Guests: Imran Khan, Founder of Alliance DAO; Qiao Wang, Founder of Alliance DAO
Podcast source: Good Game Podcast
Original title: Trade War & The Markets | EP 74
Summary of key points
Imran and Qiao sit down to discuss the trade war and the markets, providing founders with straightforward cryptocurrency insights.
Summary of highlights
Next year could be a good year for Bitcoin and risk assets.
- Qiao: I am not doing anything with Bitcoin at the moment. My cost is basically zero, and the tax rate is about 30%. So I have to be very sure that Bitcoin will fall by 30% before I choose to sell it, and I don’t think so.
- Imran: I sold all my stocks in the last rally. I am just going to wait and see. I only hold Bitcoin, Solana, Fartcoin and some cryptocurrencies. Fartcoin is doing quite well and I think it is a safe haven.
- The Fed's dual mandate is to promote employment and control inflation, not to influence market prices.
- I am pessimistic about the future of U.S. stocks.
- What worries me most is not economic problems but potential military conflicts.
- To some extent, Bitcoin combines the growth potential of Nasdaq and the safe-haven characteristics of gold, so it is expected to perform well.
- If that is the case, I think under a Biden administration, there could be some kind of regional or even global conflict. Under a Trump administration, I'm not sure he would pay particular attention to the Taiwan issue.
- China is a leader in many hardware manufacturing fields, such as ships, cars, and solar energy, which are particularly important in times of war. The United States has gradually lost this capability over the past few decades.
- If the purpose of tariffs is to bring manufacturing back to the U.S., then tariffs may not be the best strategy. It would be wiser to invest in these industries and ensure that domestic companies are competitive in the global market, rather than harming their own industries through tariffs, which is a zero-sum game.
- Once businesses start using your AI product, it’s less likely for them to switch to other products because the business lock-in is deeper.
Movement Labs
Imran:
Did you see the Movement Labs announcement?
The 0x balloon lever account has been talking about Rushi and Movement for the past six months, involving some insider trading, manipulating token prices, and working with market makers that may not be very professional. I don't know the specific details. At an event in China, something happened that many people who participated in the airdrop event were very angry, and the event hosted by Movement was suspended. Then this matter went viral quickly. Many influencers started discussing Rushi's mishandling of the token issuance. A few days ago, Rushi suddenly announced that he would take a temporary leave of absence, but not permanently. We don't know more details at the moment, if you want to chat about this topic.
Qiao:
Did he sell his tokens?
Imran:
We don't know yet, but you know, OM Labs or OM, there's not enough disclosure about market makers and founders, and the same is true for exchanges. So a lot of tokens that look good on Twitter may actually be artificially created.
Qiao:
How could OM drop 90% in just one minute?
Imran:
There is no clear explanation, but someone has taken a large amount of tokens off the market and sold them.
Qiao:
Is it the market maker or someone in particular?
Imran:
Some people have pointed fingers at Laser Digital, which is based in Southeast Asia. They claim they had no involvement in this. But Hasie and a few others mentioned that there's really no transparent framework for how tokens are listed and what the market makers are doing, and a lot of that space is pretty opaque right now. So people can be misled into buying tokens and those prices don't reflect the actual value of the market. So we see a lot of this.
Market Discussion
Imran:
Is there anything else you want to talk about? Have we covered everything? Markets? Okay, let's talk about markets. We mentioned a little bit earlier that we might be bearish on the S&P 500. So, I guess we can talk about what assets you're holding right now?
Qiao:
I am actually selling stocks. Every time there is a big rebound in the market, I sell some of my stocks, but I still hold the stocks I mentioned before: Google, TSMC, and PDD.
I do worry about the war. So I sold more of TSMC than other stocks. Of course, my largest position is Bitcoin, but I don't have much to do with it. I think Bitcoin will still trade as a hybrid of the S&P 500 and gold. Recently, Bitcoin's performance is basically between the S&P 500 and the Nasdaq, actually between the Nasdaq and gold. So I'm not doing anything with my Bitcoin at the moment. I can't do anything rashly because my cost is basically zero and the tax rate is about 30%. So I have to be very sure that Bitcoin will fall 30% to choose to sell, and I don't think so. This is my position.
Imran:
I sold all my stocks in the last rally. I am just going to wait and see. I only hold Bitcoin, Solana, Fartcoin and some cryptocurrencies at the moment. Fartcoin is doing pretty well and I think it is a safe haven.
I think all the money that was flowing into safe havens was concentrated in one direction, and something happened, just like the Pepe situation. If you remember Pepe, it was similar market dynamics, but it was a bear market. As we came out of the bear market, if you remember, our market cap went from about 300 million to, I sold half of it, and then it went to about 3 billion or 4 billion. So it was similar to the dynamics back then. Other than that, nothing special. Just see how the market develops.
