Crypto is in a washout phase, and Lao Bai does not try to soften the picture. In a long-form conversation hosted by 168X, he described August 2026 as a bottoming period marked by deep bearish conditions, exchange shutdowns, product closures, recurring security incidents across DEX and DeFi, and a broad migration of talent and money toward AI.
BitMEX and BitMart have announced the end of trading operations. Former cycle favorites such as Zapper and Fantasy Top are shutting down. At the same time, he said, venture capital in the primary market is being cleared out as capital rotates elsewhere. Against that backdrop, 168X invited Lao Bai, also known as @Wuhuoqiu, to discuss a basic question: if crypto has already won in important ways, why do so many participants still feel like losers, and what opportunities are left for the next cycle?
Lao Bai comes from a technical background, with ten years of network engineering experience. He entered crypto in 2017 after trying to buy Bytom, later worked in research and investment roles at Amber, ABCDE and OKX Ventures, and now studies a much broader range of themes spanning AI, semiconductors, and the tokenization of traditional finance. Inside crypto, he has narrowed his attention to what he sees as the few sectors that still have product-market fit: stablecoins, perpetual futures, real-world assets, and prediction markets.
From there, he built out a set of hard-edged views. Token issuance, in his words, is not real financing but a form of liability. The category of “crypto VC” will eventually disappear. Prediction markets have been overestimated and their ceiling is far lower than perpetuals. The age of treating blockchain like a hammer and every problem like a nail is over. The two biggest crypto-native inventions, he argued, are not a new Layer 1 but stablecoins and perpetual contracts. And in the long run, Robinhood looks closest to the final form of the exchange business.
From engineer to cross-asset researcher
Asked to introduce himself, Lao Bai said his career in the industry moved from research at Amber to an investment and research partner role at ABCDE, followed by several months at OKX Ventures. He said his focus today is much narrower inside crypto than it once was.
He now concentrates on Perp, prediction markets and RWA, which he sees as some of the few areas in crypto where PMF still clearly exists. Much of the rest of his time has shifted toward US equities, especially AI-related names, plus other investment opportunities. More recently, he has spent substantial time studying options because they expand the ways he can express a view on both the price and the value of an asset.
Too many projects were built “for blockchain’s sake”
When the conversation turned to prior market cycles, Lao Bai said a recurring mistake by both VCs and founders was the tendency to use blockchain as a universal answer.
His version of first principles strips away much of the standard crypto vocabulary. He said most users do not really care about decentralization, privacy or censorship resistance. What matters at the core is that blockchain became the first network capable of peer-to-peer value transfer. The internet moved information. Bitcoin, and later blockchain systems more broadly, moved value by preventing information from being copied infinitely through mechanisms such as double-spend prevention. In that sense, he said, a blockchain can be understood as an upgraded BitTorrent: a peer-to-peer download network turned into a peer-to-peer value transfer network, with programmable assets layered on top.
He pointed to posts from Satoshi Nakamoto around 2009 to 2012, saying those discussions already hinted at a future where Bitcoin could carry more than just bitcoin itself, including invoices and bonded contracts. In his reading, Bitcoin originally aimed at some of the territory Ethereum later pursued, before Satoshi disappeared and Vitalik Buterin implemented that broader programmable-asset vision through Ethereum.
That line of thinking helps explain why he was bullish on Bytom in 2017. At the time, the idea of bringing real-world assets on-chain felt to him like the exact thing blockchain should be used for. It became the first altcoin he bought. Looking back, though, he now sees Bytom as a project that was early enough to become a martyr rather than a pioneer.
He also said that during the DeFi Summer era, as prices surged into late 2020 and 2021, he lost sight of that original principle. By 2024, after reassessing the market, he concluded that nearly all of the tokens he still considered valuable were tools that improved blockchain programmability, throughput or infrastructure. From Uniswap V1 to V2 and V3, Aave across its own versions, and later names such as Morpho and Pendle, the industry kept building better rails. What it did not build, he said, was a durable base of on-chain assets with lasting value.
NFTs, GameFi items, inscriptions, ORDI and even meme tokens all failed that test in his view. He recalled that game items as blockchain assets had once looked plausible partly because Vitalik had highlighted finance and gaming as early landing zones back in 2017. But he believes that experiment has already failed. He referenced a prior essay of his on why World of Warcraft succeeded while GameFi did not, arguing that anything that can be scaled can also be industrialized, and gaming economies can be wrecked by that dynamic even before adding a blockchain layer.
