Harry MeridianX

YouTube视频链接: youtu.be/VeqmPJUMFwg 大家好我是哈利：）今天我想讲一点好玩的东西， 就是我在 4 月 10 号左右跟我姑夫打的一个赌： 我们约定在三年后的 6 月底，我的账户要从当时的大概 11,000 加币翻将近三倍，达到 3 万加币 如果我赢了，他奖励我 3 万人民币； 当然，如果我输了也没有任何惩罚，因为他觉得这根本是不可能完成的任务，但我认为这很有可能完成 不过到目前为止，我的账户只增长了 600 左右，所以按目前的进度来看我算是输定啦 但实话说，因为有这个压力在，我这半年来感觉进步飞快 简单讲讲这个过程吧。 刚开始的时候，我其实有一种莫名的骄傲，明明什么都不懂，总觉得自己认知很高，年纪轻轻就知道投资炒股，很厉害 结果紧接着就被市场狠狠“毒打”了——我买的股票跌得一塌糊涂，而看着大盘不停上涨， 当时我姑夫其实已经给我指明了方向，比如芯片股和医药股 [ 拿芯片股举个简单的例子：像德州仪器当时他提到的时候大概 186，后来最多涨到 300 多（两倍）；还有更夸张的涨了三四倍甚至五倍，做光互联，光通信的从 90 多最高涨到将近 500；就连盘子这么大的英伟达，也从 160 涨到 240 多 也就是说，当时我只要听他的随便买个芯片股，或者看看医药股（基本上翻倍的有的是）， 我的信心就不会受到那么大的打击了 当时我就想：我研究半天股票，费了大心思，结果还亏钱 这就叫“越努力越亏钱”，还不如去听成功之人的建议 但我认为有些坑是必须踩的，因为不踩一遍，你永远不知道是怎么回事 比如我们都知道散户会追涨杀跌，当听到这个词时，你可能会觉得“我怎么可能是韭菜呢？” 但实际上，绝大多数新手在股市中几乎不可避免地都会追涨杀跌 我就犯过这个严重错误：入市一年多后，碰上当时指数两个月涨了 30% 看着别人赚钱，空仓的我是特别难受、特别煎熬的，总忍不住想冲进去 于是在 6 月 5 日指数大跌时，我在跌了 3% 的时候没忍住冲进去，买了 PL（一家做航天太空的公司）和 Blackberry（当时在做国防和汽车自动驾驶的传感系统） 虽然买的位置还算可以，但因为板块不好，个股没赚到钱 之后我又犯糊涂去追高美光（Micron），在 1074 买入，最高涨到 1255 美元后就被套牢了，最多跌到 700 多，直接被套了 30% 多 这叫交易的反身性，也叫财富效应，总而言之就是定力不够，而定力这个东西真的必须吃过大亏才能醒悟过来 买入美光的惨痛教训让我记忆犹新 我最近读了几本心理学书，深刻体会到人在不知道怎么做时喜欢随大流， 但股市里往往“当大家都觉得好的时候，股价就到顶了”，随众心理几乎永远会让你亏钱 我的账户最多时从 1.2 万的高点回撤到了 6000 多，现在两个账户加起来大约 11,000 左右 段时间让我有了点醒悟的感觉 我并不认为股市存在瞬间“开窍”，只有积累到一定程度后，意念力跟控制力才会提升 对于接下来的行情，我觉得还是顺势往上看一看 对于“三年翻三倍”的目标，信心依然是有的 这段时间我付出了不少、后悔过很多次，但也真正学到了什么是“高确定性的交易” 高确定性的机会往往是要等的，是要等市场犯错、等大家恐慌的时候（比如波动率指数 VIX 在 40、50 相对绝对安全的时候）再小仓位做反弹 今天的视频就录到这里啦，感谢大家收听。

Next-gen institutional treasury like AAA level business bond is finally here. Feel free to checkout and DM for collaborations.

欢迎评论交流美股投资，交易体会

🎙️ AMA Recap — AI Creates Value. Who Owns the Upside? Today’s live AMA has ended successfully! 🎙️ We had some meaningful conversations with our guests across AI, Web3, digital ownership, RWA, decentralized infrastructure, AI agents, and the future of digital participation. A huge thank you to everyone who joined and shared their perspectives: ⛓️ @TTChain_ — RWA, on-chain verification & contributor incentives 🌐 @MarsCat_Global — decentralized identity, privacy & P2P infrastructure 🤖 @Magne_Ai — AI, hardware, decentralized infrastructure & user ownership 🧠 @4ainet — AI agents, decentralized AI & the economics of autonomous work 🎯 @dj_k25 — AI-powered entertainment, livestreaming & audience participation 🙌 @Harry_tyz — contributing to the discussion and connecting perspectives across AI & Web3 Hosted by @MeridianX_—— Stablecoin-based middle settlement layer for the global aviation industry. 💬 What did we discuss? 1️⃣ AI & Productivity If AI creates a massive productivity dividend, who actually captures the value? 2️⃣ Incentives & Contribution How can Web3 reward meaningful contributions such as data, compute, expertise, capital and community work without simply encouraging speculation or farming? 3️⃣ Transparency vs. Privacy Can on-chain transparency provide accountability while still protecting users' privacy and economic activity? 4️⃣ AI × Web3 Integration What does meaningful AI + Web3 integration actually look like, and where can decentralized infrastructure add real value? 5️⃣ The Next Web3 Narratives We explored the potential roles of RWA, DeFi, Stablecoins, PayFi and other emerging infrastructure in a more programmable digital economy. 6️⃣ Participation & Entertainment How can technology create deeper forms of participation between fans, creators and entertainment ecosystems, rather than keeping audiences purely passive? 7️⃣ The Future of Digital Ownership What is missing between simply watching, consuming or using digital content and actually becoming a participant in the ecosystem around it? 🔥 The bigger takeaway AI can make intelligence, creativity and production dramatically more efficient — but creating value and capturing value are two different things. Huge thanks to all our speakers, guests, listeners, builders and everyone who joined us today. 🤝 See you in the next Space! 🚀 #AI #Web3 #RWA #DeFi #Stablecoins #PayFi #Crypto

