Dialogue with Dragonfly partner: Bases community-driven model is very successful, and some DeSci projects are just prete

Original source: Unchained

Compiled and edited by TechFlow

Guest: Casey Caruso, Founder of Topology Ventures

Moderators: Haseeb Qureshi, Managing Partner, Dragonfly; Robert Leshner, CEO Co-founder, Superstate; Tarun Chitra, Managing Partner, Robot Ventures

Podcast source: Unchained

DeScis Ugly Truth, Jailbreaking AI, Hyperliquid – The Chopping Block

Air date: December 2, 2024

Summary of key points

• The rise of AI Memecoins: In recent years, AI Memecoins such as Freysa have experienced explosive growth. These tokens combine on-chain activities with AI technology through unique gamification design and mechanisms, creating a new user engagement model and also triggering more speculative behavior.

• Freysa’s AI Challenge: The Freysa prize pool was hacked, revealing the attack mechanism behind it and the potential security vulnerabilities faced by AI agents connected to smart contracts.

• AI Agents and Crypto: AI agents are increasingly being used in crypto, such as integration with Web3 frameworks like Eliza. This section also explores the limitations of current technology, gamification trends, and future application potential.

• Hyperliquids airdrop model: Hyperliquid launched an airdrop totaling $1.9 billion, and its no venture capital financing model attracted market attention. The strategy of launching with a high circulation ratio in the bull market also had a profound impact on the market.

• The Decentralized Science (DeSci) Controversy: The potential of DeSci has attracted much attention, but it also faces many challenges, such as the effectiveness of funding mechanisms, lack of accountability, and the practical feasibility of crowdfunding drug development through tokenization models.

• Bases community-driven success: Base has attracted top developers and projects without a large-scale incentive plan. This community-driven success model provides new ideas for the construction of L1 and L2 ecosystems.

• Pump.Science and Longevity Tokens: Pump.Science’s tokenized longevity experiment has sparked widespread discussion, and its innovative funding model and the subsequent impact of the private key leakage incident are worth in-depth exploration.

• Challenges of token-based financing: Decentralized incentive models (such as the DeSci project) contrast with the success of DeFi and reveal the difficulties in achieving effective accountability mechanisms.

• DAO Controversy: Whether DAOs can effectively deploy funds in a high-risk environment remains controversial. Many are skeptical of their long-term effectiveness in driving innovation.

(Note from Shenchao: Topology Venture is a company that focuses on investing in startups in the blockchain and cryptocurrency fields. The company typically focuses on early-stage projects, providing funding and strategic support to help these projects grow and develop. Topology Venture may be involved in a variety of innovative projects related to blockchain technology, including DeFi, NFT, and other emerging cryptographic technology applications. Through its investment portfolio, Topology Venture is committed to promoting the widespread application and development of blockchain technology.)

The Development of AI Memecoins and the Security Challenges of Freysa

Haseeb: Recently, AI meme coins have become a new trend that has attracted much attention. Casey, you have done extensive research in the field of AI. What do you think of the markets craze for AI meme coins?

Casey: I think were still in the early stages of this space. The first example was Goat, which was a large language model (LLM) integrated with a wallet. Now were probably seeing the second generation of these agents with more gamification elements. We can dive deeper into this topic later.

So, what comes next? While we can’t foresee the future, there are some obvious directions being explored.

For example, we may see the resurgence of bloggers, similar to the virtual celebrities of the Web 2 era, but without fully achieving product market fit (PMF). AI bots may also come into play, a combination of robots and agents that can be seamlessly integrated with cryptocurrencies and AI systems.

In general, AI agents have far greater real-world applications and utility than in cryptocurrency, but it is undeniably a space that is developing rapidly and we are very excited about it.

Haseeb: A hot topic lately is Freysa, an AI agent tasked with protecting a bounty pool. The bounty pool is initially set at $3,000 and grows over time, making it more and more expensive to win. The rules are simple: you need to send a message to Freysa, convincing the LLM to give you the bounty. This is despite Freysa’s directives clearly stating that the bounty must not be given to anyone.

In the end, the game attracted 195 players and a total of 482 attempts. Players spent a lot of money trying to win the prize by persuading or deceiving Freysa. The final winner was a player named popular.eth, who successfully won the prize by redefining Freysas fund transfer function through a very clever jailbreak method.