He grouped prior “star sectors” into two broad buckets. One included projects such as Bytom and Augur, where the direction may have been right but the timing was too early. The other included projects such as Bancor and Kyber, where execution fell short and Uniswap took the market instead.
His current test for a crypto product is blunt: remove the token and ask whether the product still stands. He thinks Uniswap and Aave do. He thinks GameFi names such as Axie Infinity and STEPN do not. If the product collapses without the token, he no longer treats it as a sound direction.
Why token issuance stopped working
On why token issuance once worked but has become much harder, Lao Bai pointed to two structural reasons.
First, early-stage crypto lived on dream multiples. Projects with no users and no revenue could still command high valuations as long as KOLs and major VCs believed the business model might work five or ten years down the road. Retail demand followed those signals. If a project raised from firms such as a16z or Paradigm, he said, retail investors effectively bought a “dream multiple option” on that future.
Second, there were simply fewer projects competing for incoming capital. Hundreds of billions of dollars could flow in while only dozens, or perhaps a couple hundred, projects were available to absorb it. That imbalance favored issuers. Over time, the number of projects exploded, more VCs were launched, and the standard playbook became obvious: raise from funds, secure endorsements, issue low-float high-FDV tokens, lock supply and let retail become the final liquidity exit. Once the market moved from “more meat than monks” to “more monks than meat,” the cycle broke down.
He was also asked whether successful cash-flow businesses still need tokens. Taking the Blur versus OpenSea case, he said one answer comes from retail optics and another from the company’s own incentives. Retail may view OpenSea’s failure to issue a token at the right moment as a missed chance, with a possible valuation of at least $10 billion at the time. But from the operator’s standpoint, he said, a token is not automatically helpful if the business already generates strong cash income. In his words, “the token you issue is your debt.” Token issuance does not necessarily maximize financial value or brand value for the business itself.
That logic extends to Polymarket. Lao Bai said a token could help Polymarket push harder on distribution and perhaps outrun Kalshi more clearly. But he also cited opinion, which almost reached Polymarket-like volume during its token airdrop wave from late last year into early this year, only to see both volume and token performance collapse afterward. So he does not see a hard necessity for Polymarket to issue a token. It can go either way.
Crypto won in one sense and lost in another
One of the central questions in the discussion was why many long-time participants feel they lost even as crypto has moved deeper into the mainstream. Lao Bai said both views are true at once.
Crypto has matured, he said. But its original ambition has also narrowed. The movement once aimed to build an entirely native new world. Now it increasingly looks like a reform layer for the existing financial system, or even a subordinate one.
He framed it with a thought experiment. If someone in 2018 or 2019 had been told that stablecoins would become a major piece of global dollar infrastructure, with a scale in the high hundreds of billions; that Bitcoin would have ETFs and attract pension and institutional money; that crypto firms could pursue IPOs; and that Stripe, Visa and BlackRock would all be building crypto infrastructure, public chains and PayFi systems, that person would probably conclude the industry had already won.
But from the perspective of asset returns, he said, the market did not win. The setup that once allowed another DeFi Summer or 50x and 100x altcoin years is gone. And in the broader ten-year effort to create native on-chain assets, from NFTs to inscriptions to memes, he thinks crypto failed to establish long-term winners. What it built instead was a financial rail, and now that rail is being used to bring real-world assets on-chain.
Even so, he sees two clear victories: stablecoins and perpetual contracts. Those, he said, are the strongest crypto-native PMF stories the industry has produced.
On the future of crypto VC, he broadly agreed with a point made by Dragonfly managing partner Haseeb that the label itself may not survive the next decade. The reason is not that blockchain disappears, but that it becomes infrastructure rather than a standalone sector. Just as few firms now present themselves as “internet VCs,” Lao Bai expects fewer investors to define themselves primarily as “crypto VCs” as blockchain tech sinks deeper into wider commercial systems.
Perp DEX will likely consolidate into a few winners
Lao Bai was especially direct about perpetual futures. He thinks the market ultimately needs only four or five major Perp platforms, much as the exchange market itself never needed dozens of lasting leaders.
