A Space discussion that goes deep into the big picture—macro, AI, TradFi, quant trading, and Bitcoin cycles
Last night I happened to come across a Space session with a very direct title: “So, who is sucking the liquidity out of crypto?”
I originally just wanted to go in and listen for a bit, but ended up listening from 10 p.m. all the way through. The topics went from macro liquidity, AI, and US stocks, and kept extending to quant trading, the four-year Bitcoin cycle, Satoshi Nakamoto’s original design, and OP_CAT.
The whole session had a huge amount of information. I reorganized the parts I personally felt were valuable, and also added some of my own understanding. The content of this article mainly summarizes the on-site views of the Space guest. It does not mean I completely agree with all viewpoints, nor does it constitute any investment advice—only for communication and reference.
The article is quite long. If you’re also researching Crypto, AI, TradFi, and quantitative trading, I suggest you like and save first, then read slowly when you have time later.
──── ✦ ────
Table of contents
01｜Who exactly drained Crypto’s liquidity?
From four angles: macro, AI, TradFi, and industry credibility.
02｜Macro is short of money: why high-interest-rate environments are least friendly to Crypto?
Capital is concentrating to both ends: low-risk assets and high-growth assets.
03｜Are stablecoins still suitable for judging Crypto liquidity?
The use cases for USDT and USDC have changed noticeably.
04｜Money hasn’t disappeared—it’s just that traders changed markets.
US stocks, Korean stocks, gold, and crude oil are competing for Crypto-native trading users.
05｜What Crypto may be losing might not just be capital, but also “legitimacy/orthodoxy.”
Scams, Ponzi schemes, and excessive speculation cause long-term damage to industry credibility.
06｜What is truly worth期待 (to look forward to) in AI + Crypto?
In the future Agent Economy may naturally require on-chain accounts and settlement systems.
07｜The problem might also be with Crypto itself
Why, after all these years, are there still not enough products that are truly used for high-frequency trading?
08｜The four-year Bitcoin cycle— is it really a规律 (rule)?
What’s the relationship between Social Consensus, the halving cycle, and the monetary policy cycle?
09｜I asked on mic a question I’ve been researching for about half a year: can AI quant really be done?
7.4 billion tokens, a 60% win rate, and why the strategy suddenly fails.
10｜The hardest problem for quant might not even be finding strategies
Market Regime, switching between multiple strategies, and trading frequency.
11｜What was Satoshi Nakamoto initially trying to make Bitcoin into?
Block rewards, transaction fees, scaling, and Bitcoin’s original path.
12｜OP_CAT and the Bitcoin endgame as seen by Bruce
The truly successful infrastructure should be “forgotten” by users, just like HTTPS.
13｜With US stocks and prediction markets, why trade Crypto still?
Volatility, leverage, and Crypto’s unique risk-reward structure.
14｜After listening to the whole session, I left with three judgments
Crypto, TradFi, and AI may be reshaping into a new financial map.
──── ✦ ────
01｜Who exactly drained Crypto’s liquidity?
At the start of this Space, the host raised a really interesting question: today Crypto is so bearish—who exactly siphoned the money in the market away?
The possible directions included macro, AI, cycles, and problems within Crypto itself. The backgrounds of the guests were completely different, so the answers they ended up with were also very different.
@PhyrexNi Ni Da leans more macro, Adam leans more toward trading and asset rotation, Gary approaches from industry credibility and “orthodoxy,” and the guests from Bruce and SOLAI also believe that Crypto’s own development path over the years has had many serious problems.
By the end, I actually feel it’s hard to pick out a single unique answer. Today’s Crypto gloom looks more like the result of several factors happening at the same time.
──── ✦ ────
02｜Macro is short of money: why high-interest-rate environments are least friendly to Crypto?
Ni Da’s judgment is quite direct: one of the most core issues in the global market right now is that overall liquidity isn’t as abundant as it used to be.
When interest rates are high, a lot of capital can stay comfortably in low-risk assets like US Treasuries. There’s basically no need to take on high-volatility risks like Crypto just to earn a little more return.