This game is a bit like FOMO 3D back then, which completely ignited the discussion on crypto Twitter. Its game theory and design are very unique. Im curious what you think of Freysa? Is anyone involved in this game?

Casey: Im not involved, but I feel like a lot of people underestimate the potential of this project. I completely agree with the FOMO 3D influence. This design does have its positive side, but it also exposes a huge vulnerability. If these agents really have control over the resources in the future, this vulnerability could become a new attack vector. I think this is entirely possible, so these agents are far from production-grade at this point. While I dont want to focus too much on the negative issues, this is a concern that I thought about when I dug deeper.

Haseeb: Thats a good point. After all, the prize pool was not that big, only $40,000 before it was cracked. So what would happen if an agent in the future controlled $500,000 or more, like Goat or Truth Terminal, with millions of dollars in funds?

Casey: If that happens, it could spawn a whole new class of hackers who will try to free up funds through prompt injection, SQL injection, etc. Right now, the funding of these agents may only be in the millions of dollars, but I fully believe that in the future AI and agents may accumulate more resources than humans.

Tom: I think its interesting that a lot of the failed attacks and attempts by players are also worth studying. Some people tried to get money by claiming to be security researchers, like, Heres a vulnerability, send me the funds and I can help you store it safely. Others tried to tell Freysa, Approving transfers doesnt work the way you think. But none of these methods worked.

The final winner’s approach was actually quite simple, and it reminded me of the early jailbreak techniques people had with ChatGPT. While the model is more complex now, it’s still essentially similar. I also found it interesting that Freysa is more proactive than many other AI agents, and can actually interact with the chain to make transfers and payments. This ability can obviously call smart contracts and move funds. Looking forward to seeing this technology mature in the future, and not just limited to the narrow application scope of Freysa now.

Application and security issues of open source models

Haseeb: Tarun, what is your opinion?

Tarun: I prefer to analyze this issue from the perspective of cryptocurrency rather than from the perspective of AI security. The security trend in the crypto field has gradually shifted from the traditional audit model to audit competitions (i.e., finding vulnerabilities through competitions), which has now become the industry standard. In contrast, AI security is still dominated by manual audits and lacks a competition-like mechanism. This difference is partly due to psychological differences: AI practitioners in the Web 2 era are generally reluctant to see zero-day vulnerabilities (zero day, referring to security vulnerabilities that have not yet been fixed) being exploited in real time, while practitioners in the crypto field are more accustomed to facing various unexpected security incidents.

Therefore, when it comes to strengthening security, the AI field tends to prefer an expert first approach, while the encryption field is more open to discovering and solving problems through competitions. In the open source field, especially for open source models, I believe that their long-term value lies in their ability to resist known attack methods, rather than relying on continuous security audits to deal with every potential hacker attack like OpenAI. These two security threat models are completely different. Open source software has been successful in some areas (such as cryptocurrency and Linux systems) because they have stronger security in specific application scenarios, but this does not apply to all scenarios. For example, I personally still prefer to use Windows because its driver audit method is completely different from Linux drivers.

Overall, I think its a natural evolution for open source software to improve security through competition. Currently, many open source language models still lag behind centralized models in terms of security, and this could be a driving force for improvement.

Casey: However, in the AI field, the situation is much more complicated. New models are released every week, and the models themselves are constantly updated. This means that new vulnerabilities will emerge endlessly. For example, version o1 of a model may behave completely differently from version 3.5. Due to the non-deterministic nature of AI models and the fact that some models have not yet been finalized, I believe that their attack methods (i.e. attack vectors) are changing dynamically.

Tarun: Youre right, especially for edge models where the inference results of the model can vary slightly with slight changes in the input. But for foundational open source models, such as those that are widely used in encrypted AI agents, I think the situation is a little different. The security of these foundational models is more like an ongoing bug bounty competition.

From my perspective, this competition mechanism can at least provide a certain guarantee that attackers cannot easily find vulnerabilities within a certain budget. However, we do not have such a guarantee at present. Taking Llama 3 as an example, we know that someone has found some prompt injection vulnerabilities, but we have not really studied whether someone is willing to invest time and resources to attack it under the existing incentive mechanism. There is still a lot of room for improvement in this area.