In his ranking, Hyperliquid sits alone in the T0 tier. Behind it, among the names he likes in the next layer, he listed Aster first, then Lighter, then Variational. He said Variational’s RFQ model paired with off-exchange hedging makes strong sense conceptually, and he also likes the team, founder, and the capital and institutional network behind it.
His broader expectation is that the market will converge around a few leaders: Hyperliquid at T0, and platforms such as Lighter, edgeX and Variational occupying the T1 race behind it.
He also floated a more unusual possibility for the next two to three years. As crypto perpetuals and centralized exchange products come to resemble leveraged equity contracts in traditional finance, low-liquidity small and mid-cap stocks may start using a playbook similar to Binance Alpha. In that structure, operators accumulate 60% to 70% of spot supply, keep buying and lifting the price, attach a new narrative to the company — for example, a struggling firm suddenly pivoting to AI or to digital asset reserves — and then build a large amount of OI on Hyperliquid or Binance. Shareholders or aligned insiders can then use the perp market as the exit venue.
He argued that regulators might find little to object to on the spot side if the company appears only to be buying or buying back shares, while the effective distribution happens via perpetuals. He described that as a long-term regulatory arbitrage setup that could create a new path for illiquid listed names around the world.
As for how later entrants can compete with Hyperliquid, he agreed that copycat CLOB-plus-market-maker models such as Lighter or edgeX need differentiation. Geographic specialization is one option. When asked about adding Asian equities — Taiwan, South Korea, Japan, Hong Kong — he said those routes make sense and mentioned Trasia, built by Mable, as one effort in that direction. He also said a friend is considering launching an on-chain FX perp business after leaving venture. If stablecoins ever reach the multi-trillion-dollar level on-chain, he said, the TAM for on-chain foreign exchange could become very large.
But he drew a distinction between those strategies and platforms such as Variational or Ondo Perp. In his view, these models may actually challenge Hyperliquid head-on because their execution and hedging setup can produce tighter spreads and lower slippage than Hyperliquid, and in some cases even Binance. He said Hyperliquid relies in essence on a single market-maker model for liquidity, while an RFQ-style system such as Ondo can quote and hedge through traditional finance channels. That makes it easier to lock and offset risk in-house. Lao Bai said that on many RWA-heavy names, especially large-cap stocks such as NVIDIA and Micron, Ondo Perp’s spread and slippage look better than those of Hyperliquid or Binance. He also said Variational’s mechanism should, in theory, be able to replicate that setup. Last month, its swap product already introduced traditional finance hedging mechanisms into commodities such as crude oil and gold.
Inside the Hyperliquid ecosystem itself, he said the HIP-3 market appears to be moving toward a strong winner-take-most pattern. On the front end, deployers such as TradeXYZ and Paragon are already visible. Paragon may be faster and more tactical in chasing hot tickers, but Lao Bai noted that the volume gap remains huge. Some Paragon-listed assets are still only doing hundreds of thousands in volume, while TradeXYZ can post names tied to volume in the hundreds of millions.
He explained that HIP-3 ticker slots originally represented a deployment opportunity, with the first three slots free and later ones auctioned, while the winner decides what asset to launch. But because market making is handled by deployers rather than by Hyperliquid itself, he expects strong flywheel effects to develop. If TradeXYZ delivers better depth and tighter spreads on the same asset, users will keep concentrating there. Unlike in earlier AMM battles, where protocol design could produce a meaningful edge, HIP-3 looks more like a direct contest over market-maker capital and liquidity quality.
Security: time matters, and perp platforms should avoid lending sprawl
Security incidents were another major topic. The discussion referenced the Hynix wick event on TradeXYZ, a recent hack of about $18 million at RWA project Ostium, and the hack affecting Paradex. Lao Bai said user trust in DEX infrastructure depends on two things: time, and avoiding the temptation to fold lending complexity into the perp core.
His first point was simple. Most users will never read code. They judge systems by reputation and survival time. He pointed to Aave as a case where trust has remained strong because the lending protocol itself has not suffered a direct core security failure, even if some issues have reached it through underlying assets.
He said most lending-platform attacks end up revolving around lending itself: low-quality assets are manipulated through oracle flaws, price pumping, or governance capture, then used as collateral to drain stablecoins. Mango on Solana was one example. Cream, associated with Machi, was hit multiple times.