At the same time, AI and semiconductors are absorbing a huge amount of high-risk-tolerant capital. So the market starts to form an obvious polarization: one side pursues safety, the other pursues extremely high growth. Assets in the middle—“not safe enough, growth not sexy enough”—end up feeling the worst.
To a certain extent, Crypto gets stuck right here. It’s still a high-risk asset, but at the current stage it doesn’t provide strong enough trends or positive “trading gains” effects—so capital with risk appetite naturally looks for new exits.
Ni Da also used an example with Nike. He views Nike as a reference asset for observing US traditional consumer and non-technology economic conditions. When even large traditional enterprises have been weak for a long time, it shows that the issue isn’t happening only in Crypto.
──── ✦ ────
03｜Are stablecoins still suitable for judging Crypto liquidity?
I think this part is worth noting because before, many people—including myself—would get used to understanding stablecoin market cap growth as “off-market capital entering the coin market.”
USDT and USDC issuance increased, and in the past that really often meant more dollar buying power entering Crypto. But today the scope of stablecoin usage has clearly expanded—needs like cross-border payments, settlement, and savings are also increasing.
Ni Da mentioned in the Space that now a large amount of stablecoin funds have moved away from pure Crypto trading scenarios. So if you only look at stablecoins’ total market cap, the reference value is declining. Stablecoin growth can’t simply be equated with these funds going on to buy BTC or ETH next.
So he now pays even more attention to ETFs, real exchange fund flows, and whether traditional investors are actually selling or buying.
──── ✦ ────
After BTC got close to $60,000, he observed a change: the funds that had been withdrawing continuously started to slow down, and some ETFs and traditional investors began buying again.
ETH also shows some similar net inflow behavior. At the very least, this suggests that for some long-term capital, around $60,000 has entered a price zone where they are willing to reallocate BTC.
But he didn’t directly define this area as a “major bottom.” I agree with this point, because the easiest mistake in a bear market is: when you get a bullish indicator, you immediately announce that a bottom has formed.
Bottoms are always clearer only after they’ve already moved out. Very few people can confirm a bottom in advance accurately via a single indicator.
──── ✦ ────
04｜Money hasn’t disappeared—the traders just switched markets.
Adam’s answer is completely different from the macro view.
His view was that a large part of Crypto’s current problem comes from traders migrating to other assets.
And the reason is also very practical: traders need volatility.
When BTC can move up or down by 1% in a day and that already counts as somewhat of a trading opportunity, meanwhile assets like US stocks, Korean stocks, crude oil, and indices start showing larger and larger volatility frequently. For people whose main purpose is trading, wherever it’s easier to make money, that’s where they go.
Adam also shared some DEX data they’ve observed. Among the top assets by trading volume on certain on-chain platforms, TradFi assets like Micron, NASDAQ, crude oil, S&P 500, and Micron are already showing up in large numbers.
──── ✦ ────
What’s truly worth paying attention to is that many Crypto-native users haven’t left the Crypto funding ecosystem.
They might still hold BTC; even if they don’t sell BTC at all, they may pledge BTC as collateral, borrow stablecoins, and then use those stablecoins to trade US stocks, indices, or crude oil futures/contracts.
In other words, the funds are still on-chain, the accounts are still Crypto’s, and stablecoins are still being used—but the underlying assets that truly generate trading volume have already changed.
This could deeply change the shapes of exchanges and wallets later on. In the future, it may become increasingly normal for one account to simultaneously show BTC, ETH, NVDA, NASDAQ, gold, and crude oil.
The only question users truly care about is: where is the行情 (market action), and where are the opportunities?
──── ✦ ────
05｜What Crypto may be losing isn’t just capital—it may also be “legitimacy/orthodoxy.”
Gary offered a really interesting term: legitimacy/orthodoxy.
What he means is that in the past few cycles, too many scam projects, Ponzi schemes, rug pulls, and projects that exist purely to extract value have appeared inside Crypto. If an industry repeats these things long enough, it will consume not only investors’ money, but also the industry’s overall credibility.
When an external investor sees Crypto, their first reaction becomes “how many scams are here?” rather than “will the next generation of financial innovation appear here?”—so it’s completely normal that the money leaves.