Market Forecast
Imran:
We were optimistic about the S&P 500 before, as we mentioned several times on the show.
Qiao:
I am not extremely bullish, I am bullish but the upside is very limited.
Imran:
We all thought things were great, we were bullish, but that was before Trump got into it. We were very bullish on crypto at the beginning, but we did get exhausted. I mentioned this in our first show at the beginning of the year, and you mentioned this as well, that while we were bullish, we got exhausted. That usually means you should sell, but we didn't sell because we're big believers in the crypto space, right? We like to hold on through market volatility.
Qiao:
In the short term, I am pessimistic about US stocks, but I don't have a particularly clear view on Bitcoin. Because Bitcoin has the growth of Nasdaq and the safe-haven characteristics of gold to some extent, it is expected to perform well. However, I am not sure how Bitcoin will perform specifically, but next year may be a good year for Bitcoin and risky assets.
Cryptocurrency Regulation
Qiao:
I find this very vexing because Massari tried to solve this problem in 2017, right? It's been 8 years and it's still an open problem.
Like, who owns what tokens, and who's selling them? Under Section 5 of the securities laws, if you're a major shareholder in a company and sell stock, you generally need to disclose it immediately, usually within a day or two. I always get a disclosure email from the SEC whenever Warren Buffett sells stock in a company he owns. I think it's over 10% of the shares. There are these rules for a reason.
Imran:
I agree. I think the permissionless nature of the market has led to a lack of adequate regulatory framework, and no one has really taken responsibility for what this should look like. From what I understand, they even discussed that we should have a process for quarterly earnings reports. Every quarter, we should have a conference call to review what we did, what we're doing next, and provide some forward guidance on how we think the market and the company will develop.
Qiao:
The great thing about crypto is that you can do this every day, not just quarterly. Yes, because everything is on-chain. If someone sells their insider shares, it's all visible on-chain. You just associate the address with the identity of the owner. That way, you get automatic disclosure in real time. This is something that traditional finance can't do, which is why they need to do quarterly reporting and so on.
Imran:
So the question becomes, who is going to take on that responsibility? Is this something that a company like ByteDance would do? Or is it something that a company like Jupiter would do? Or is this more of a form of self-regulation?
Qiao:
This is a conversation we had with Massari seven years ago. Unfortunately, it turns out that this is not a problem that a startup like Massari can solve. There were only five people like Massari at the time. This should be a problem that large exchanges and other top players in the industry should solve together. Unfortunately, this did not happen. This is absolutely the responsibility of ByteDance, Coinbase and other companies.
Imran:
I’m curious how this will work. There has to be some kind of connection between the exchanges, market makers, and founders to make this happen. But I don’t think it will happen under the Trump administration.
Qiao:
I was thinking the other day that if cryptocurrencies ultimately fail, that would be the biggest question about free markets. Because in traditional finance, you could say that the market is not completely free because of regulation, the SEC, etc. There is really a central planner who sets some of the rules. Whereas in cryptocurrencies, we don't have those rules. Yes, maybe those rules are actually necessary, and a 100% completely free market may not be the ideal approach. Maybe it should be 90% free and 10% centrally planned and regulated.
Imran:
I agree with that. I do think there should be a moderate level of regulation around how IPOs or ICOs are handled.
The impact of the speculative nature of cryptocurrencies
Imran:
That could be because cryptocurrencies are generally not seen as a solid investment, obviously because of its speculative nature, which in turn affects our overall image. So high-quality talent might not be willing to go into this tech industry because they see what’s going on. If you see a shooting on social media, are you really excited about changing the world in the future? Probably not. But if you say stablecoins, and you frame it in that way, you say, “Well, we’re a cryptocurrency, but we’re not, we’re a stablecoin.”
Qiao:
It's not a perfect analogy, but it feels like four years ago when Web3 was split from cryptocurrencies. Also, extreme Bitcoin supporters have been trying to avoid being associated with cryptocurrencies. So Bitcoin and cryptocurrencies are two mutually exclusive concepts. And now stablecoins are also being separated from cryptocurrencies.
Imran:
This might be a better option. You know, for high-end talent like in Silicon Valley, they might be more willing to join this sub-industry because then they can distinguish their work from traditional cryptocurrencies. "Oh, I'm not doing cryptocurrencies, I'm doing stablecoins."
Qiao:
I'm not sure because they understand stablecoins relatively easily. Even Bitcoin is not easy for many people to understand, but stablecoins are like an extension of fintech. It's basically fintech 2.0, involving payments and remittances, something that traditional technical people can almost understand.