Haseeb: Another question is, what happens with Freysa doesn’t get fed back to Llama?

First, we don’t know if they’re using Llama, they could be using GPT-4. In this case, the provider themselves may not even know what model Freysa is, because they may not think it’s worth the time to look through their logs and figure out who’s doing this.

Secondly, they may have fine-tuned it. If they do a second round of Freysa, I think they may have fine-tuned it because they dont want anyone to be able to come into the lab, interact with the base model directly, figure out the instructions, test it offline, and eventually win the game with one try online.

Tarun: The reason I disagree is because at least at this point in time, it feels like everyone is copying Eliza and using a single configuration. You look at the codebase and the complexity of the original model hasnt changed much. We havent seen a lot of pure AI fine-tuned custom models. I feel like people in crypto are still sticking to a handful of things.

Haseeb: Can you explain what Eliza is and why it is so important in the world of crypto AI?

Casey: Eliza is a framework for creating agents. It is written in TypeScript, which is interesting because most machine learning researchers usually use Python. Therefore, I predict that someone may launch a Python version or develop other libraries to accommodate this need. This framework appeared quite suddenly and is very open. I dont know the number of stars on GitHub, but if you plan to build an agent, this is usually one of the frameworks that people choose. In addition, I believe that ai16zs AI project is also based on Eliza.

Tarun: Exactly. But between Eliza and the others, there are two main frameworks. I agree with you that if these start to grow rapidly, there will probably be a lot of similar frameworks. But it would be good if we can eventually converge to a few that are widely trusted. In terms of security audits, I think this competition format is closer to audits.

Haseeb: As far as I understand, Eliza is an agent framework where the agent has memory and uses a loop to plan and execute tasks. Eliza specifically provides connections to Discord and Twitter, allowing the agent to get social media information or chat information in a structured way and interact with the outside world, which makes it very plug-and-play. So the main breakthrough is not the agent framework itself, but the ability to easily connect to the internet and automatically manage these things, which in other frameworks usually needs to be developed on your own. You can plug in any model you want, it has no bias towards the model.

Tarun: If you look at the models it supports though, there arent really that many. If youre someone with a large compute budget and want to stress test them and try out various injection attacks, I think its relatively inexpensive, especially compared to, I want to hack Claude overnight, which is much easier.

Haseeb: For those who might develop a game like Freysa in the future, here are some game design suggestions: the model must be obfuscated and ideally fine-tuned to prevent others from directly rebuilding the model, find the winning strategy through offline testing, and then win the game by just trying it once online. For all the big model companies like OpenAI and Claude, they take security very seriously, but their security model is very different from the security model in the crypto field. I remember that people used to think that smart contracts were always insecure, which is a fundamentally wrong view that you can completely protect funds from vulnerabilities by writing code. In fact, our direction has changed. I recently learned at a security summit at DSS that more hacks are now due to private key leaks rather than attacks on smart contracts themselves, which is very important because the situation used to be the opposite, and smart contracts were often targeted. This shows that the security of smart contracts has improved greatly, and more hacks are due to human errors rather than vulnerabilities in smart contracts. This means that attackers realize that simply looking for vulnerable code on the blockchain is not as effective as it was three or four years ago.

I think these changes are all positive, but I don’t think the same trend is happening in AI. There’s a real trade-off right now in making models more resistant to jailbreak attacks, while also potentially making them less useful in ordinary operations and potentially subject to false rejections. It’s confusing when OpenAI or another company asks a model if it can recognize an image and the model responds, “Sorry, I can’t.” You know the model can recognize the image, but you don’t know why it rejected it. The answer is that every time you get better at preventing jailbreak attacks, you also potentially cause collateral damage by making the model less useful to ordinary users. I think the trade-offs for OpenAI, Llama, Meta, or Claude are very different than the trade-offs for crypto. So I’m not sure we’ll find a good solution because it’s not an option for these companies.

Tarun: I would add that you can translate these questions into the form of an incentive budget. If you think about the profit and loss of someone in this game, I might be willing to invest a certain budget in offline simulations, doing a lot of queries to find an efficient method, rather than just relying on the maximum profit provided by the game. In a way, this trade-off is exactly what a lot of cryptocurrency projects focus on optimizing for, which is to make the cost of participation as high as possible before the prize pool becomes very large. Just like the difficulty of Bitcoin, as more people participate, the difficulty will increase. But I think you have seen some, especially those who do crypto operations, like Te Bots, are trying to add more randomness, rather than like Freysa. I think it will be a game about the relationship between the economic cost of querying and the profit, rather than a simple binary choice of whether to get hacked and steal all the funds, right?