For perp operators, his advice is to stay disciplined. Do not let the product become a do-everything venue that adds collateralized lending and creates new attack surfaces. In his view, if a perp protocol expands into that territory, problems become a matter of time. Beyond that, he said, there is no shortcut: if Hyperliquid goes four or five years without a security accident while competitors do not, user capital will naturally concentrate there.
Will Hyperliquid build spot markets?
Lao Bai does not think Hyperliquid is likely to build a native crypto spot business centered on assets such as BTC, ETH or SOL. The economics, in his reading, no longer justify that focus.
He said 80% to 90% of Binance revenue now comes from perpetuals, and starting last month more than 60% of perpetual revenue at Binance and HTX was already tied to US equities — in other words, RWA — rather than native crypto assets. If Hyperliquid does move into “spot,” he said, the more plausible path is stock spot designed to support its perp business, similar to Binance’s bStock idea.
He is unsure whether something like “hyperstock” will appear, but said he would not be surprised if Hyperliquid eventually launches a spot layer for stock-based arbitrage and internal hedging. The point would be to support perp liquidity, not to become a major crypto spot venue.
That line of thought led into the comparison between tokenized stocks and perpetuals. In an ideal world, Lao Bai said, tokenized stocks would mean real shares, clear legal ownership and free on-chain transferability. He added that Nasdaq itself may move in that direction and mentioned that the US Securities and Exchange Commission had recently approved Nasdaq to work on on-chain stocks.
But in reality, he argued, most users do not need full legal ownership or direct stock possession. They want a clean price exposure linked to the underlying. On that basis, perpetuals remain the dominant format. That is also why so much exchange revenue still comes from perp products, with a large share inside that coming from RWA-related names.
Stablecoins remain his highest-conviction crypto sector
If there is one theme Lao Bai sounds most unequivocal about, it is stablecoins. He called himself “super bullish” and described stablecoins as the most important invention in crypto outside Bitcoin — even saying that crypto’s biggest contribution to the world may be stablecoins rather than Bitcoin itself.
His reasoning is straightforward. Bitcoin now functions largely as a game for a relatively small number of whales, institutions and large holders. Stablecoins, by contrast, are the technology that has actually changed how value moves in the real world. Last year, he said, the market discussed the role stablecoins could play in absorbing US debt when Treasury debt stood at $36 trillion. That figure is now $39 trillion. In that setting, if stablecoins can reach $1 trillion to $2 trillion before 2030, they could absorb roughly $2 trillion of Treasury demand and ease part of the burden on the US government. Ten or twenty years out, he said, the scale could potentially reach $5 trillion or even $10 trillion.
On the business side, he noted that firms such as Stripe and Visa are already entering the sector directly, building around payments and yield-bearing use cases. On the consumer side, stablecoins are already being used to access and trade a growing range of global assets. In his view, stablecoins and tokenized US stocks will ultimately work together, extending both dollar hegemony and dollar-asset hegemony into more countries and more corners of the world.
For would-be new entrants, however, he set the bar extremely high. He said only players that already control key payment rails, financial gateways or settlement positions in traditional finance — firms such as Stripe, BlackRock and Visa — are realistically positioned to take meaningful share in stablecoins. He does not think ordinary startups should try to force their way into that business.
Agent payments and machine economics look real
Lao Bai also took the AI agent payment theme seriously. He said the narrative around agent payments and stablecoins is viable, even if it is not clear whether the market will ultimately describe it as a crypto story.
He broke the question into two parts. First, will an agent economy emerge at all? He thinks yes. To support that, he cited a recent interview with Cloudflare’s CEO, who said that on one day in July, bot traffic on the global internet exceeded human traffic for the first time. Lao Bai called that a major milestone.
He said the shift makes sense in the age of large language models. A human shopping for a camera might compare five websites and generate a handful of HTTP requests. A model asked to recommend a camera around RMB 5,000 with certain features might make thousands of requests across ecommerce and review sites to optimize its answer. In that context, he noted, Cloudflare’s CEO expects agent traffic to rise another 1,000x over the next five years.
The second part is monetization. The old web economy relied on Google Ads and the click-based loop linking search, recommendation and advertising revenue. But when an agent makes the recommendation on the user’s behalf, that loop weakens. Lao Bai said the most plausible answer he sees is micropayments for data access. If an agent pulls information from a merchant site or a review site thousands of times, the agent could pay a tiny amount each time it uses that data.