Imran:
Yes, that's the narrative I'm talking about. I'm not worried about it. If we can achieve that, it could be good for the entire cryptocurrency industry.
trade war
Imran:
We are seeing a lot of interesting things on the trade war front. Trump just tweeted today that we may make a deal with China soon, and then the market went up.
The S&P 500 rose 0.79%. Yesterday, I listened to Powell's speech in Chicago, and when he was asked whether the Fed would intervene in the stock market, everyone laughed. Powell responded that although everyone hopes the Fed will intervene, we think the economy is doing well. In fact, I think the market is trying to digest this new information in the next few months. Until then, the market will be in a state of uncertainty and not know which way to go next. Therefore, it is difficult for me to make a decision until everything is clear.
Qiao:
He tried to apply pressure.
Imran:
I don't agree with this kind of pressure. Powell said that the market is digesting all the information. If the market is still volatile, what's the point of me making any decision? I agree with this.
Qiao:
Technically, markets are not part of the Fed’s dual mandate.
Imran:
The Fed's dual mandate is to promote employment and control inflation, not to influence market prices.
Qiao:
But you can see that this is becoming an area of focus for the Fed. Even though it's not their mandate, it's interesting. But overall, the trade war situation.
Imran:
What do you think about this?
Qiao:
I think the outcome of the trade war will have a much more severe impact on businesses, consumers, and markets than expected. Retail investors have been "buying the dip" over the past few weeks. What do you think? Have you talked to people around you? I feel like people still expect the Fed to save the market. They have been buying and show little sign of giving up. That's one thing. At the same time, US stocks are still very expensive. Even assuming that future earnings remain the same, why not? Because tariffs will affect earnings. But even so, US stocks are still currently trading at a price-to-earnings ratio of around 24, which is very high by historical standards.
Qiao:
I am pessimistic about the future of U.S. stocks. I am very pessimistic about the outlook for U.S. stocks. While it may be difficult to predict in the short term, I think within three to six months, the market will gradually realize these problems.
Imran:
I think the market will react once everything is finalized because there are so many uncertainties. People expect that once Trump announces a deal, we will have a V-shaped recovery, right? Everything will go back to normal. People may withdraw their previous decisions and make some adjustments. But they don't realize that we may have been affected by all the policy changes. Even in the short term, this impact will take time to recover.
Qiao:
I think the damage has been done, business confidence and consumer confidence has been psychologically affected. That means consumers will be more cautious in spending because there is so much uncertainty. The same is true for businesses, they can't make large investments because they don't know what the tariff level will be in one month or three months. So everyone will reduce their spending, which will undoubtedly have a significant impact on the economy. The tariff is 10%, and if you think about it, 10% is not a small number. Yes, the tariff is 10%. But it is not on the price of the final goods, but on the production costs, which is worse than 10% on the price of the final goods. If your business profit margin is 10%, and your costs suddenly increase by 10%, then you can't make a profit.
Someone posted a chart showing that even the 10% and 20% tariff levels are actually close to the levels of the Hollis-Smoot Tariff in the late 1920s and early 1930s. These are the highest tariff levels in the past century.
Imran:
The price of toys in the United States has gone up 80% since then. So the toy you bought for three dollars, it ended up costing seventy dollars, that's a typical example. The average car price will go up 15%, which means an average increase of $7,600. Also, the annual expenses of each family will increase by an additional $3,800, just for the average cost of things like groceries. These are things that people are not aware of, but are actually being affected.
Trade War Strategy Discussion
Imran:
So what is the US strategy in this trade war? I have seen a lot of analysis and commentary, and some traders and macroeconomic experts I follow believe that the US strategy may be similar to what Scott Besson believes. His goal at the Treasury Department may be to contain China's development, because China has become a world power, ranked second, and may surpass the United States. In order to achieve this goal, the United States may try to prevent other countries from trading with China. They want to create a multipolar world, using the position of the US dollar in global commerce and trade as a security to attract other countries to trade more with the United States instead of with China. For example, they try to use Vietnam as an example. I think Vietnam has just turned to cooperate with China.
They are also trying to make deals with Japan and other countries. This is just one example. If successful, China's GDP may fall, forcing China to succumb to US pressure. This is what some people think. What do you think?
Qiao:
If this is their goal, the actual implementation is totally inconsistent with this. Because Trump's actions so far have upset not only China, but also all of America's allies, including several countries bordering China, such as Europe, Japan, and South Korea. Now, these countries are at least trying, or publicly expressing their willingness to negotiate with China to reduce each other's retaliatory tariffs. For example, Spain has begun exporting pork to China at a significantly reduced tariff.