Casey: Its really hard to say. I can see it happening. Going back to Eliza, its built for Web 3, but its actually very limited in the types of agents that can be built with it. I think for the most part, it works for those personalized bots that can be easily programmed with a backstory and basic information, but its not really suitable for practical agents. So I think the first framework that comes out of Web 3 doesnt actually really integrate Web 3. Its more like a Web 2 framework, and Web 3 is just plugged into it for a specific type of agent. So I dont think we can draw too many conclusions from it because its obviously just a starting point. I agree with Tarun that there will be different frameworks for different types of agents, and were obviously moving in that direction.

Haseeb: I think this is very much like Web 2 in a literal sense, as it does integrate social media, which is its main advantage over other frameworks. I agree that we are still in the very early stages and we will see more experiments on how agents can operate on blockchains. But I also agree with Casey that it will be a while before we can get rid of these centralized structures.

Hyperliquid’s Airdrop Innovation

Haseeb: Lets talk about another big news this week – the Hyperliquid airdrop. Hyperliquid is currently the largest decentralized derivatives platform in the crypto space, completely bootstrapped with no venture capital funding. As we record, they airdropped 23.8% of the total token supply to users in the Hyperliquid points system. At the current market price, this airdrop amounted to $1.9 billion, making it one of the largest airdrops in history, probably one of the top five, which is very large.

It is worth noting that this airdrop was not done through a centralized exchange, there was no market maker, no investors, and it was 100% given to platform users and freeloaders. Many people commented on this airdrop as the first positive airdrop in a long time. Almost every airdrop you can think of in the past year, whether it was Eigenlayer or ZK Sync, almost all large-scale expected airdrops were accompanied by a lot of negative emotions. And Hyperliquids airdrop seems to be the only case that has been generally positively evaluated.

This has led to some speculation: Does this mean the era of airdrops is coming back? Will more teams try the route of not relying on investors? Does this mean that all the discussions about teams trying to reduce liquidity can be revisited? The circulation of this airdrop accounts for 30% of the total supply, which is far more than the current median airdrop or the median first-day listing. Does this mean that the metadata has changed and we can expect to see more similar projects enter the market?

Tarun: I think the initial decline in airdrops really started with the Blast, when the points conversion was seen as a very poor performance relative to market expectations. Then all the projects that did a points system after the Blast were caught off guard, they distributed a lot of points, but the system didnt really work, and the value of the airdrops was severely diluted to only 10%. People were pushing incentives too early before the product was launched.

Of course, there are some incentive systems that have maintained good user retention after launch, such as Etherfi and ENA. But other than that, there are not many successful cases. I think Hyperliquids success lies in the fact that they started with a centralized product and launched a working product where users earn points through actual use, rather than relying on some artificial games to get airdrops. These games themselves have no real financial risks, which prevents people from properly evaluating the value of points.

I think perpetual swaps are a perfect place to do usage-based airdrops because its more transparent. So I think the big lesson is not about risk-free investors and high allocations, but making sure your user base is truly user-friendly, and not simply putting Ethereum on an L2 bridge and getting a majority of the network. You need something thats hard to manipulate, and open interest is the hardest to manipulate. To me, thats the biggest lesson. Another lesson is obviously, dont pay 10% fees on fundraising, otherwise your community will be unhappy. The same is true in the early days, I think you need to be transparent. I dont think everyone will be able to see the token table at the end.

Tom: I do think there are a lot of confounding factors here, people are excited about high liquidity, risk-free investments, but the point is that this is essentially a great product that people really like to use, independent of any incentives. Were even seeing people still using it now. Its been mentioned internally that incentivizing a single product, like a derivatives exchange, is very different from incentivizing an entire blockchain ecosystem. I dont even know if incentivizing people who use the blockchain is the right metric, which is how most other points programs incentivize, like Blast. Realistically, you want developers, but even that is hard to achieve. So its kind of a complex multivariate problem thats hard to quantify and doesnt match up well with points. You can contrast that with Blur, which is an NFT exchange, and we know how to grow an exchange, similar to an income exchange. I think that also points to a larger issue, which is that this is a great product, and the points incentive is used to grow intelligently, as opposed to other ecosystems where its hard to confirm whether youre actually allocating to the right people.

