It was a Tuesday evening in Makati — September 9, if the timestamp on my screen was telling the truth — and I was sitting in a coffee shop off Ayala Avenue with a cold brew going warm, watching a Telegram group of Manila traders go quiet for the better part of a minute. That never happens. What silenced them was a post from Eugene Ng Ah Sio, a name this corner of the internet has been quoting for four years. He wasn't calling a top. He wasn't shilling a coin. He was saying he had spent the past six months systematically learning AI equities, that his trading edge in crypto had visibly weakened, that he did not regret the shift, and that a 10x through tech stocks over the next three years looked more attainable to him than a 10x through crypto. Somewhere in the thread he folded in Anthropic's IPO and the long road toward ASI.
Nobody in that group said "he's wrong." That is the part I keep turning over.
Because Eugene is not a random account. He earned his following inside the crypto cycle — the 2021 grind, the thread-by-thread track record, the kind of trader whose position updates get screenshotted into group chats within minutes of posting. When someone with that résumé tells a crypto-native audience that the better bet lives somewhere else, it isn't a price call. It's a migration note. And migration notes are the most underpriced documents in this industry.
The map matters more than the portfolio. On one side sits AI and semiconductors, an industrial cycle with revenue lines you can audit — Nvidia, AMD, TSMC, Broadcom, the hyperscaler capital expenditure numbers that anchor the entire complex. On the other side sits crypto, where the last eighteen months delivered a $10 billion spot Bitcoin ETF inflow, a layer-2 landscape so fragmented that bridging now costs more attention than it returns in yield, and an AI narrative that arrived as a theme roughly eighteen months before it arrived as a cash flow.
Then there's the third layer, the one I keep circling. The crypto-native AI tokens. FET and AGIX merged themselves into the ASI brand. RENDER sells decentralized GPU capacity. TAO pays for machine learning work at the protocol level. AKT, GRASS, and a lengthening list of others. Not one of them appeared in Eugene's post. He pointed at the real thing and left the shadow assets unmentioned. That omission tells you exactly how he ranks them — and it should tell you something too.
The thing nobody wants to say out loud is that crypto's trading edge didn't get taken away by AI. It eroded from the inside, and it eroded precisely because the infrastructure got good.
I've felt this in my own hands. Back in 2020, during DeFi Summer, I ran a 15 ETH portfolio across SushiSwap and Uniswap, chasing APYs that changed every few hours, refreshing a Discord feed that never slept. That was a genuine edge — not because I was smarter than anyone, but because the market was slow, the information was asymmetric, and the plumbing leaked yield to anyone willing to stand under it. Then the plumbing got fixed. Aggregators ate the routing margin. MEV bots ate the timing margin. Uniswap v3 concentrated liquidity turned a game of instinct into a game of parameter tuning. Every layer of polish removed a layer of sloppiness, and sloppiness is where retail alpha lives.
The same story is playing out one level up. Parallel EVMs are shipping. ZK proving is moving from testnet theater to production. Rollup costs keep falling. Every one of those upgrades is fantastic for users and quietly terrible for directional traders, because the inefficiencies they eliminate were the same inefficiencies that paid for lunch. A market that gets easier to use gets harder to beat. That is the trade-off nobody puts on the roadmap slide.
Which brings us to the argument Eugene actually made — one he framed as sector selection beating trading edge. He's right about the mechanics and probably understating the implication. In any given cycle, beta decides more outcomes than alpha. Pick the fastest-growing sector and you will beat the best execution in the slowest one, most of the time, especially if your capital is small enough to rotate without moving the tape. From 2017 through 2021, crypto was that sector. It wasn't a strategy, it was a buoyancy. Everyone who bought the wave looked like a genius.
I know that feeling from the inside. In late 2017 I was at a crypto conference in Makati, three drinks in, watching a room of people cheer for slide decks with no code behind them. I put ₱50,000 into Icon and Waves on charisma and crowd energy, sold into the first spike for a 200% gain, and walked out believing I had read the market. What I had actually read was the room. That lesson set the shape of everything I've written since: sentiment arrives before fundamentals, and the person who notices the temperature shift gets paid before the person who finishes the spreadsheet.
So when a proven operator says the buoyancy has moved, I don't dismiss it. But I do want to interrogate the claim that the new buoyancy is easier to ride. It isn't. It's just differently hard.
Here's what the AI equity trade actually asks of someone shaped by crypto. You lose the 24/7 clock and its beautiful forgiveness — no more waking at 3 a.m. to exit a position that went wrong while you slept, because the market closed and nothing happened. You gain a calendar: earnings dates, guidance revisions, conference calls, filings, Fed meetings that move the whole complex in a single sentence. You trade against desks with quants, not against a Discord of degens with leverage. Your exit liquidity is a pension fund, and pension funds do not buy your bags because a chart looks right.
That's the learning curve he called steep, and I think he's right about its steepness and maybe wrong about its difficulty relative to crypto. Trading equities is not harder because the numbers are complicated. It's harder because the feedback loop is slower, and slow feedback punishes instinct. Crypto taught a generation to trust their gut on a fifteen-minute timeframe. Equities ask the same gut to sit still for three quarters and be judged by a CFO you'll never meet.
There's also a crowding question. AI semiconductors have been in a narrative acceleration phase for well over a year. Every retail brokerage account in the world now has an opinion about Nvidia's next quarter. The moment a trade becomes the consensus answer to "where's the 10x," the 10x has usually already been paid out to the people who arrived when it was still a weird question. That doesn't mean the cycle is over. It means the easy part of the beta is behind, and what's left looks a lot more like work.