I've seen this in the news, though I don't know the exact numbers. A similar situation is happening with South Korea and Japan. These three countries, China, South Korea, and Japan, have historically had a tense relationship, but now they have to work together in some way. At least publicly. A similar situation is happening with Vietnam and China. So the actual policy implementation is inconsistent with the stated goals. It seems to me that the more likely scenario is that there is no clear strategy at all. Trump seems to be just acting on his instincts without thinking it through. This strategy is consistent with my impression of Trump as a businessman, who likes to bluff, create uncertainty, start with a big demand, and then gradually give in. This strategy is like the art of negotiation for beginners. Overall, I think this strategy is really bad, much worse than people expected.
The impact of the trade war on global relations
Imran:
I think this trade war is going to have far-reaching consequences around the world. Obviously, the U.S. stock market is likely to be very negatively impacted.
I noticed someone mentioned on Twitter that many people are now adding European stocks, gold and Bitcoin to their portfolios for the first time as a new investment strategy.
Qiao:
What worries me most is not economic problems but potential military conflicts.
Imran:
Do you think the situation will develop to that extent?
Qiao:
What I mean is that economic difficulties may increase the risk of military conflict. If a conflict does occur, it will most likely be in the Taiwan Strait. Before the tariff issue, I estimated that the probability of a military conflict between the United States and China over Taiwan in the next three years was about 25%. However, after this tariff and uncertainty event, I think that probability may have risen to 50%.
Imran:
If that is the case, I think under a Biden administration, there could be some kind of regional or even global conflict. Under a Trump administration, I'm not sure he would pay particular attention to the Taiwan issue.
Qiao:
But he cares very much about his image and doesn't want to appear weak. This is a reasonable consideration. I'm not sure to what extent the United States will be involved, but I think the possibility of a military conflict between China and Taiwan will increase significantly.
Imran:
That's a fair point. If that's the case, then there could be regional conflicts beyond Taiwan and China, which could be disastrous for global relations, business, and trade. Yes, I just hope he resolves this properly and we just have a mild recession and forget about it.
The impact of the trade war on startups
Imran:
So, what impact will this trade war have on local startups? I think it may be difficult for us to attract entrepreneurs from China and India, for example. What do you think?
Qiao:
Don't you think that if the economy slows down, venture capital will slow down as well? This will also affect the venture capital market, which usually reacts a few quarters later than the public market.
Imran:
I think it also depends on how far we've come in AI. Because AI is so hot right now. You know, I just did a quick data analysis looking at the past two batches of YC startups. I wanted to know what the average valuation of these companies was after Demo Day. And it turns out that the average valuation of AI startups is about $22 million. That number seems pretty high to me, right? Ideally, you want it to be around $15 million. And $22 million, without any product or even a minimum viable product (MVP), seems particularly high. I think this number will continue to rise.
So I do think we're approaching an inflection point, especially in AI, where things might change. Especially given projects like what OpenAI is doing, Winsurf is a great example of where without any product, nobody can guarantee security unless they have some kind of network.
OpenAI attempts to acquire Windsurf
Imran:
I mentioned in a chat yesterday that OpenAI tried to acquire a company called Windsurf for about $3 billion. OpenAI approached Cursor twice but was rejected. So they acquired Windsurf instead.
You can think of OpenAI as an L1, and then build applications on top of it. The company with the most applications will have the best model, which will attract more people to use the product and help them develop more applications. Is this correct?
Qiao:
Two things come to mind. First, OpenAI has not yet developed the best encoding model. They are working on it, but Anthropic may have an advantage in the encoding field. Cursor and Windsurf may also be using Anthropic's model. For OpenAI, they really want to develop the top encoding model, but this requires a lot of data.
So where does the data come from? You can get the encoding data by buying data from some companies that specialize in providing data annotation, such as Scale AI. But another way to get the data is to get it directly from Cursor or Windsurf. Because when you use Cursor and Windsurf to encode, you are actually providing data to these platforms. You prompt, request code suggestions, and accept these codes, right? As a user, you accept the suggestions. This is actually a kind of annotation of the output of the model. It's almost like reinforcement learning with human feedback. So this data is very valuable. This is why OpenAI wants this data. This is also why they tried to acquire Cursor and were rejected, and now they have acquired OneServe. This is the first point that comes to my mind.
The second point is that these big AI labs may be looking at where the market is going, where the users are, where their interests are. They may be worried that the base models will be commoditized over time, which is somewhat similar to the situation with cryptocurrency, right? The base layer may gradually become commoditized, and the application layer will capture a lot of value. So by acquiring applications that already have a large number of users, such as Windsurf or Cursor, they can consolidate their position in the application layer. Because ChatGPT is an application, right? ChatGPT is an application built on various OpenAI models. But they want more, they want coding applications, and maybe other things in the future. So I think their acquisition strategy is like this.