This morning, @MeridianX_ and I was honored to join @AlphaWireHQ AMA. Here is the Recap: How Advanced AI Agents Could Reshape Web3 1. Why do advanced AI agents matter? Advanced AI agents matter because AI is evolving from a passive information tool into an active participant in digital markets. A traditional chatbot can explain market conditions, but an advanced agent may be able to monitor those conditions continuously, compare multiple sources, make a decision, and act within permissions established by the user. Crypto provides a natural environment for this development. Blockchains operate continuously, while wallets, stablecoins, smart contracts, and decentralized applications offer programmable infrastructure. An agent could potentially research an opportunity, pay for data, interact with a protocol, and verify the result using the same digital rails. Agents could also make Web3 easier to use. Instead of requiring people to manually understand networks, bridges, gas fees, and smart-contract approvals, an agent could translate a user’s intended outcome into a series of technical actions. However, the narrative is developing faster than the technology. The long-term opportunity will depend on whether agents can act safely, consistently, and in the user’s best interest. 2. What separates a real agent from the hype? The main difference is that a chatbot responds, an automation follows fixed instructions, and an agent works toward a goal. A genuine agent should be able to observe its environment, evaluate possible actions, use appropriate tools, execute a decision, check the result, and adjust its next step. It should also be capable of responding when conditions change. For example, if a transaction suddenly becomes too expensive or risky, the agent should pause or find an alternative. However, autonomy alone does not make an agent useful. A system that can perform many actions without behaving reliably may simply be a dangerous bot. Real agents need clear objectives, reliable data, access to relevant tools, understandable decision-making, and enforceable permissions. Users should know which assets an agent can use, how much it can spend, and which actions require approval. Agent capabilities also exist on a spectrum. Some agents may only conduct research, while others prepare transactions or execute limited actions. The true test is not how impressive the demonstration looks, but how safely the agent handles incomplete data, changing conditions, and unexpected failures. 3. How could agents change wallets and payments? In an agent-driven economy, wallets could become much more than places to store assets and sign transactions. They could become programmable policy layers that define exactly what an agent is allowed to do. Rather than giving an agent unrestricted access to a primary wallet, users could create separate accounts with limited balances and specific permissions. An agent might be allowed to spend a certain amount per day, use approved assets, and interact only with trusted contracts. Larger or unusual actions could still require human confirmation. This approach could be supported through smart accounts, session keys, transaction limits, contract allowlists, and automatic permission expiration. Agents may also create demand for machine-to-machine payments. They could pay for data, computing resources, storage, specialized services, or assistance from other agents. Stablecoins and blockchain payments are well suited to this activity because they are programmable, global, and available around the clock. The safest model is not to give an agent complete access and hope it behaves correctly. It is to give the agent the minimum authority it needs while ensuring that every action is visible, auditable, and reversible whenever possible. 4. What data do agents need? Agents may need market prices, liquidity, trading volume, volatility, on-chain activity, wallet movements, smart-contract events, security reports, news, social sentiment, and macroeconomic information. They may also need personal context, including the user’s portfolio, objectives, risk tolerance, and previous permissions. The risk is that this information can be delayed, incomplete, inaccurate, or deliberately manipulated. Fake social engagement, artificial trading volume, misleading news, and coordinated wallet activity could all influence an agent’s decisions. Agents can also be targeted through prompt-injection attacks, in which malicious content attempts to change the agent’s instructions or convince it to take an unsafe action. For this reason, agents should compare independent sources, verify when information was collected, and communicate uncertainty. They should also simulate transactions and recheck market conditions before executing important actions. The most capable agent will not necessarily be the one with the most data. It will be the one that can identify reliable information, recognize uncertainty, and refuse to act when the evidence is insufficient. 5. What can go wrong? The biggest risk comes from combining speed, access, and unclear authority. A person might make one bad transaction and then stop. An agent could repeat the same mistake many times before the user becomes aware of it. Security threats include malicious smart contracts, wallet drainers, compromised tools, fraudulent services, and manipulated data. An agent could also misunderstand the user’s objective or make an overconfident decision based on incomplete information. There may be broader market risks as well. If thousands of agents use similar models and respond to the same signals, they could act simultaneously, increasing volatility and creating automated feedback loops. Accountability is another unresolved issue. When an agent causes a financial loss, responsibility could potentially involve the user, developer, wallet provider, data source, or application. Necessary safeguards include spending and rate limits, transaction simulations, approved asset lists, contract allowlists, audit logs, real-time alerts, emergency controls, and human approval for high-risk actions. A well-designed agent should fail safely and never have more authority than it genuinely requires. 6. Who controls the agent infrastructure? Control could be distributed across AI model providers, cloud companies, data services, wallet developers, application platforms, payment networks, and blockchains. Model providers influence how agents reason. Data providers influence what agents believe to be true. Wallet providers control which actions agents can authorize, while application and API providers determine what services agents can access. This means an agent system could use decentralized blockchains while still depending heavily on a small number of centralized gateways. Web3 has an opportunity to create more open and portable infrastructure. Users should ideally be able to change an AI model, wallet, or application without losing their identity, permissions, reputation, or transaction history. However, decentralization alone does not guarantee accuracy or security. The goal should be meaningful user choice, transparent dependencies, interoperability, and clear accountability at every layer. 7. Where will agents create real value first? Research and market monitoring are likely to be among the earliest successful use cases. Agents can track markets continuously, filter large amounts of information, identify unusual activity, and explain what matters to a particular user. Transaction preparation may follow. An agent could compare routes, calculate fees and price impact, inspect contracts, and prepare a transaction for human approval. Payments, treasury management, and routine wallet operations are also promising because they can operate within clearly defined limits. In DeFi, agents could monitor lending positions, warn users about liquidation risks, claim rewards, or rebalance positions. Fully autonomous trading will attract attention, but it is more difficult because market conditions can change rapidly. The next cycle will likely move from experimentation to consolidation and then infrastructure development. Successful projects will be measured by whether they save time, reduce costs, improve decisions, and keep users in control—not simply by whether they use the “AI agent” label. Again, thanks everyone for sharing! :) @Growthy_Web3 @aetheriumX_fun @gary_yangge @fortuneglobal_ @nitromachines @BuildOnVerity @ULTILAND @ScaleVC_group @MeridianX_ @pai3ai