So from Gary’s perspective, AI, US stocks, and traditional finance only took these liquidity along; the one that truly handed liquidity over and let it out is actually Crypto itself.
I think this angle is worth pondering, because macro liquidity will eventually cycle, but if industry credibility is damaged, the time needed to repair it could be longer.
──── ✦ ────
06｜What is truly worth looking forward to in AI + Crypto?
When Gary later talked about AI + Crypto, there was one line that really stuck with me:
AI is productivity in the digital world; Crypto is the financial foundation and production relations for the digital world.
When people talk about AI + Crypto now, what they often think of is AI helping analyze the market, automatically trading, or putting Nvidia and AI company stocks on-chain.
But the truly big future use case might be in the Agent Economy.
──── ✦ ────
If in the future there are massive numbers of AI Agents, they may buy each other’s data, APIs, models, code, skills, compute power, and all kinds of digital services—then AI Agents themselves will also need accounts, payment, and settlement systems.
If you put an Agent into the traditional finance system, it has to open an account, do KYC, bind a card, and handle a bunch of permissions. If it instead directly creates an on-chain wallet—from generating an address to starting trading—it could be done within seconds.
So for AI Agents, Crypto has a very natural advantage: it is financial infrastructure that machines can directly call.
If this direction really scales up, the liquidity it brings to Crypto in the future might not be only human traders, but also economic activity between machines.
──── ✦ ────
07｜The problem might also be with Crypto itself
Bruce and the SOLAI guest’s viewpoints are more tilted toward the industry itself.
Over the past few years, Crypto has been extremely good at creating assets and manufacturing stories, and it’s also been very good at creating opportunities for speculation. But the products that actually let ordinary people open them every day, repeatedly use them, and—after using them—really improve efficiency or reduce costs are still not enough.
User behavior on many Web3 products is actually very simple: they come when there are airdrops, they come when the coin price is rising, and they come when there are yields.
Once the market turns off, users leave quickly.
This shows that many products are still truly connected to “making money demand,” and they haven’t really entered ordinary people’s lives.
If all industry users rely on speculation as the driver, liquidity will naturally swing violently along with the bull/bear cycle.
──── ✦ ────
08｜The four-year Bitcoin cycle— is it really a规律?
When Bruce later talked about the Bitcoin cycle, he shared a very interesting viewpoint.
People often talk about the four-year cycle as if it’s a deterministic rule, but in Bruce’s view it’s neither Newton’s laws nor thermodynamics—it’s just Social Consensus.
Everyone believes it will go up after halving, so they position early; as more and more people act on this logic, eventually it really forms a cycle.
But since this pattern comes from consensus, it’s obviously possible for it to be broken.
I think this sentence is really worth remembering: consensus can form patterns, and consensus can also change patterns.
──── ✦ ────
Ni Da then proposed another perspective.
Could it be that the so-called four-year cycle itself overlaps highly with the US monetary policy cycle, the election cycle, and global liquidity cycles?
When the Fed is easing, money flows from low-risk assets to risk assets; when interest rates stay high, funds return to bonds and cash again.
Crypto’s own halving cycle happens to often overlap with these macro cycles, so in the end it results in a very obvious “four-year pattern.”
If this judgment holds, then in future research on Crypto cycles, just watching the halving might really no longer be enough.
──── ✦ ────
09｜I asked on mic a question I’ve been researching for about half a year: can AI quant really be done?
When we got to this point in the Space discussion, I also requested to join on mic.
Because we’ve been talking about AI, Crypto, and trading, I happened to have a question that I’ve been tinkering with for half a year but never found a satisfactory answer to—so I took this opportunity to ask the guests:
Can AI really build a quantitative trading system that’s effective long-term?
I myself have roughly 20 years of R&D experience, and this year I’ve started using AI for quant in large quantities. From January to now, the tokens consumed just by Codex have already reached about 7.4 billion.
──── ✦ ────
Whether it’s Chan theory, raw K, order flow—including liquidation maps, liquidation data, and exchange candlesticks—anything I think is related to trading, I basically try to put into it.