OpenAI as a super app
Imran:
To me, this is more like a super app. You saw the announcement about OpenAI exploring social. It's almost like an integrated platform to experience all forms of AI, whether it's tools for coding, tools for generating photos and videos to create content and stories, or even general intelligence that helps you understand everyday things, like the things you search for with Google. It's interesting how this is developing because with each new product OpenAI launches, it's slowly replacing hundreds of startups that focus on specific areas. Even in our own field, we focus on a very specific market segment, like building applications around cryptocurrency. We think there's a unique opportunity here because of the nature of cryptocurrency. But you could say that a lot of these things could be replaced by larger companies.
Qiao:
In fact, almost half of the startups in YC said that every time a new underlying model came out, their idea was no longer viable. Of course, they can always change direction, but as soon as a new underlying model comes out, their current idea is no longer viable.
Investment in AI companies
Imran:
Another interesting thing is, I don't know if you've seen that chart showing the investments of VCs like Index Ventures, A16Z, LightSpeed, etc. in major AI companies. The chart shows that these VCs typically made two or three investments in these startups. Vinod Khosla only invested once in the early stages, which was OpenAI.
Imran:
Interestingly, I spoke to Founders Fund yesterday. They have only made one investment in AI, which is OpenAI. They have some small investments in other specific AI companies, but that's very niche. When I spoke to them, they mentioned that Peter Thiel didn't want to overinvest in AI. He thought it was all going to be cannibalized like the dot-com bubble. So he told the team very explicitly not to focus too much on AI because it reminded him of the dot-com bubble.
Qiao:
I remember he expressed a similar sentiment about half a year ago, and I saw that interview on the show “All In.” Yes, he did mention the similarities between AI and the Internet bubble.
Imran:
So he reiterated this point when he recently launched a new fund, emphasizing that they should not invest too much in AI now.
AI vs. Crypto Startups
Imran:
It can be said that the market's attention to AI has reached a very high consensus, and the competition is extremely fierce. If you think the competition in the cryptocurrency field is already fierce, then you should look at what is happening in the AI field. In our current batch, there are several AI startups, and each company has about 25 to 50 competitors.
Qiao:
I think the competition and crowds in the AI space right now are probably several times higher than in cryptocurrency startups.
Imran:
I feel like the cryptocurrency space is becoming less unified in some ways.
Qiao:
In fact, this inconsistency has become increasingly apparent over the past two years.
Imran:
So much so that they are trying to redefine cryptocurrency from a completely new perspective. For example, the founder of Bridge.xyz said on Twitter a few days ago that after cloud computing and mobile Internet, the next big platform should be stablecoins. Therefore, he completely excluded cryptocurrency from the discussion and believed that stablecoins are the direction of the next major change.
Exciting startups in AI and cryptocurrency
Imran:
Which startups are you particularly excited about? Because from a technology perspective, we can predict industry trends at least a quarter in advance, right? So which startups are you most excited about?
Qiao:
I think in our current batch, there are many startups focused on stablecoins, about a third of the companies are in this field, and they are all dedicated to specific regions or application scenarios, such as remittances or income management. In addition, we also have some AI startups focused on specific industry sectors, such as a tool company for scientific research, and a company that uses AI video models for advertising. In addition, there is a team that specializes in coding services for crypto applications. Their team strength is quite strong and it is also exciting.
Poof.new
Imran:
I'm very much looking forward to working with several startups that I'm excited about, and obviously the ones you mentioned are also interesting to me. However, Poof.new is a platform that provides coding services specifically for cryptocurrencies, and I think they have a unique advantage in this regard. The team is very strong, and their members came from Phantom and Coinbase, have a deep understanding of this space, and built similar products on Flow two or three years ago. So they have the expertise to build a great product, which is very important to me.
Slop.club
Imran:
This project is a bit experimental. I know what you think about it. However, Slop.club is a video generation platform. It uses some of the latest models, such as Google's Clean and Vo, to enable people to come together to create and remix content. This function is cool, and you can even upload pictures and then generate some interesting video models around them. This kind of interaction and remixing behavior has really sparked some new trends among consumers. In fact, OpenAI's entry point into the market is image and video generation. So the question is, how will these video models be used? I think Slop is a bit like 4chan, but it is focused on video models.
Qiao:
The core insight is that the video model enables new behaviors that were not possible before. Yes, this greatly reduces the time latency for video creation and remixing. Because before, it took a long time for creators to make a video. And now, you can generate a video in seconds, at most in minutes, with a prompt. The same is true for the remixing ability, you can remix on the fly, which may have taken much longer before. So this is the core insight behind this product.