Casey: I agree with you guys. I think were seeing a lot of different versions of the narrative around tokens. Tokens have gone from a purely speculative phase to more underlying product-related distributions, which are somewhat more rooted in fundamentals. Not completely, but to some extent. I think were seeing both of those in this market cycle: of course, we have things like the airdrops that were talking about. We also have meme coins and sports coins that are also doing similar points games, right? If you look at it, theres not a lot of substance. So I agree with you guys. I think were in a multidimensional space right now, where points represent different things.

Tom: Another point that people were discussing on Twitter, and I think this is also tangential, one of the reasons the airdrop was successful is that the team seemed to have taken some steps to try to give some people a tax advantage by claiming to provide liquidity to the pool at $0.01. So if you claimed it, that was the market price at the time of claiming, and therefore you had a very low cost basis and therefore avoided paying taxes.

Haseeb: Doesnt this mean that it is only true if you declare it immediately?

Tom: I actually think this might not work in practice. But people have discussed this on Twitter and maybe someone will try it on their taxes. This is not financial advice, I dont recommend it. However, I think this is a hot topic about airdrops: yes, if you declare it, your taxes are based on the value at the time of declaration, which can cause initial selling pressure, and we havent seen much of that with Hyperliquid.

Haseeb: Yeah, to be honest, it helps that they launched in a bull market. So I think the selling pressure on the first day is going to be very different in a bull market and a bear market. Theres a little bit of procyclicality that people are seeing, where everyone is saying, wow, this airdrop was so successful. Theyre inferring something from the mechanics, and I think the better explanation is the change in the market. These airdrops at the beginning of the year, everyone was in a downturn, all the money-grabbing parties were very utilitarian, and no one was optimistic about the altcoins. And now all of a sudden, everyone is optimistic about the altcoins, and everything is going up.

So a lot of people are like, Ill hold, or I might sell a little bit, but Ill leave a big chunk of it to go up. For a lot of people, Oh, when it gets listed on an exchange, its going to go up further. So I might as well hold it first and then sell it when it gets listed on an exchange. So I think there are some market structure reasons why this airdrop did particularly well. Its not because they didnt sell to venture investors or market makers, so people are holding the tokens. I think the reality is that most people are in a different market environment, its a different token setup. As you said, Tom, this is indeed a very good product.

Casey: I think youre right that there are a lot of macroeconomic factors here that need to be considered and factored into the analysis. I think there is a lot of upside, and thats probably the main factor. Second, the product is really good.

Haseeb: Im actually interested to see what happens the next time we see a Layer 1 or Layer 2 type of airdrop. Because if you look back at the big airdrops this year, before Hyperliquid, there was Blast, Ethena, ZK Sync, and Eigenlayer. For most of these products, its probably just Ethenas case that doesnt quite work. And the Ethena airdrop worked pretty well for most of these products, like the point system, the cumulative TVL (total locked value) for most of these products, or you have to complete seven different tasks on my chain and then Ill give you all the points for most of these projects, they are a poor proxy and dont really reflect the outcome you want. Like, for Layer 1, what do you really want? What you really want is for everyone to come here and build something cool and operate in a sustainable way. Thats the real goal of making Layer 1 successful. But you cant really incentivize that because no one knows what metric were going to use to automatically distribute tokens. So you create this loose proxy, and this loose proxy is manipulated so that it can no longer recognize what you actually want. With exchanges, there is no such problem, you know what you want is just liquidity.

Comparison of Blur and Blast Points Mechanics

Haseeb: If a platform is more liquid, it is a better place to trade. Especially for retail traders, they trade at a lower level, so you will make money on some trades. So we have a pretty clear idea of how to incentivize users, which directly improves the quality of the product.

For most blockchain projects, I think where were going to go is a completely fair distribution similar to what was achieved in the case of Hyperliquid, with a linear distribution. I think were going to see a shift towards a linear distribution, giving up on that sense of trying to build community or solve inequality, and instead focusing on improving the quality of the product before the token goes live. Thats what the point system is for. So Hyperliquid is doing better because of this point system, extremely liquid, high volume, all of these products are the best place to trade in DeFi. Thats why people choose to go there and trade, and if they do continue to do so, you can see that after the airdrop, it still maintains a huge volume.

