The following text is organized from the Twitter Space series #Dialogue with Traders, hosted by FC, founding partner of SevenX Ventures, Twitter @FC_0X 0.
Guest of this episode: Timo, AI investor, Twitter @timotimo 007.
To obtain Alpha, one should play with things they can understand.
In Web2, Timo is a VC specializing in investments in the first-level AI market, having invested in many domestic AI companies in different directions.
In Web3, his trading journey began in 2017 when he bought Bitcoin, and in 2019 he played with contracts, reaching a maximum floating profit of over ten million, but lost most of it in several liquidation events. Since early last year, he has been playing on-chain, engaging in inscriptions, token swaps, and pre-sales, and then moved on to zoos and AI.
Looking at AI targets is completely different from looking at those zoo coins; one should still play with things they can understand. If you just follow others blindly, you will only become someone else's liquidity in this market.
Timo's strategy for trading AI targets on-chain can be summarized as focusing on playing the second stage, which means looking for potential returns of over ten times and with a relatively high ceiling for Alpha, while managing position sizes so that single-coin investments do not exceed 15%. So far, this strategy has an above 90% win rate.
Aside from seeking Alpha on-chain, Timo's large positions in Beta are all placed on exchanges, including BTC, ETH, SOL, DOGE, PEPE, and AAVE.
What to buy: Which AI targets are worth paying attention to?
Overall, Web3 AI is actually following Web2 AI, from technical applications to narratives, and then to talent levels; thus, the trend of Web3 AI can largely refer to the development of Web2 AI.
Specifically, looking at the industrial chain, every link of Web3 AI has problems that need to be solved; where there are problems, there are new opportunities.
1. Protocol Layer
When the intelligence level of the Agent itself reaches a certain state, Agents should be able to interact intelligently across ecosystems/platforms, and this interaction requires underlying protocols. Currently, no one is doing this.
2. Model Layer
Currently, many models are general models that understand a little about everything but not deeply. Therefore, if you want a model to perform better and more professionally in a specific field, you need to train the model with proprietary data. Previously, there were some models specifically for Solana in the market, and it is likely that more proprietary models will emerge, presenting many opportunities. Additionally, the inference cost of models is also an area that can be improved.
3. Data Layer
Just as one needs to find high-quality learning materials, models also require high-quality data. However, most data is ineffective and needs to be filtered, cleaned, and labeled. A company in Web2 that specializes in data cleaning, Scale AI, is already valued at over 1 billion USD.
4. Tool Layer
Comparing the model to a human brain, it needs tools to complete tasks. Currently, many projects that want to develop agents face two issues regarding tools: first, whether they have the ability to intervene; second, whether the tools are willing to let them intervene. Therefore, Timo believes that if an open platform can be created where everyone can call various tools through this platform, it would be very promising.
5. Application Layer
In Timo's definition, the application layer is essentially various AI agents. In Web3, we need to focus on applications that can attract the attention of Web3 users and encourage them to use them frequently, which is also why many projects are talking about AI + DeFi. However, the current developments are relatively basic, remaining at fundamental and technical analysis of tokens, and cannot automatically execute desired trading strategies, so there are many opportunities here, but high-quality on-chain data is needed to train the model.
How to buy: How should one operate in the second stage?
The play of the second stage can be summarized in two steps: first analyze the fundamentals, then analyze the market conditions. Fundamental analysis determines whether a target can be bought, while market analysis determines where it should be bought.
Fundamental analysis can start from four aspects:
The first point is narrative, which essentially is the product's positioning.
Narrative determines a project's ceiling; a good narrative can attract enough funds in the market to FOMO. Timo suggests focusing on three types of narratives: 1) those that users are likely to use frequently, 2) those that can tap liquidity, and 3) infrastructure-related.
The second point is control, which is whether there are strong hands.
Why play projects with high control levels (strong hands)? In previous conversations with Mai, he explained in detail; in summary, strong hands have a relatively longer lifecycle and greater imagination space, making it a win-win game for both institutional and retail investors. Judging whether it is a strong hand is also very simple: go to GMGN to see the holding wallets; if the front row wallets are mostly small fish that bought early and profited a lot, they are generally all strong hands.
The third point is products and technology.
As the market matures, pure narratives are becoming hard to buy into; attention must also be paid to products and technology. For analysis in these two areas, Timo suggests that if you cannot understand it on your own, you can form a team with friends, where some look at code, while others are responsible for 'feeling' the chain; it is extremely difficult for one person to complete all analysis.
The fourth point is the background of the team.
Timo believes that if the DEV of Open AI can score 100 points, the average level of AI DEV in Web3 might only be 30 points. However, this industry is gradually attracting better talent, so it’s essential to look for 'regular troops' and thoroughly verify resumes and research achievements.
In fact, talking directly with the DEV team to obtain first-hand information will greatly help in assessing the fundamentals of the project; many DEV teams are willing to engage in this kind of communication.
Regarding positioning, in Timo's experience, many projects' second stages do not need to be rushed because there are usually at least two bottoms, typically when market makers are accumulating; generally, this is a reasonable time to build positions. Market maker operations can be monitored and tracked through their wallets.
When playing the second stage, how can one avoid mistakes?
Pay attention to two aspects.
1. Research must be sufficiently thorough.
When playing the second stage, there is plenty of time to analyze the project, so never be lazy. Additionally, Timo recommends collecting first-hand information, such as direct conversations with Dev, as this type of information is more valuable.