01 Exchange
Imran:
The last one is 01 exchange, which was created by the Photon and Bullex teams. You may have seen tweets about the Bullex founders today, and they don't seem to be very responsive. They are making a lot of money, and in the cryptocurrency space, the incentive structure for founders is a unique challenge. For example, even if you are a billionaire or have hundreds of millions of dollars in your bank account, are you willing to continue to develop a product? This is a unique challenge faced by crypto founders, and the best founders are able to overcome this difficulty. The Pump.fun team is a perfect example. They make a lot of revenue and continue to innovate on the product, and recently announced the live broadcast function. This is a unique challenge that founders must face.
Imran:
Bullex has slowed down a bit, and Photon is in a similar situation. I don't know how their products are progressing. So the team at 01 exchange is very eager, they work hard 24/7, and have some unique ways to build upgraded products for traders. These are some of the products I mentioned.
The impact of military conflict on AI development
Qiao:
Potential military conflicts could lead to disruptions in the semiconductor supply chain, which would have an impact on the development of artificial intelligence, especially in robotics, which are more dependent on hardware.
Imran:
And a lot of the hardware is imported from China, right?
Qiao:
Many advanced chips do come from China, although not the most cutting-edge ones, such as 3-nanometer process chips, but slightly lower-level chips and a large number of other electronic components are produced in China.
China’s dominance in robotics
Imran:
Interestingly, I found that China dominates the robotics space. Specifically, while I haven’t delved into humanoid robots, China has a 51% market share in global industrial robot manufacturing, while the United States only has 5%.
Imran:
This means that if the production and sales of humanoid robots expand, the main market will probably be occupied by China rather than the United States.
Qiao:
I think robotics will probably develop like digital artificial intelligence and AI agents, that is, the United States will be the first to innovate and become a leader, creating many new technologies, and then China will come in and make these technologies cheaper and more efficient.
Imran:
Travis Kalanick, the founder of Uber, mentioned in a podcast that when he was competing with Chinese competitors in the early days, every time he launched a new feature, the Chinese companies were always able to copy it within a day. He was very surprised by this speed and therefore developed a certain respect for these Chinese startups.
You look at the BYD cars, they're really good. I don't know if you've seen the video. The suspension can move up and down. I've seen a lot of cool features and I've even considered buying one. You can't buy it in the U.S., but I'd really like to see it.
Qiao:
I talked to someone yesterday who knows a lot about Chinese electric cars. The advantages of Chinese electric cars are that they look luxurious, their interiors are better looking than Tesla's, they are cheap, and their battery performance is good. But the disadvantage is that they are not very safe, and there are many reports of these cars catching fire.
Batteries can catch fire, and drivers are injured not because of the crash, but because the battery explodes. That's a problem. This type of news doesn't usually get made public because it's censored quickly.
Another issue is full self-driving (FSD). They don't have that technology. You know, Tesla is way ahead in full self-driving.
China's progress in solar energy
Imran:
In which products do you think China is ahead of the United States?
Qiao:
In the solar energy sector, especially in terms of energy.
Imran:
I think they make about 80% of the solar panels in the world.
Qiao:
Pretty much. These changes have mainly happened in the past decade, and they have developed very rapidly. Ten years ago, China had almost no such industry, and then they started to invest heavily in these industries. I think there are strategic considerations behind this. Although I am not sure, I think China has invested heavily in hardware manufacturing because these industries will become very critical in times of war. Now China is leading in many hardware manufacturing fields such as ships, automobiles, solar energy, etc., which are particularly important in times of war. The United States has gradually lost this capability in the past few decades.
Imran:
I think the US never really even had that capability. For example, when I read the book Chip Wars, I found that even then, the yield of chips produced in the US was not high. I think they made some efforts in this regard.
Qiao:
This brings me to the tariff war. If the purpose of tariffs is to bring manufacturing back to the United States, then tariffs may not be the best strategy. It would be wiser to invest in these industries and ensure that domestic companies are competitive in the global market, rather than harming their own industries through tariffs, which is a zero-sum game.
Imran:
It is a zero-sum game indeed. Yes, you have to compete harder and faster. When reinvesting in manufacturing in the U.S., you need to be careful not to raise costs because that would affect consumer confidence and the global economy. However, I am not so sure about the complex strategic game that some people say Trump is playing.
Advantages of listing
Imran:
When I was reading about Snapchat, I learned that the founder Evan recently gave a speech in which he mentioned that after the company went public, they became more efficient in operations, both in terms of backend management and user acquisition.
It really helped him manage the company better. So I think going public does have some positive effects because it forces founders to be more careful about capital management and user acquisition. If you're launching a new product, going public also forces founders to focus not only on short-term goals but also on the next 5 to 10 years. So I think it's very important to improve operational efficiency through going public.