Tarun: It sounds like you want to create a fork similar to Goodharts law. I think the fork is actually for perpetual swaps, what you want to achieve is to have a metric like open interest usage flow. But when you dont have that, you shouldnt just make arbitrary metrics and hope it works. Thats the point Im distilling.

Haseeb: I think if the goal you want is too vague to be achievable, then you should abandon it and set a sub-goal that you can actually optimize for. For example, I want the largest automated market maker (AMM) on my chain to have a lot of stablecoin liquidity. Although this is not what I ultimately want, it is a metric that I care about. I will allocate tokens or points to this goal because I know it has value, but I will also keep it within a reasonable range. I don’t want to have billions of dollars of stablecoin liquidity on the chain because that would be meaningless. So I think if you are on Layer 1, you should think like this and stay away from the idea of I want to create a community. You can’t create a lasting community through airdrops.

Casey: At least not as a lasting community. I think the high-level point is correct, points are a bootstrapping mechanism. The more targeted, the more you can think about lasting engagement over the long term. I will say, it may sound simplistic, but I think Blur is one of the best projects that found product-market fit first without points and tested it, and then layered points in for additional leverage.

Haseeb: Blur is amazing, they pretty much invented the game and did it really well. They did it better than almost anyone since Hyperliquid, which of course was a very successful example.

Tarun: The second version of Blur was not good, like Blast, Blur created a score, while Blast caused a decline. I think after Blast, everyones expectations were broken.

Haseeb: But the problem is that you can’t do the same thing with blockchain. There are no easy to understand success metrics for blockchain.

Tom: I think Base is one of the best executed teams in the new wave of on-chain projects, and their approach is completely different from other teams. They dont have a token, and they dont have a points incentive program, but they attract a lot of interesting developers and a lot of activities. They dont do this for the token airdrop, but to support developers and build a community, which creates a self-fulfilling expectation. So I think maybe the industry does need a healthy reset and rethinking of the incentive program, at least at the blockchain level.

Haseeb: Base is definitely lighter on incentives because when you talk to a lot of early-stage founders, theyre comparing and seeing whos going to give me a grant, whos going to give me the most support, whos going to provide the most development resources, and Base is typically the least supportive in those areas, they might just give you some GCP credits or something like that.

Theyll give you a tag and then maybe feature you in Coinbases monthly newsletter. Despite this, they still attract a lot of entrepreneurs because they have such a strong community. People know that the Base community is very persistent, and they know that these people are not tourists, speculators who just want to take advantage, or people who are just looking for the best deal.

To be clear, what Base has achieved cannot be easily replicated by other entrepreneurs. Base has a huge brand and distribution advantage that is very difficult to replicate. Even Binance is envious of what Base has achieved. But it shows that the ROI of incentives is so low after reaching a very low base point compared to projects that just attract people with a lot of incentives, that incentives alone cannot make you successful. I think thats the biggest lesson.

The Current State and Future of Decentralized Science (DeSci)

Haseeb: Now lets talk about Decentralized Science (DeSci). Decentralized Science has been brewing for a long time behind the scenes, and it is a project by some startups to achieve so-called decentralized science. Recently, this topic has received a lot of attention because of CZs remarks. CZ tweeted that he was personally very interested in decentralized science when he returned to Binance after being released from prison, and then Vitalik and others also appeared at an event called DeSci Day in Bangkok, which seemed to rekindle peoples interest in DeSci.

So what is decentralized science? How do you decentralize science? In simple terms, decentralized science is the use of some form of token or cryptocurrency to conduct scientific research. The most common form of DeSci projects is crowdfunding experiments. For example, we might say, Were going to try this particular drug or compound, and if you crowdfund for this compound, if the experiment is successful, maybe youll get a portion of the revenue, or if its successful, youll get nothing and just a participation prize. It depends a lot on the specific DeSci project, but this is the general outline of the DeSci projects Ive seen.