He expects systems such as x402, or stablecoin-based payment mechanisms rolled out gradually by Cloudflare, Visa or Stripe, to be natural candidates for that model. It may not look like a classic card payment and may not even be marketed as “crypto,” but he thinks a stablecoin on some chain is the likely settlement layer.
Public chains are moving toward functional segmentation
As for what stablecoins and agent payments mean for public-chain competition, Lao Bai does not expect a single winner. He expects segmentation.
In his framework, gray-market activity may continue to run on TRON and USDT. Retail crypto users will probably stay with USDC on Ethereum and Solana to trade DeFi and memes. Institutions, B2B users, corporate payment flows and foreign-exchange activity may migrate toward institution-focused chains such as Tempo, Arbitrum and the privacy-oriented Canton network, which he said has also been developing well.
That leaves a market structure where a few institutional chains serve business users, Ethereum and Solana serve consumer crypto users, and TRON continues to absorb gray-market demand.
He was equally clear that public-chain tokens are unlikely to remain the main beta trade for VCs or retail in the next cycle. Hyperliquid, he said, has already set a new benchmark: a chain must exist in service of a powerful application. The age of the universal general-purpose chain may have effectively ended with Ethereum and Solana, even if the market kept trying through projects such as Aptos, Sui, Monad and MegaETH.
He pointed to a recent post from MegaETH’s Shuyao saying the MegaMafia incubation plan would stop funding new on-chain projects and instead build applications directly. To Lao Bai, that was another sign that the era ahead belongs more to app-chains than to generic chain narratives.
Large models are starting to look like public chains once did
On AI itself, Lao Bai said the large-model market already resembles the old public-chain race. Different models are beginning to differentiate on capability, cost-performance and censorship boundaries.
He cited Chinese names such as Kimi, DeepSeek and MiniMax as examples. MiniMax’s recent model launch, which leaned toward a less restrictive setup, immediately appealed to some video creators. In his reading, these are exactly the kinds of feature-level distinctions that once separated blockchain ecosystems.
But the long-term winners in AI, he said, will be decided by two things. One is enterprise integration. OpenAI and Anthropic have both started building field-deployment engineering teams and are learning from the Palantir playbook, trying to embed AI much more deeply into business workflows. Despite the hype, he said, more than 90% of enterprises have so far done little more than hand employees tools from Anthropic, OpenAI or Doubao. AI may improve individual productivity by more than 500%, but overall enterprise improvement may still be below 20% to 30%.
The second battleground is consumer killer apps. Beyond products such as ChatGPT, Codex and Claude, the market still needs to see whether any other truly mass-market applications emerge.
He also referenced a recent interview in which Palantir’s CEO said Anthropic is now telling the public stories that Palantir had effectively been executing already. To Lao Bai, that helps explain why Palantir stock has performed well: the market is recognizing that major SaaS platforms are not easily displaced by frontier models.
Prediction markets are real, but their ceiling is much lower than perp
Prediction markets drew one of Lao Bai’s sharpest critiques. He said the category is real, and that projects such as Kalshi and Polymarket proved after the election that prediction-market products can work. But he also argued that the sector became overheated before the World Cup and is now in a rational reset.
He referenced a recent post of his describing four stages for the sector: underrated before the election, validated after the election by Kalshi and Polymarket, overrated before the World Cup, and now normalizing. He added that Polymarket volume has basically been falling since July 18, after the World Cup. If Polymarket really reaches a valuation around $20 billion, or if it issues a token that trades in the $20 billion to $30 billion range, he said he would short it without hesitation.
His case against the sector rests on several points. First, prediction markets do not benefit much from liquidity spillovers. DeFi Summer exploded partly because zero rates and massive liquidity pushed money from US equities into crypto and then from Bitcoin into altcoins. Prediction markets are expressions of opinion, not assets in the same sense, so they do not absorb that kind of spillover as effectively.
Second, the attention frequency is too low. Sports and politics produce major events every few days or every few weeks. Meme markets and broader trading culture can generate new narratives every day, even every hour. He specifically contrasted this with products such as pump.fun and the endless stream of meme catalysts.