For main coins with relatively good liquidity like BTC, ETH, SOL, and BNB, I also pulled the historical candlestick charts locally and used programs to conduct lots of backtesting and validation.
My original thought was pretty simple.
Since AI can read so much trading data and also write code, run data, and find patterns, is it possible that in the end we can summarize from these theories a relatively stable trading system?
After messing around for half a year, I’ve become more and more doubtful about whether a “perfect strategy” even exists.
──── ✦ ────
Around May this year, I did indeed make a version of a strategy that performed pretty well.
At the time, putting together information such as candlesticks, on-chain data, liquidation data, and data on liquidations/positions, and replaying from January to June, the win rate could reach roughly 60%, with a profit-loss ratio around 1:1.2.
From the perspective of mathematical expectation, this result can already make money.
The problem showed up in July.
After the structure of the market changes, strategies that previously performed well start entering drawdowns collectively, and nobody knows how long this drawdown will continue.
So the question I really want to ask the guests is actually:
If the market itself keeps changing, is there really a strategy that can cover the full bull and bear cycles?
──── ✦ ────
Ni Da first pulled the question back to the most basic layer.
Whether it’s Chan theory, raw candlesticks, order flow—none of these by themselves offers a 100% sure way to make money. If these trading theories are probabilistic to begin with, handing all of them to AI won’t suddenly turn them into a deterministic answer either.
He said something I found quite interesting:
“The biggest appeal of finance is precisely that it’s full of uncertainty.”
So when doing quant, there’s no need to pursue a 100% win rate.
If you can maintain a win rate above 50% long-term, and pair it with a reasonable profit-loss ratio, that’s already very difficult.
──── ✦ ────
10｜The hardest problem for quant might not even be finding strategies
Adam’s answer afterward is much closer to the problem I’ve genuinely encountered over these past six months.
He mentioned a term: Market Regime.
In simple terms, it’s basically what structure and environment the current market is in.
Trend markets require trend strategies; range-bound markets need range strategies; event-driven trading has its own event-driven playbook.
So in many cases, the strategy itself hasn’t suddenly “broken.” What’s really changed is that the market environment that fits this strategy has disappeared.
──── ✦ ────
Trend strategies run into persistent ranges, and they keep incurring losses/decay.
The box strategy originally ran very well; then suddenly it hit a strong one-direction breakout, and the same could lead to consecutive stop-losses.
Event-driven dynamics are even more direct. If your speed in getting information is one step slower than others, by the time you place your order, the opportunity might already be gone.
So the problem ultimately becomes:
When should you use which strategy set?
And if a model can always correctly identify the Market Regime and always switch to the right strategy at the right time, it might be harder for the model itself than finding some “perfect strategy.”
──── ✦ ────
This also corresponds exactly to the multi-strategy hybrid mode I’m currently researching.
I want AI first to judge whether the current market is actually a trend, ranging, or some other structure—then decide which strategy set to open.
But soon you run into a second-layer problem:
Can the model for judging the Market Regime itself be wrong?
You think you solved the strategy’s failure issue, but in the end you’ve only pushed the “uncertainty” from the first layer to the second layer.
That’s also why I’m increasingly finding quant interesting, and at the same time increasingly feeling how hard quant is.
──── ✦ ────
Another viewpoint that left a deep impression on me is Adam’s take on trading frequency.
After many people build robots, they naturally hope the robots will trade continuously. If they don’t place orders for a day, it feels like the system you built isn’t doing its job.
But if you’re not actually doing true HFT, once trading frequency increases, transaction fees, slippage, fake signals, and consecutive stop-losses will also increase.
True high-frequency trading has already entered another dimension: it’s not just models anymore, but how close your servers are to the exchange’s matching engine room, and how low the network latency can get.
Ordinary quant and true HFT are actually two completely different industries.
──── ✦ ────
This point also highlighted a contradiction I had when I used to work on strategies.
If you use only candlesticks and add very strict filtering conditions, some strategy results can indeed be made decent—but in the end you might only trade one or two times a month, or even go without trades for one or two months.
Back then, I would think: the robots are already built; making just one or two trades a month seems pointless.