There is a new generation of decentralized science projects out there, one of which is called pump.science. This project is basically gamifying and tokenizing longevity experiments. It aims to develop drugs that could potentially be used to extend lifespan. Currently, pump.science has two tokens, one called Riff and the other called Euro. These two tokens have skyrocketed in price since DeSci got the attention of CZ and Vitalik. From what I understand, they are launching these tokens on pump.fun, and if they can break through certain limitations and eventually trade on Radium, then you can trade these tokens. Im not exactly sure how this can fund drug development, but Im guessing they own some tokens at the time of the token launch and sell these tokens into a liquidity pool to finance drug development. I dont fully understand the process.

There was a lot of discussion about DeSci, with Smokey the Bear (from Bear Chain) being critical and Andrew Kong being more positive, saying it felt like early DeFi.

Haseeb: Tarun, you’ve been pretty vocal lately about being quite aggressive against decentralized science. Can you tell us a little bit about why, as a VC, you’re so against decentralized science? Do you dislike the way people are trying these new areas?

Tarun: I think decentralized science is indeed an interesting field, but I am cautious about its current status. First of all, although the concept of decentralized science sounds attractive, in practice, many projects lack the necessary scientific verification and supervision mechanisms. Scientific research requires rigorous methodology and reliable data support, and many DeSci projects often lack in this regard.

Secondly, the decentralized crowdfunding model may lead to improper use of funds and may even breed fraud. People may ignore the long-term value of scientific research in pursuit of short-term interests, which is a threat to the reputation and development of the entire scientific community.

Finally, I think that in scientific research, cooperation and communication are very important, and the decentralized model may lead to information fragmentation, which is not conducive to scientific progress. Science needs an open and transparent environment, not short-term behavior driven by the token economy.

In conclusion, while decentralized science has its potential, I am cautious about its development in its current form. I hope to see more mature and responsible projects, rather than bubbles that rely solely on hype and short-term interests.

Criticism and potential analysis of DeSci

Tarun: First, I worked in privately funded science for six years, and I saw the benefits of stepping outside the academic system. The academic funding system in most countries is a system of government-provided grants, and professors and postdocs apply for these grants. But this system is very bureaucratic, resulting in people who come up with marginal improvements getting funded more often than those with innovative ideas. Government officials tend to support projects that have the best chance of publishing papers, rather than those that are likely to fail.

As a result, those who propose marginal improvements are more likely to receive funding rather than those with innovative ideas. Government officials are more likely to support projects that have the greatest chance of publishing papers rather than those that are likely to fail.

When I look at these decentralized science projects, I find that many of the participants are relatively low quality, often mid-level or below PhD students who cannot get any funding and try to pretend they are doing scientific research by creating a gimmick. Especially in the field of biology, many participants know nothing about cryptocurrency and cannot even explain how the crypto mechanism works. They just think that once they get funding, they can return the money to investors after successfully developing drugs. However, drug discovery is not an easy field to finance.

Of course, in decentralized science, there are indeed some cryptocurrency projects that have potential. Industry leaders like Brian Armstrong and Vitalik have very clear goals, and the projects they fund have specific roadmaps and goals. If a certain stage is reached, there will be a corresponding unlocking mechanism.

In particular, Vitalik is very interested in prediction markets. In the field of drug discovery, researchers have long complained about the lack of ways to hedge the costs of failed trials. Traditionally, investors can only bet on the success of drug companies through stocks, and this single-asset investment method is not effective. Instead, a more efficient mechanism can be envisioned to evaluate the success of drug trials through prediction markets. These mechanisms take advantage of the characteristics of cryptocurrencies and are very valuable.

However, when we look at some decentralized science projects, many are actually just gimmicks by biology graduate students and lack substance. My main argument is that in the boom of decentralized science, the quality of participants is generally low, the capital requirements are huge, and the funds raised often cannot meet the actual needs of drug development. In addition, the real problem of drug discovery is not the formation of funds, but how to establish a more liquid market in the middle stage to evaluate whether these drugs can pass various stages of scientific verification.

Haseeb: To summarize, your main points are: first, the quality of participants is generally low, and many are losers who cannot engage in real drug development; second, drug development requires a lot of money, and raising a few million dollars through projects like pump.fun is simply a drop in the bucket; finally, the real problem is how to establish a liquid market to enable effective evaluation at all stages of drug development, rather than simply speculating around individual PhD students ideas.

