. Judging by sentiment, Crypto AI is in its first cycle—its breakout year likely 2024—now entering short-term over-optimism. It may not have peaked, but historically, when bear markets hit, most first-cycle projects drop 90–95%, like previous DeFi or GameFi cycles. But I believe AI’s longevity will exceed GameFi’s, which had heavier Ponzi characteristics. So long-term, after the bear market deflates the bubble, most AI projects may fall over 95%. Not necessarily from today’s TAO price of ~700 to 70, but perhaps from 2000 back to 200 or 100. That could be a solid opportunity.
Sharing Favorite AI Tools
Alex: Let’s end with a lighter topic—possibly unrelated to Web3. Since our theme is AI, I assume both of you use many AI tools daily. Which ones do you rely on, how do you use them, and what roles do they play?
Lydia: My top tool is GPT, used in two ways. First, unrelated to productivity—I use it to practice English. I ask, “How can I express X in 10 different ways?” Very helpful. Second, as a mental health companion. I enjoy chatting with GPT. When it launched, many treated it as a casual chatbot. I still do—and frequent chats mean GPT now understands me deeply. With a simple query, it infers my underlying issues based on past conversations. It provides remarkable emotional comfort. Second tool: Perplexity, mainly for search. Excellent—comprehensive, especially for English content. If I discover a project but lack time to read the whitepaper, I ask: “Compare Project A and B’s tokenomics” or “Do they differ in veToken models?” It finds answers fast. Sources are cited—if unclear, I click through to original pages, boosting research efficiency. Third: a ByteDance plugin called Doubao. When watching YouTube, it generates a timeline summary on the side, letting me jump to key sections. These are my three go-to tools.
Max: I’m a heavy ChatGPT user. For me, it’s a powerful knowledge absorption tool. When reading articles or listening to podcasts, I usually listen fully—ChatGPT summarizes well, but I prefer hearing nuances firsthand. But for critical materials I lack time for—say, a 20-page PDF—I paste it in and ask ChatGPT to summarize. Fantastic for information gathering and organization. Also, as I produce YouTube videos and research reports, I avoid letting ChatGPT edit my writing—my voice matters, and AI edits lose authenticity. But for images, I use ChatGPT to generate visuals based on scenarios—no need to hire designers for thumbnails or report covers. Those are my main uses. Future plan: input my investment research framework, train an AI agent to draft reports, and debate with it. Still experimenting.
Alex: I relate—GPT is my most-used tool too. Two main uses: first, I run a Telegram channel, updating 3–4 Web3 investment memos weekly—covering key news and my analysis. These are long, covering a dozen items. I ask GPT to add small icons before each headline—makes the layout richer. Second, when reading books, I encounter concepts worth deeper reflection. Recently, I read JD Vance’s memoir *Hillbilly Elegy*, filled with unfamiliar religious terms—neo-evangelicalism, secularism. Pre-GPT, I’d skip them—seemed irrelevant or too tedious to research. Now, I ask: “What is secularism? Neo-evangelicalism?” GPT explains origins, definitions, historical evolution—fully. It’s like having an infinitely patient, 24/7 expert tutor. During learning, you access unlimited educational resources. Crucially, it’s personalized—everyone’s questions differ. So in education, GPT’s potential is boundless. As it integrates virtual humans and spatial interaction, traditional classroom teachers may fade. People may prefer interacting with virtual tutors—their patience and personalized teaching surpass human limits. Like Lydia, I also use Perplexity. Since adopting it, I’ve stopped using Baidu and Google. It scans broad sources, delivers precise answers with citations. I now pay for Perplexity Pro. Today I heard Google’s CEO Sundar Pichai say next year’s search will undergo massive AI-driven transformation. Users will feel the impact. I believe future search will be AI-first. So mastering AI tools will soon be as essential as computer literacy was 20 years ago—a fundamental productivity upgrade.
Thank you both for your insightful, diverse, and thoughtful contributions. We hope to invite you again for future episodes. Appreciate it.
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
Mint Ventures