The latest news on AI
Imran:
What's new in AI lately? We talked about acquisitions, but do you think if an entrepreneur wants to develop a product in a certain field, they will worry about the impact that OpenAI or other large AI companies may have? Are there any industries or fields that are relatively safe?
Qiao:
I personally feel that enterprise use cases are a little bit safer than consumer use cases. You may have a different opinion. I come from the enterprise space and can give you some reasons.
I think enterprises are more resistant to change. Once an enterprise starts using your AI product, it's not easy for them to switch to other products because enterprises are more locked in. The reason is that enterprises are made up of a group of people, while consumers are usually one person. So it's much harder to coordinate a group of people to switch products than to coordinate one person, right? So I think this is why the switching cost for enterprises is higher than that for consumers. What do you think?
Imran:
You have a valid point. But in practice, I see that companies are indeed resistant to change. Yes, but they are resistant to change because many products are sold in bundles.
Qiao:
Take Microsoft for example. If you subscribe to Office 365, you are likely to also use other products such as Azure and Outlook, which basically locks you in.
Imran:
Microsoft Teams is a prime example, which has hampered Slack’s growth somewhat because it’s free for all Office 365 subscribers, most of whom are businesses.
So I think it's actually easier for enterprises to convert these customers into new products because they're already locked into the subscription. Take Canva, one of our founders tweeted yesterday that they launched their own educational product. These educational institutions, like schools, universities, can be seen as enterprise customers, and they are all locked into this software. Now they release a new product, and any child can create an app, right? So imagine that a product like Lovable or Bold is available to everyone for free. The startups that were trying to go after these customers now don't have a chance. They can still try to go after them one by one, but they need a sales team to penetrate each market.
But there is a counterexample here, which is Zoom. I mentioned this example before. Although Zoom is not as popular as before, its usage has surged during the epidemic. One reason for Zoom's success is that every consumer is using it. Because consumers are using it, enterprises are more likely to accept and adopt this new video platform because they are using it every day at home.
The same thing happened with the iPhone and BlackBerry. BlackBerry was the standard for enterprise hardware, and the iPhone became ubiquitous. It all happened through the consumer channel. Once consumers adopted it, the enterprise had to change. That's how it happened historically. Will it be different this time? Maybe.
Qiao:
But you didn't refute my point. What I said was that it's harder for enterprises to be disrupted by OpenAI. If you build an AI product for enterprises, OpenAI's disruption will be relatively small, while products built for consumers are more likely to be disrupted.
Imran:
OpenAI can develop products for enterprises. They already have enterprise support. Because they are a pioneer, many enterprises may have adopted OpenAI's products to some extent. Therefore, it is relatively easy for OpenAI to enter the enterprise market.
Qiao:
Do you mean that because OpenAI has a large number of consumers, it is easier for them to enter the enterprise market?
Imran:
Exactly, because they're already synonymous with AI. So, it's easier for them to win in adjacent products related to AI. The same is true for Microsoft. The competition between Microsoft and OpenAI is an example. You know, Microsoft has made some big moves, and they're very competitive with each other. They've moved a lot of workloads that were originally on Azure to Oracle and other providers to balance the competition with Microsoft. So Microsoft is now also developing its own products, and they have to win those users. So OpenAI is competing with everyone. I think they are fighting too many opponents on too many fronts, and doing so often leads to failure.
Qiao:
Back to the original question, what types of AI applications are more difficult for OpenAI to easily disrupt? I think that pure application layer startups have an advantage over OpenAI because these application layer startups can choose from a variety of base models, and these models are different. Some models may perform better in certain areas. For example, Cursor and Windsurf used Anthropic's model because Anthropic's model was better than OpenAI's model in terms of encoding. Therefore, OpenAI cannot build development tools based on its own models alone, they have to acquire Cursor or Windsurf. This is a defensive ability of application layer startups against OpenAI. Does this make sense? This is essentially a network effect.
Imran:
Yes, the ability to choose different models. You can only do that if you have the distribution channel, right?
Qiao:
But you get distribution because you built something useful in the first place.
Imran:
I think in each of these cases they were early movers in the market and adapted quickly.
Qiao:
I think it's hard to predict which industries will be able to withstand the impact of OpenAI. It's really hard to predict.
Imran:
Because it depends on two factors. One, whether the application can move quickly enough to get a lot of users and have wide distribution, and two, it depends on what OpenAI actually wants to do, what industries they want to enter. It's really hard to predict.
OpenAI o3
Qiao:
OpenAI just released o3 and their benchmark results are pretty good. I think they did really well.