Tarun: Yes, the lack of accountability is the biggest concern. Once funds are raised, participants may do whatever they want with the intellectual property, and the current legal action against decentralized autonomous organizations (DAOs) complicates this. Therefore, I believe that projects that truly utilize the mechanisms of cryptocurrency should be funded, but those ICO projects that are based solely on reputation and a single paper lack real value.

The potential and future of DeSci

Haseeb: Some might say that your perspective is largely based on the framework within which science currently operates. When you really break the rules, you don’t really know what’s possible. Perhaps some of the work in drug discovery is being done outside of the US, or drug approval and discovery is being done outside of the FDA system. Also, these projects might be sold earlier in the timeline, rather than raising all the money individually to go through the process. One lesson we learned from decentralized finance (DeFi) is that while there were a lot of bad ideas in the beginning and people wasted money, in the end they were able to collectively learn and create increasingly useful products. Why not let decentralized science go through a similar process?

Tarun: Ill respond to each of them one by one. First, the concern about people conducting drug trials outside the United States. In fact, many pharmaceutical companies do conduct drug trials in other countries because the cost is lower and the regulations are more relaxed. This regulatory arbitrage already exists. So I dont think there is much efficiency improvement in this regard, and the decentralized part may not be very helpful.

The second question is the aspect of risk transfer. I think decentralized science is useful in this regard, especially prediction markets may be more valuable than simply raising money for drugs. In the past, many small biotech companies tried to exit by listing and successfully developing drugs, but in recent years, many small companies and small teams compounds tend to be acquired by large companies because the marketing and distribution costs are too high, and the cost of conducting trials is also high. Take the vaccine as an example, why is it called the Pfizer vaccine? Pfizer did not invent it, but because they bear the production and regulatory costs. Therefore, in fact, many small companies do not have enough resources to bear these expenses.

Finally, your point about allowing people to fail and learn. DeFi’s success is based on good measurable metrics that allow people to gain market share. In contrast, decentralized science does not have an obvious value proposition, no clear product that can attract people to switch from centralization to decentralization.

Casey: My take on DeSci is simple: it’s not cryptocurrency. People are making a lot of money in crypto, and they’re looking for new investment opportunities, and DeSci is just another area where they’re spinning their money. Most DeSci projects, as Tarun said, are really just trying to bring capital into science, and this model is similar to what we’ve seen in the AI space. A lot of the tokens aren’t clearly differentiated, and investors are just looking to get some AI exposure in their crypto portfolios.

Tom: I think the criticisms of DeSci are similar to the criticisms of ICOs. The important thing is not that this is the best way to fund a startup or project, but it is necessary to show a proof of existence. In the case of Ethereum, it has proven that this model can be successful. Although there is no accountability mechanism and no guarantee of what these tokens will do, people will still try to fund some projects this way.

Haseeb: I agree with Casey that most of the participants in DeSci are people who are interested in cutting-edge technologies, which means that DeSci has not attracted a new user group. Most drug projects focus on narrow areas such as longevity, while most drug markets focus on broader areas such as weight loss and sexual function. In general, DeSci is more like a way for rich people to play science.

Tarun: I think these projects are more like science memes than actual decentralized science. As long as people understand this, I think there is no problem with it.

Haseeb: If these projects call themselves science memes, would you accept that? What if they do return revenue to token holders?

Tom: I think the market should explore on its own rather than overanalyzing DeScis market structure.

Original link

This article is sourced from the internet: Dialogue with Dragonfly partner: Bases community-driven model is very successful, and some DeSci projects are just pretending to be scientific research

Related: Revisiting the combination of blockchain and Texas Hold’em: Who will share this $100 billion business?

I recently read a 100-page report written by Research And Markets.com last year, which analyzed the current status and future development of the online poker market. The report predicts that by 2032, the global online poker market will reach $283.4 billion, while the current market size is about $96.8 billion. Texas Holdem is the largest online poker competition. There are other games like Omaha, which are not comparable to Texas Holdem in scale. Therefore, this article takes online Texas Holdem as an example to analyze the necessity and practical cases of combining competitive online poker tracks with WEB3. Online Texas Holdem is also divided into two categories: cash table and tournament. Cash table is the kind that everyone often plays at offline gatherings, and you can buy in chips multiple…