Third, the upside ceiling is lower because prediction markets lack the classic “lead car” effect that creates outsized returns elsewhere. In memes or traditional markets, users often follow prominent traders or narratives. He pointed to Dogecoin’s old run toward a $70 billion market cap under Elon Musk-fueled momentum, where some participants made 100x or even 1,000x returns. Prediction markets have no comparable upside. In theory, moving from a 10% probability to 100% is a 10x outcome. In practice, he said, most mispricing opportunities are closer to 20% or 30%, and even a 2x is already a very strong outcome.
Fourth, prediction markets rely too heavily on rational decision-making. People only trade when they believe they have better information or understanding than the market. That naturally limits the number of events each person wants to trade. A user may have five or ten themes they understand well. A meme, by contrast, can spread across an entire feed and attract ten thousand, hundreds of thousands or even millions of holders in a short span.
For all those reasons, Lao Bai said he is much less bullish on prediction markets than he was last year. He accepts that the PMF is genuine. He does not think the TAM justifies current hype, especially when compared with perp.
He was also asked about projects blending prediction markets with meme mechanics, including 42space, a YZi Labs-backed project that he has invested in and that introduces bonding curves and meme-like elements. Lao Bai said he used to be strongly against memes, but seeing the limitations of prediction markets has made him less dismissive. In his framing, one major difference between crypto markets and traditional finance is that crypto has more users seeking high-volatility bets and more tools built to satisfy that appetite. Prediction-market functions in traditional finance can often be replaced by options or CDS, while in crypto prediction markets still do not feel volatile enough for many users. That is part of why he sees experiments like 42space — making prediction markets more “degen-friendly,” in his words — as a worthwhile avenue to explore.
Robinhood as the exchange endgame
Lao Bai described Robinhood’s strategy as unusually clear. In his view, Robinhood effectively understands that users want two very different things and serves both: RWA for investors, and meme exposure for speculators.
Long-term investors want assets such as Micron, NVIDIA and Hynix, not what he called empty VC-backed altcoins. More aggressive traders want volatility, and memes provide it. He said this dual-track setup is exactly what makes the platform effective. It captures attention first, then gives users a place to rotate capital once they want to keep gains in more conventional assets.
He compared that to Base, which also used memes to attract early attention because meme markets naturally produce 100x or 1,000x stories in the beginning. Robinhood, in his reading, is simply more explicit about the full funnel: bring users and money in, then keep them there by offering mainstream assets as the next step.
That leads to one of his bigger strategic conclusions. Exchanges should stop thinking of themselves as crypto exchanges. Their future competitors are not just Binance, OKX, Gate and other existing venues, but also Robinhood, Interactive Brokers, Tiger Brokers and even traditional banks. Users care about who offers the most assets, the best liquidity, the lowest cost and the easiest experience. The exchange of the future, he said, should be a single entry point into global risk assets.
That is why he repeated a line he has used before: Robinhood is the ultimate form of the exchange business. In the long run, he thinks users will expect one place where they can fund with USDT or USDC, buy US stocks, buy crypto, use leverage, trade futures and access prediction markets in one integrated app.
This bear-market cleanup feels worse because growth is gone
Every cycle bottom brings exchange closures and product shutdowns, Lao Bai said, but this one feels worse. BitMEX and BitMart stepping away from trading, and products such as Zapper and Fantasy Top shutting down, are all signs that crypto has entered a more mature and harsher phase.
In earlier bottoms, he said, an established exchange without a major security failure was unlikely to disappear. Products such as Zapper, which had real users and real utility during DeFi Summer, might have survived as well. The difference now is that the old condition of “growth covering everything” is gone.
When new money and new users kept arriving, even weak businesses could survive and many structural problems stayed hidden. Now there is little fresh growth. Users are leaving, talent is leaving, and many traders and KOLs have shifted their attention to US equities. Without growth, every problem is exposed at once, and the circle gets smaller.
His view on the talent drain is even more pessimistic. He mentioned that Sui’s co-founder recently moved to Anthropic. Compared with 2017 or 2020, when some of the world’s brightest people flowed into crypto, the direction has reversed.
He recalled that in 2023, while at ABCDE, he had seen just how deep that talent pool was. The firm recruited several interns from Tsinghua and Peking University. Colleagues included a Hong Kong University of Science and Technology database PhD and a University of Pennsylvania graduate. Over more than two years, he said, he spoke with roughly 1,300 to 1,500 projects, and around half the founders he encountered came from Ivy League schools or places such as Harvard, Stanford, Berkeley and MIT. That concentration of elite talent was one reason he had once been deeply bullish on crypto.