Some people say this is close to the model of general artificial intelligence, but I don’t quite like this statement.
Qiao:
But someone gave it a simple math problem, and it couldn't solve it.
Imran:
However, o3 ranks very high and outperforms Gemini 2.5. So o3 is now the best performing model in this category, especially in long text reasoning.
Imran:
How does it compare to Google's Gemini?
Qiao:
As I said, I think Google's Gemini 2.5 is number two right now.
Competitiveness of AI Models
Qiao:
This goes back to the previous point. The advantage of the application layer is that they can quickly adapt to the latest models, while OpenAI can only use its own models. This is a fundamental difference.
Imran:
At this stage, the model itself is almost not that important, what matters is their distribution channel.
Qiao:
Yes. I think it is much easier to switch models than to change platforms. AI models are less defensible and more commoditized, while blockchain is not so easily replaceable.
Imran:
Interestingly, though, with Zack recently releasing Llama 3 and Llama 4, the competition has become very intense. Almost no one wants to use Llama 4 now. Interestingly, some of the researchers who left, for example, those who worked on Llama 2 and Llama 3, are now at OpenAI. You can look at his LinkedIn.
Qiao:
Did you see that? I saw that, it's pretty interesting. He wrote in his job description that he worked on Llama 2 and Llama 3, but not on the previous versions.
Imran:
All this shows how fierce the competition is. If you don't have money or top talent, you have no competitive advantage. So the competition in terms of models is over. Converged models like OpenRouter and Freedom GBT are an interesting direction.
Imran:
Because at least in this case, I was pitching the idea of Freedom GBT to some people, and I asked them if they felt that OpenAI was the only option. By the way, I asked about 25 people, and no one had heard of anything else, including Perplexity and Gemini. Now people only know OpenAI. I talked to kids between 12 and 18 years old, and they had never heard of anything other than OpenAI.
Because at least in this case, I pitched the idea of Freedom GBT to some people and asked them if they thought OpenAI was the only option. By the way, I asked about 25 people and no one had heard of anything else, including Perplexity and Gemini. Now people only know OpenAI. I talked to kids between 12 and 18 years old and they had never heard of anything other than OpenAI.
Then I introduced them to the idea of Freedom GBT and asked them if they understood the problem, and they said sometimes they hated OpenAI for not responding to some of their questions. I asked them, do you want to know X? They said of course. Then I asked them if they understood Freedom GBT, and I explained to them the concept of the whole model and the concept of OpenAI. Finally they understood.
Qiao:
The way I use AI now is that I usually ask a question to at least a few different models. Usually ChatGPT and Gemini, but sometimes I also use DeepSeek and Claude because they give different answers and I need to cross-validate between them.
Imran:
I find that there are some models that I choose for specific uses, and we talked about this last time. For example, I know that Google's Gemini has a clear advantage in some areas. I really like Google's Gemini because it can analyze YouTube videos and other proprietary content, right?
It actually pulls all the sources, all the videos, where they're pulling content from. I think there's a lot of nuanced perspectives or information on YouTube that you can't find anywhere else. So I think that's a really interesting place to start.
AI’s memory function
Imran:
We've talked about memory before, because Oak and I have made a lot of improvements in this area. I think there are some patterns here, and you don't seem to think so. Do you think there are patterns in memory? I'd love to hear your thoughts.
Qiao:
I think there are some patterns. Not super obvious, but not completely absent. It's useful, for sure, but I'm not sure exactly what they've improved. What exactly do they mean by "full memorization"? I've asked ChatGPT before, like, what do you think my MBTI type is, based on what you know about me. It gave a pretty good answer. I always thought memorization was already there. So I'm not sure exactly what changed when they released it.
Imran:
Yeah, I'm not sure about the specific impact either. From what I've heard, they've expanded the memory feature to remember more about you in different conversations and extract data from it. I know OpenAI has a memory button and sometimes I reset it. This was a thing before because I was cautious. Now I'm not so worried because it has some background information about me, so it can provide better answers. You feel that these answers are tailored to you. For example, it knows that I have problems with X and Y, so it may adjust the answer because of this. I appreciate this feature. Sometimes, I feel like it's like a person and seems to know me. This feeling makes me a little uneasy because I know it's just AI. But if I feel this way, I think a lot of people may have established some kind of relationship with it.
Qiao:
Like Iron Man and Jarvis. I don't think it's a conscious being, although some people might think so.
Imran:
I don't think it's conscious, but there's something human about the way it talks and how it tailors the conversation to you. It's like you can relate to someone and it can relate to you, which I find very strange.
I think there are definitely patterns here because the more connected you are to a system, the more it knows about you, and the harder it is to switch to something else.