Last night, @MeridianX_ and I were honored to attend @fortuneglobal_ AMA. And here is an Event Recap of this "AI × DeFi × Prediction Markets in 2026." @Dragonfly_io s way of using autonomous AI agents to facilitate our daily lives was also introduced. During the AMA, we disscussed: AI agents have quickly become one of the biggest talking points in Web3. They are being described as autonomous traders, yield managers, liquidity providers and even full-time participants in prediction markets. During the session, however, the guests took a step back from the excitement to ask a more practical question: How much of this is real adoption, and how much is still driven by hype? The conversation began with the state of DeFi in mid-2026. DeFi continues to support tens of billions of dollars in capital and billions in daily trading activity, so there is no question that the underlying market is real. What has changed is how people judge new products. Users and investors are becoming less impressed by flashy demonstrations, ambitious roadmaps or the number of agents launched. They increasingly want to see sustainable revenue, returning users and performance that remains attractive after fees, slippage and token incentives are taken into account. The guests agreed that simply connecting an AI model to a crypto wallet is no longer enough to call something a production-ready financial agent. A reliable agent must understand instructions, operate within clearly defined limits and respond safely when something goes wrong. It should be able to deal with failed transactions, sudden market movements and changing liquidity conditions. Most importantly, it should be able to explain what it did and why. This is why some of the most meaningful progress is happening behind the scenes. Spending limits, approved-contract lists, transaction simulations, human approval systems and emergency stop controls may not generate as much attention as an “autonomous AI fund,” but they are essential if agents are going to manage real money safely. Prediction markets stood out as one of the clearest examples of genuine adoption. Polymarket has recorded billions of dollars in monthly trading volume, showing that demand for event-based markets is no longer theoretical. At the same time, that activity remains concentrated around major platforms and popular events. Smaller markets still struggle with limited liquidity, wide spreads, confusing wording and uncertain resolution rules. DefiLlama’s Polymarket data AI can help by following information around the clock, comparing related markets and updating prices as new information appears. It can also make smaller, more specialized markets cheaper to research and maintain. For everyday users, agents could turn complicated news and market movements into clearer explanations. Still, the panel emphasized that AI cannot solve every problem. It cannot turn an ambiguous contract into a fair market, and it cannot create a trustworthy outcome when the underlying data or resolution process is unreliable. Strong market design, dependable oracles and transparent settlement rules remain necessary. The discussion ultimately separated autonomous execution from intelligent execution. Agents can already place trades and manage wallets, but that does not mean they can consistently generate profits. For builders, the strongest opportunities may therefore lie in secure wallets, risk controls, agent reputation, verifiable performance and liquidity infrastructure. The closing message was clear: AI agents are likely to become important interfaces to DeFi before they become dominant investors. The most successful systems will combine human judgment, machine-speed execution and enforceable on-chain safeguards. Thanks @DeFiPastry @Tom_cwallet @BrcToTheMoon for joining too.