The crypto AI tokens are a separate story, and an uglier one. FET/ASI, RENDER, TAO, AKT — these trade as shadow beta. When Nvidia gaps up overnight, they gap up on the open, and when the semiconductor complex wobbles, they wobble harder because there's no earnings floor underneath them. No cash flow. No guidance. Just a narrative wrapped around a real product that has not yet found a customer base large enough to justify the multiple. If I were going to express an AI thesis in size, I would not do it through the shadow. I would do it through the thing that reports revenue.
I watched the same dynamic play out with NFTs in 2021, from the other direction. I spent weekends at launch parties in BGC where the art was beside the point and the guest list was the product. Three pieces for twelve ETH, bought as entry tickets to rooms I wanted to be in. When the market cooled I kept them anyway, because they were still working — just not as assets. Cultural utility and financial utility are not the same ledger, and the people who confuse them are always the last to sell. That's the trap sitting under every AI narrative token right now: the story keeps working socially long after it stops working financially.
None of which means crypto is out of ideas. It means the ideas have gotten smaller and more technical, which is exactly what a maturing market looks like from the inside. Look at Bitcoin. The inscription wave from two years ago was mocked as a novelty, a JPEG infestation on a settlement layer. But that wave did something structural: it created a persistent fee market on a chain that had a genuine security-budget problem waiting for it around the halving cycle. The people calling Ordinals a fad were ignoring the only revenue line that keeps Bitcoin's security model honest. Narrative injection wasn't a bug in that story. It was the entire plot.
Now hold that next to DeFi, which has spent five years pretending its oracle problem is solved. It isn't. Feed latency is still the Achilles heel — the gap between what the price is and what the chain thinks it is, and that gap is where every liquidation cascade gets its start. The industry's answer was to decentralize oracle networks by running them on a limited set of permissioned nodes with strong reputational incentives, which is a polite way of saying we solved centralization with centralized nodes that promise to behave. It works. It also means a critical piece of the stack rests on a social contract, not a cryptographic one. That is not a solved problem. That is a solved problem's press release.
I mention it because it's the kind of thing a trader leaving the market would never cite in a farewell post — too granular, too unglamorous, too far from the price. But it's exactly where the remaining alpha sits for anyone willing to do work that doesn't screenshot well. Oracle latency arbitrage. Liquidation-cascade positioning. Fee-market analysis on Bitcoin during inscription surges. Nobody tweets about it. That's the point.
So what can we actually verify? Not the sentiment — the flows. Watch the stablecoin reserves sitting on exchanges. If the people who share Eugene's view are genuinely rotating capital into brokerage accounts, that number declines and stays declined. Watch BTC and ETH net exchange inflows, which tell you whether the "attention left, capital stayed" story holds or whether the two move together. Watch the share of total crypto volume that AI-related tokens command. If that share climbs while Bitcoin's dominance holds, the narrative has transmitted into the token market. If it climbs while everything else bleeds, we're just watching rotation inside a shrinking pool.
And watch derivatives open interest on the crypto AI names, because open interest is where conviction shows up in size. Talk is cheap and free. Leverage is neither. When I was at the Singapore financial forums in 2024, the institutional desks arriving through the ETF door told me the same thing in different words: they weren't watching Twitter sentiment, they were watching custody flows. That's how the serious money reads a rotation — not by who says they're leaving, but by who actually moves the wire.
The contrarian read on all of this is that Eugene's exit is not a bearish signal for crypto. It's a maturity signal, and maturity always feels like abandonment to the people still dancing.
Think about the pattern. In 2019, the smartest traders I knew had left for traditional markets — the ICO wreckage was too ugly, the infrastructure too broken, the vibes too dead. Some never came back. The ones who did return in 2020 and 2021 found a market slower, thinner, and more forgiving than the one they had abandoned, because the tourists were gone and the prices had stopped lying. Attention departing a market at the peak of narrative fatigue is one of the most reliable setups in this industry. It has happened before every major leg up. It will happen again.
The difference this time is that the destination is real. The AI trade isn't a pump-and-dump wearing a lab coat. It's an industrial cycle with actual earnings, and that makes the migration more durable than the 2019 exit was. Some of the people leaving will stay gone. That's not a tragedy. That's a market finally having to earn its participants back instead of assuming them.
During the 2022 crash I did the least analytical thing possible — I organized monthly meetups in BGC and let people talk over drinks while the charts bled. It looked like avoidance, and it partly was. But it also taught me something a spreadsheet never would: the people who stayed through that winter weren't there for the yield. They were there for each other. That cohort is still here, still building, still quietly shipping while the attention cycles elsewhere. We didn't see them on any dashboard. We only saw them when everyone else left the room.
What I'd push back on hardest is the framing that you have to pick one side. You don't. The people who got hurt in 2022 weren't the ones holding both crypto and equities — they were the ones who held only the thing that was going up and called it a strategy. A trader who spends six months learning semiconductors and returns to crypto a year later with a wider lens is a better trader, not a traitor to the asset class.
If more headline traders follow Eugene's path, the short-term effect is real: thinner books, quieter social volume, a marginal bid that disappears. But the medium-term effect is a market priced by people who chose to be there. I'll take that trade every cycle.
Which leaves us with the actual question, and it isn't whether Eugene is right about AI. It's whether we're watching a permanent reallocation or a seasonal rotation — capital leaving crypto for a cycle and returning when the beta flips back. We didn't ask whether AI was a real industry. We asked who would still be holding crypto when the music slowed. The way to know the answer is not to argue about it. It's to watch the flows, follow the ones who publish verifiable track records instead of convictions, and notice that the loudest exits usually mark the quietest entries.
We didn't get into this market for the years when everyone agreed with us. We got in for the years when they left.