I believe that the biggest problem in the second stage is that the Dev does not proceed; you think they will, but they don’t, or they were never going to, but you never contacted them and don’t know the project’s situation. If you still go to buy in the second stage, that would be foolish.
2. Positioning and position management.
Always remember not to go all in. As Munger said, if I knew where I would die, I would never go there.
Additionally, building positions requires patience; do not rush to buy enough of what you want. At the same time, avoid so-called technical analysis, as the capacity of on-chain pools is limited. If a large holder suddenly dumps or tricks others, it can easily break the K-line 'support level,' but that doesn't mean you should enter at that moment. Focus on two points: first, have a thorough understanding of the fundamentals; second, observe the behavior of market maker wallets. Due to the high volatility on-chain, only by mastering these two points can you ensure that 'actions do not become distorted' amidst fluctuations, thus achieving a high win rate.
From Timo's own experience, most of the targets he buys initially incur floating losses, for example, buying 1% and floating a loss of 30%, but he analyzes thoroughly enough to accept it. 'Because I know it will definitely rebound, so I might not care too much about it.'
How to value AI targets?
The reason for asking this question is that valuation determines how to take profits. Timo believes there is no absolute answer or standard for valuation, and he attempts to classify it.
The first type is for a new target in an existing track, using a comparable valuation method.
Basically, for an already existing narrative/track, the market cap of the leading project in this track is the upper limit of all targets in this track, and the second leader is about 20% of the first leader's valuation. However, in the AI field, a new situation may arise where projects that come out later exceed all previous targets, in which case the principle of 'the second leader is 20% of the first leader's valuation' may no longer apply.
For example, Timo believes that AI targets in the current market only have some leading advantages and lack a moat. Suppose the core Dev of Open AI comes out to do a framework project, then it needs to be reassessed; the probability is high that the project's valuation will surpass that of ai16z.
The second type, for a completely new project without a reference, can only be valued through cognition.
For example, when AI16Z first came out, based on my understanding of this thing, I thought it was worth 1 billion; some people thought it was worth more. TRUMP is similar; some people think it’s worth 1 billion, others believe it’s worth 10 billion, and some think it could reach 100 billion, which is why they dared to take large positions when it was at 10 billion; this can only be said to be a matter of personal opinion.
For fast-moving markets, the most referable thing is sentiment, judging where the communication chain has reached through group chats and social media content.
Returning to how to take profits, Timo's method is:
Give three valuations for the project: the first stage valuation, the second stage valuation, and the final valuation. If the first stage reaches the target position, sell 30%, and when it reaches the second target position, sell another 30%.
Of course, this is a changing matter and cannot become a fixed standard. In terms of taking profits, everyone must do so based on their acceptance of the profit level and their acceptance of drawdowns. For most people, it might be better to sell early than to experience a drawdown.
Is the fundamental itself important, or is it more important what the market perceives it to be?
If it is an ultra-short-term target, then what the market thinks it is is more important than what it actually is, because when speculating on the first wave of FOMO, essentially pushing up the price requires attracting everyone's attention and liquidity.
However, if you are holding a target for the long term or medium term, you should be more concerned about what it actually is. Because the iteration of this market is very fast, if you can't keep up, you will quickly be disproven.
When playing the second stage, it is necessary to spend more time caring about a project's foundational capabilities and assessing its fundamentals, because filtering out some targets and missing opportunities must be accepted, and one must adhere to their trading system.
How to continue growing?
Timo raised four points:
The first point is to shrink one's ego. This viewpoint has been expressed by many, including Zhang Yiming, in different ways; essentially, it is about letting go of one’s emotions and external noise to see the essence of things.
Second, maintain curiosity and an open mindset. In the Crypto market, the things being played at each point in time are different; always keep a curious and open mindset to accept changes. If the market's hotspots change and mainstream trends shift, one must be able to quickly follow suit.
Although I still firmly believe that the main trend will be AI, if it's something else, I will also quickly change my own methods and strategies.
Third, learn to review and reflect. Don’t repeat mistakes; it seems simple but is actually very difficult to achieve.
The fourth point is to accept opinions and doubts; do not get entangled in whether to agree or disagree; it is more important to absorb some good external things.
The three people and three books that have had the greatest influence on Timo.
The first is Mises, the founder of the Austrian School of Economics, who helped Timo understand the relationship between monetary cycles, government, and human economic behavior during his university years, aiding his understanding of the world.
The second person is Duan Yongping. 'A person can continue to win in different fields is absolutely not luck but ability. One of his sayings is to do the right thing and stop immediately if wrong. I think this is far more important than being smart; being hardworking is more important.'
The third is Taleb, the author of (The Random Walker). 'Many events are actually random; essentially, no one has assigned the so-called causal relationship to them. I think the core is to respect objective laws, respect probabilities, and also respect black swans.'
The three recommended books are: (The Random Walker), (The Crowd), and (Built to Last), among which the third book greatly helped Timo, who was investing in Web2 at the time, understand what a lasting company is.
In conclusion
Timo mentioned that he spent a long time learning how on-chain actually works from last year, covering everything from dev to narratives to market making, and understanding the market rules before participating is very important.
I completely agree.
Previously, when I was playing games at a VC friend's house, he also told me something similar: first determine the game designer's objectives before starting to play the game. You need to know the game rules and understand how different aspects think and operate to play well; this is also a crucial reason why I engage in dialogues with traders.
Thanks again to Timo for participating in the dialogue with traders; previous Space audio will be updated on Xiaoyuzhou, search for 'Dialogue with Traders.'