Looking back later, this idea itself might have been wrong.
The value of robots is in executing based on advantages—not in proving they work every day.
──── ✦ ────
So after this discussion ends, I actually think more clearly about AI quant.
Right now, AI’s most valuable role might not be suddenly helping you create a strategy that forever makes money, but greatly improving the efficiency of the entire research workflow.
Writing code, organizing data, generating strategies, running batch backtests, adjusting parameters, validating assumptions, maintaining programs—these were things that previously might have required cooperation from many people across algorithms, development, and operations/trading. Now, one person working with a few Agents can try it.
Just from this perspective, AI is already a huge efficiency boost for ordinary trading researchers.
As for the market’s own uncertainty, it still can’t help us eliminate it.
──── ✦ ────
The advice Adam finally gave is also fairly realistic.
Before building strategies, first clarify what type of trader you are.
How much return do you want to get?
How much drawdown can you tolerate at most?
How much risk are you prepared to take?
Under what circumstances should you admit the strategy no longer fits the current market?
Only then do you choose the strategy that fits you.
At the end of the day, it comes down to one thing:
Control position size, control drawdown—first make sure you can stay in the market long term.
──── ✦ ────
11｜What was Satoshi Nakamoto initially trying to make Bitcoin into?
After the quant discussion ended, Bruce’s mic came back on, and the topic returned to Bitcoin.
The history he shared later was interesting, because his core judgment was actually different from many of today’s Crypto narratives.
According to Bruce’s understanding, when Satoshi Nakamoto designed Bitcoin at the very beginning, what he truly wanted to build was an Electronic Cash System that could be used at massive scale.
Block rewards are only subsidies used early on to attract miners to participate in the network.
As halvings keep happening, future block rewards will become smaller and smaller. In the end, miners should maintain network security through transaction fees generated by massive trading.
──── ✦ ────
If you push this logic further, a very crucial problem appears.
If in the future miners need to make money from transaction fees, there must be massive trading volume.
If you want to generate massive trading volume, the fee for each trade has to be low enough.
In Bruce’s view, Bitcoin’s later development path diverged from what appears here.
After blocks get limited, as users and trading increase, the network becomes congested and transaction fees keep rising. This directly affects Bitcoin’s potential to serve as a payment and application infrastructure.
──── ✦ ────
He specifically mentioned 2017.
At the time, trading demand grew quickly. The Bitcoin network started to get congested, transaction fees rose noticeably, and large numbers of users and applications began looking for other networks.
Including USDT—it actually ran on the Bitcoin network in the early days, and then gradually migrated to other chains.
Later, Ethereum and many other public chains rose. Of course there are many complicated reasons, but from Bruce’s Bitcoin-Native perspective, the fact that Bitcoin didn’t continue down the path of large-scale, low-cost transactions is a very important historical turning point.
That’s also why he keeps emphasizing “going back to Satoshi Nakamoto’s original design.”
──── ✦ ────
12｜OP_CAT and the Bitcoin endgame as seen by Bruce
Bruce is currently pushing forward the OP_CAT and related Layer technical roadmap.
His goal is complicated if you make it complicated; simple if you keep it simple:
Make Bitcoin capable of carrying lots of low-cost transactions again.
Transaction fees are so low that ordinary users basically can’t even feel them.
Networks can support enough applications and transactions, allowing Bitcoin to become real underlying infrastructure, not just an asset used to trade prices.
──── ✦ ────
He used HTTPS as an example, and I found that especially vivid.
Today we use HTTPS every day.
When you log in to a bank, make payments, open a website, or shop, a lot of communications behind the scenes are done via HTTPS.
But ordinary people almost never discuss:
“Today I used HTTPS again.”
Truly mature infrastructure should gradually become invisible.
──── ✦ ────
So Bruce’s final expectation for Bitcoin is also like this.
One day everyone uses it every day, yet they don’t even need to know that Bitcoin is running underneath, and they won’t emphasize “this is a Crypto product” or “this is a Web3 project” every day.
It’s like today no company would deliberately say:
“We’re an Internet company.”
Because all companies are already built on the Internet.
