After the tide receded, I took another look at Web3. For the past two weeks, I've repeatedly seen that article that went viral— "I Wasted 8 Years of My Life in Crypto". To be honest, I completely understand that feeling. The year I graduated, I turned down offers from five major internet companies and chose to go all in on Web3. Those years included the lockdown of Wuhan and the Shanghai incident. At that time, I was full of enthusiasm and sincerity, wanting to join this decentralized new world and build a future "that belongs to us". Over the years, I've worked on Web3 projects, done venture capital, and even started my own business. I've witnessed the rise and fall of GameFi, NFT, DePIN, and the BTC ecosystem. I've worked in prediction markets, with projects reaching 7 million users on Telegram. I've experienced the rise and fall of projects and watched many smart people around me slowly leave the scene. A growing consensus, yet one that is rarely articulated clearly, is taking shape: It seems like there's nothing "real" to do on Web3. It's not that we can't make money, but rather— There are very few projects that truly have product-market fit (PMF), real revenue, and the ability to evolve over the long term. The remaining projects are mostly highly homogenized assembly line projects that rely on emotions, narratives, and timelines. But I am not pessimistic. Or rather, I've moved past the stage where I need to judge an industry based on pessimism or optimism. I prefer to see the present as the second half of Web3. Speculation will still exist, but it will only be among the minority; the real opportunities come from understanding the misalignment of regulation, structure, and technology, rather than creating a new concept. Looking ahead to the end of 2025, based on my experience over the past few years, I only want to talk about three things that deserve serious attention in the future. I. The essence of Perp DEX: Regulatory arbitrage, not technological romance. Many people still understand Perp DEX in terms of "on-chain derivatives" and "degree of decentralization". I believe this is a fundamental misjudgment. The only reason for the establishment of Perp DEX is regulatory arbitrage. Binance's success is essentially an arbitrage opportunity exploiting Coinbase's onshore regulatory oversight. The rise of Hyperliquid is essentially a response to the increasingly stringent compliance environment of centralized exchanges (CEXs) globally. While CEXs inevitably move towards compliance, the core needs of contract trading users have not disappeared: Anonymity, fund security, extremely high leverage, and execution efficiency. These demands will only be systematically suppressed in compliant CEXs. Thus, a very clear migration path emerged, but one that had previously been impossible to implement: Professional traders and whale are moving from CEX to on-chain Perp DEX. The demand has always been there. dYdX tried an on-chain order book, while GMX used an AMM to bypass the performance bottleneck. But reality is cruel— When users find themselves facing similar liquidation risks on the blockchain but cannot access CEX-level liquidity and matching efficiency, their willingness to migrate will quickly drop to zero. After FTX crashed in 2022, GMX and dYdX did experience a surge in data, but they quickly fell silent again. The problem has never been whether there is demand for on-chain derivatives, but rather that the product form is always wrong. This is precisely the significance of Hyperliquid. It doesn't do "more decentralized," but rather directly bypasses the performance ceiling of existing public blockchains: Self-built application chain, fully on-chain order book, and millisecond-level matching bring the trading experience directly in line with CEX. Essentially, it's not creating a "protocol," but rather a complete trading system: They control the brokerage, the exchange, the clearing and settlement—all of it. The result needs no further explanation: Daily trading volume once surged to tens of billions of US dollars, with cumulative trading volume exceeding trillions, and revenue even surpassing that of many second-tier CEXs. This illustrates one thing: There is never a shortage of demand for on-chain derivatives; what is lacking is a product that is willing to sacrifice architectural ideals for the sake of trading experience. After HIP-3, anyone who stakes enough HYPE can create their own derivatives market on-chain, which is a crucial step. This means that Perp DEX no longer only serves crypto-native assets, but has begun to support any consensus asset. Of course, this market is destined not to be monopolized by a single entity. Due to differences in channels and regions, the competitive landscape of Perp DEX is more likely to mirror that of CEX: Binance : OKX : Bybit ≈ 3 : 1 : 1. Players like Aster and Lighter are already building momentum in different dimensions. Among them, Aster has a unique advantage in terms of money, resources, and channels. The competition among Perp DEXs has never been about "who is more decentralized." The question is who can provide a "sustainable gray area" for professional traders within the regulatory cracks. The only thing hanging over our heads is the sword of Damocles that is regulation. But at least until the next US presidential election in 2028, this remains a track with a very clear time window. II. RWA × Perp: Assets are not the key, liquidity is. RWA has been talked about too much and misunderstood too much in the past two years. There is only one type of RWA that has truly been validated by the market: US Treasury bonds, gold, and indices These are highly liquid assets with a high degree of