Now, he does not see a near-term reversal. The issue is not only compensation. It is also that many of the most compelling technical problems in blockchain have already been worked through. Since Ethereum, the sector had hard problems in distributed consensus, open finance, on-chain trading, MEV, scaling, censorship resistance and privacy. By 2024, he said, most of those had been solved enough to stop feeling fresh. New chains and new Layer 2s often repeat the same playbook: launch a meme funnel, build a wallet, design tokenomics, add yield pools and lending, then port over Uniswap and Aave. To top-tier founders and engineers, that no longer feels exciting. AI, by contrast, still has years of unsolved problems ahead.
How his own trading framework changed
Lao Bai said his move into AI and semiconductor equities, along with a deeper focus on options, has changed his trading framework in three major ways.
First, he has worked to build a more mature system. In crypto, he said, he often traded through narrative and intuition. After making money during DeFi Summer, he developed path dependence and took concentrated positions in altcoins he thought were especially creative. He gave Luna in 2022 and GOAT in late 2024 as examples of periods when he gave back substantial gains. Even without using leverage or futures, he lost enough in altcoin spot that he concluded his process had been too immature. US equities, with real PMF, real markets and real users, forced him to think more carefully about drawdowns, capital deployment and the cost of tying money up.
Second, he said mature markets teach respect for not knowing. In crypto, an early-stage investor can read a white paper, inspect DeFiLlama data, talk with the founder for an hour or two, review tokenomics and come away feeling informed. In his view, that feeling is often misleading. Public equities are far more complex because stock prices reflect revenue, margins, inventory, capex, rates, options positioning, implied volatility, supply chains and competitive shifts all at once. Nobody can know everything.
Third, and most importantly, the number of ways to express a view expands dramatically with options. In crypto, bullishness usually means buying the token, and stronger bullishness means levering up on perp. In traditional markets, he said, the toolset is much broader. If he likes NVIDIA but thinks it is expensive, he can sell a put, as Duan Yongping has done. If he likes Hynix or Micron over the long run but worries about a near-term pullback, he can use a call spread. If he wants to trade volatility rather than direction — he gave a SpaceX launch setup as an example, even though that strategy did not work for him — he can build a straddle. Those structures force a trader to ask a more exact question: am I bullish on price, bullish on volatility, bearish on time, or some combination?
At the same time, he said he avoids pre-IPO perp contracts on Hyperliquid for assets such as SpaceX or Yushu Technology. Without an underlying listed stock, those markets become pure short-term sentiment trades. Even if the directional view is right, a single wick can wipe out both longs and shorts.
Advice for founders, traders, VCs and exchanges
At the end of the discussion, Lao Bai offered different advice for each major group in the industry.
For founders, he said the era when a good idea alone — or simply porting a traditional finance concept on-chain — was enough has passed. Building now requires stronger institutional resources, especially in North America, and a much clearer path to real PMF and actual users. If removing crypto makes the product better, then remove crypto. There is no reason to preserve the label at the expense of the product. Issuing a token and designing tokenomics are not mandatory.
For traders, his message was not to form an identity around any one market. Whether the opportunity is in crypto, US equities, Taiwan stocks, Korean stocks or A-shares does not matter. The job is to allocate risk budget where the edge and the payoff are best.
For VCs, he joked that he may not be the best person to advise anyone because “our own VC shut down.” Even so, his lesson is that funds should not remain pure primary-market crypto VCs. They should combine primary and secondary strategies. In primary, they can buy equity rather than insisting on token exposure. In secondary, they can own Bitcoin or Ethereum, hedge with perp, or buy public stocks such as Robinhood and Coinbase. The point is to stop limiting the role to a single narrow category.
For exchanges, he reserved his most serious warning. He said the challenge there is the toughest of all. Exchanges can no longer think of themselves as crypto exchanges. Their competitors are expanding from Binance and OKX to brokerages and banks. The future exchange has to become a global entry point to risk assets, not just a place to list tokens. That is the logic behind his final refrain: Robinhood, not a traditional crypto venue, looks closest to the end state.