If one day Bitcoin can reach this kind of state, then according to Bruce, that would count as truly completing the shift from a technical experiment to infrastructure.
In the end, he used a single term to summarize:
Infrastructure.
──── ✦ ────
13｜With US stocks and prediction markets, why trade Crypto still?
At the end of the Space, a listener who had just entered the scene asked a pretty realistic question.
Now you can already buy AI and US stocks, trade prediction markets, and more and more TradFi assets are moving on-chain.
So why trade Crypto today anyway?
Ni Da’s answer is very direct:
Volatility and leverage.
──── ✦ ────
The reason everyone thinks Crypto has no momentum is very important: we’re currently in a low-liquidity phase.
If we go back to the past when liquidity was extremely abundant, Crypto’s price elasticity was in a completely different league from traditional markets.
Some altcoins move up and down by 20% or 30% in a day—under strong market conditions, that’s not uncommon.
Meanwhile, traditional large stocks usually can’t sustain that kind of volatility for most of the time.
For pure traders, this kind of volatility is itself a source of opportunity.
──── ✦ ────
Second is leverage.
In traditional stock markets, 3x leverage is already considered pretty aggressive.
In Crypto, 10x or 20x leverage is very common.
That of course implies higher risk—but it also determines that Crypto naturally attracts some traders with high-risk preferences.
As long as this market can still provide high volatility, high leverage, and a large number of new tradable assets, it will still have a segment of users that traditional markets find hard to fully replace.
──── ✦ ────
14｜After listening to the whole session, I left with three judgments.
After listening to the whole Space, I think I’m left with about three impressions.
First, there isn’t a single unique “culprit” for Crypto’s problems today.
Macro liquidity tightness, AI attracting capital, TradFi taking away trading volume, industry credibility consumed by the past few market rounds, and not enough products that can truly enter ordinary people’s lives—when these things all happen at the same time, they ultimately form the market environment we’re seeing today.
Attributing all questions to the Fed, or attributing everything to the four-year cycle—I think both explanations are too simplistic.
──── ✦ ────
Second, the boundary between Crypto and TradFi is disappearing fast.
In the future, wallets and exchanges will likely look more and more like a global asset account.
If you hold stablecoins, you can trade BTC, trade NVDA, buy gold, or do crude oil and index trading.
When all these assets appear in the same account, the same settlement system, and the same trading interface, users probably won’t even care whether they’re “trading coins” or “trading stocks.”
In the end, everyone still cares about where the opportunities are.
──── ✦ ────
Third, and also the part I’m personally most interested in: the space behind AI + Crypto could be far larger than what we see today.
If in the future hundreds of millions or even billions of AI Agents begin buying each other’s data, APIs, compute, models, code, and services, then they too will need accounts and settlement systems in the digital world.
By that time, what Crypto truly carries might not be only financial transactions between people.
It could also be economic activity between machines and machines.
If this scenario really happens, the imagination space it brings could far exceed “what coin to pump in the next round.”
──── ✦ ────
Last night I happened to scroll into this Space, just went in and listened for a bit.
In the end, after listening to it fully—from macro, to TradFi, to AI, to quant—I really had my mind refreshed about quite a few things, all the way back to Bitcoin’s original design.
As for whether this is the bottom of the bear market right now, when the next cycle will start, and whether BTC will return to previous highs again—right now I still don’t have a definite answer.
But one thing is becoming clearer and clearer to me:
When the next round really comes, the playbook will probably be different from the previous one.
──── ✦ ────
One last reminder:
This article is mainly compiled based on the on-site discussion in a Space on August 12. It includes personal viewpoints from multiple guests, as well as my understanding and summaries of some parts. There is uncertainty in market judgments, price expectations, and strategy discussions. For communication and reference only, and does not constitute any investment advice.
This content is indeed fairly long, covering many different directions including macro, AI, TradFi, quant, and Bitcoin.
If you find it useful, please like + save/bookmark.
If you come back after half a year or a year, look at which of today’s viewpoints were right and which the market refuted—I think that question itself might be more interesting than arguing who is right or wrong right now.#大漠茶馆