consensus. The "on-chain" registration of real estate, private equity, and various low-liquidity assets is essentially just a change in the method of ownership registration, and it does not solve the core problems such as information asymmetry, valuation, and exit mechanisms. The truly interesting direction is RWA × perpetual contracts. Perpetual contracts are naturally suited to meet the needs of high leverage, high-frequency trading, and liquidity. When combined with assets such as macroeconomic data, interest rates, inflation expectations, and credit spreads, it essentially builds an on-chain macroeconomic financial toolbox. If AI agents are further integrated to automate strategies and portfolio trading, it is no longer about "trading a single asset," but about providing structured financial products. From what I've seen, the vast majority of RWA teams are still at the "asset issuance" level. Teams that truly understand trading, strategy, and AI are extremely rare. This is not a problem, but an opportunity. Third, market prediction: I remain optimistic, but only if we face its harsh reality. I have a deep emotional investment in market prediction. I have worked on two prediction market projects, one of which reached 7 million users on Telegram. Although it ultimately did not continue operating, it instilled in me a deep respect for this field. The core barrier to market prediction has never been the rules of the game. Rather, it's about liquidity and compliance. Current prediction markets can be broadly divided into two categories. One type is sports prediction. This is a market where Product-Market Fit (PMF) has already been validated by Web2, with genuine and stable user demand and a sufficiently high growth potential. The real barriers to entry are not in the product itself, but in compliance, licensing, and operational capabilities. Next year's World Cup could very well be a crucial window of opportunity. Another type is event prediction. Polymarket started with political events, while Kalshi has gone a step further in the compliance field. Their appeal lies in the statement: "You can bet on anything." But the reality is also harsh – in the three areas of TVL, branding, and compliance, the leading position is already very obvious, and the chances of a latecomer winning in a head-on competition are not high. Where is the opportunity? I believe there are still three entry points worth paying attention to: - Currently, only UMA is available as an oracle, and it supports a very limited number of chains, including Polygon, Base, and Ethereum. Aggregators can combine events from different prediction markets, allowing for quick participation. However, the future of aggregators is analogous to that of DEX aggregators; for aggregators to succeed, a relatively fragmented market competition landscape is essential. We may have to wait and see. In the sports prediction market, next year's World Cup will be a major deciding factor. If compliance and operations can be handled well, the sports prediction market has the potential to see new leading players emerge. Prediction markets don't necessarily have to be able to "bet on everything"; focusing on a sufficiently large vertical and achieving excellence is often more effective in building a competitive advantage. As an aside: AI is not a new track, but rather an efficiency dividend in the second half of the game. Many people still regard AI × Web3 as an independent track. In my view, this is a typical example of "first-half thinking." Looking back at the earliest projects that touted AI × Web3, many stopped short of success after issuing tokens. The real role of AI is not to create a new protocol, but to scale up and productize what was originally something that only a very small number of professional traders, researchers, and quantitative teams could do. In the traditional AI market, lovable helps humans generate product prototypes with a single click, while cursor amplifies R&D efficiency many times over. In Web3, the scenarios that truly achieve Product-Market Fit (PMF), whether it's CEX/DEX, Perp DEX, various DeFi projects such as lending, or prediction markets, all share a common characteristic: The products themselves are not scarce; what is scarce is the ability to consistently make the right decisions. This is precisely where AI excels. In Perp DEX, what truly determines long-term returns is never the open position button, but rather risk control, position management, strategy combination, and execution efficiency. Judgment itself is scarce in the market, but what is even scarcer is stable, reusable, and amplifiable judgment. AI is not meant to replace humans, but rather to transform the judgment of a select few into assets that can be combined and amplified. The agreement determines the lower limit, while intelligence determines the upper limit. AI × trading/asset management may be a more realistic and worthwhile direction to bet on in the second half of Web3. Looking back in conclusion, has this industry ever let anyone down? They are those passionate people who want to change the world. Or are they people who want to get rich overnight? This industry seems to have let down many people, yet it also seems to have let down no one. It just started slower and calmer than we imagined. The second half of Web3 No longer reward the first person to tell the story. Instead, they reward those who truly understand how the system works. People who are also willing to take responsibility for long-term structures. If you are interested in trading, liquidity, and long-term thinking Instead of just trying to capitalize on the narrative boom, So perhaps— Now is a quieter and more honest starting point.