In this episode, Haseeb Qureshi, Managing Partner ad Dragonfly, reflects on the year 2022 in crypto investments and shares valuable insights and lessons learned. From the importance of diligence in research to avoiding unforced errors, Haseeb offers advice for navigating the bull and bear markets.

The podcast also explores the intersection of AI and crypto, discussing the potential for privacy-preserving machine learning, decentralized inference, and decentralized training. Additionally, the conversation delves into the concept of network equity and data sharing, as well as the evolution of crypto investment strategies from DeFi to real-world assets.

Finally, Haseeb examines the future of blockchain with a focus on proof-of-useful and physical-work protocols, and debates the importance of network effects versus better tech in determining crypto success. Whether you're a seasoned investor or just starting in the crypto world, this podcast offers valuable insights and perspectives on the current state and future potential of this exciting space.

​

**Dragonfly’s 2022 Introspection 🪞 (0:24 - 3:22)**

 

- “2022 was the year of macro.”

    - Macroeconomics effected nearly every different asset class (stocks, bonds, crypto, etc.)

- “Diligence really matters”

    - In a bull market, you can get away with less research - in a bear market, those mistakes will cost you

    - The main reasons for passing on certain investments came from issues find during deep dive research

- “Pretty much everything that did not make sense to many people turned out to be actual nonsense.”

    - It’s been easy to tell yourself that you may not understand a certain protocol because you are not smart enough, especially when token prices are going up

    - If you did not / do not invest in things that you don’t understand, you did better than most people and are in a much better position for the long-term

 

**Positive Aspects / Investments of 2022 🌈 (3:37 - 6:42)**

 

- Avoided most of the big blow ups of the year (Luna, FTX, BlockFi, Celsius, Genesis, etc.)

    - “Most of the big unforced errors of the year, we steered clear of.”

- Saw multiple theses materialize

    - Made bets on Layer-1s being where the most technological innovation would take place

    - Investing in zero-knowledge, specifically zk-EVM (large investment into Matter Labs - creator of zkSync)

- For most investments, not enough time has passed to know the outcome

    - Invested in Aptos, Axelar, Bitget

    - You can look at interim prices and volumes, but these are long-term bets (will be longer than 1 year to know if right or not)

 

**DeFi Investment Thesis 🏦 (6:47 - 10:00)**

 

- Have not been many breakout products gain traction / market share

- Many blue chip projects (DEXs, lending/borrowing protocols) are growing in volume, but not users

- Yield farming protocols have collapsed due to increase in interest rates

    - When interest rates increase, the “risk-free rate” becomes much higher than it used to be - meaning that you should not be taking on a lot of risk without getting a lot of reward

    - DeFi will be sensitive to interest rates but as interest rates trend lower, there will be a resurgence in DeFi

 

- We have seen resilience / promise in Decentralized Options Vaults, Liquid Staking, and on-chain Real-World Assets

 

**Real-World Assets on Blockchain 🌎 (10:01 - 16:26)**

 

- “People tend to have magical thinking about blockchain and real-world assets.”

    - People see the blockchain as a place where they can tokenize / sell anything - but there is still limited / no liquidity for it just because it is now tokenized on a blockchain

 

- The blockchain is building a parallel financial system

    - Lower barriers to entry, can programmatically provide liquidity, and is available for anyone in the world to use 24/7

    - Need to think about what people all around the world would want to trade - probably treasuries (one of the most liquid assets in the world)

    - The only on-chain real-world asset today, built to scale, is USD

 

- U.S. Treasuries seem to be the most logical assets to tokenize

    - On-chain treasuries lead to bonds, which lead to securities, etc.

    - Once these assets come on-chain, they will be met with a global marketplace of people who want to own / interact with them

    - Many of these assets will be KYC-only but will change the way that on-chain financial markets operate

    - Currently, there is no real access to a risk-free rate on-chain

    

- Blockchain has the potential to allow for non-US citizens to interact with US assets

    - Non-US users may be more interested in buying US assets (treasuries, equities, etc.) before buying crypto

    - But this will serve as an on-ramp for bringing people into crypto

 

**1 Million TPS by 2030? 🏃 (16:27 - 19:17)**

 

- We will not see a single blockchain capable of 1 million TPS

    - We can probably get to 10,000 TPS (Visa processes roughly 3,000 TPS but peaks at roughly 30,000 TPS)

    - Only possible for the most throughput-optimized chains (likely not EVM-compatible)

 

**Investing in Blockchains vs. dApps 🥊 (19:18 - 28:53)**

 

- Risk-reward is a function of price

    - You can get a great outcome investing in a dApp or an L1 based on what valuation you invest in

 

- Fat Protocol Thesis is more of an observation than a thesis

    - Purely an observation when it was written in 2016 - Ethereum was worth a lot, but nothing built on top of it was worth much

    - L1s are still worth much more than the applications built on top of them

    - “The question is: Is this observation a function of the fact that blockchains are still new and we are waiting for the applications that will become more valuable than the underlying blockchain or is this a long-term phenomenon?”

 

- “FDV is a little bit of a mirage”

    - Different than the market cap of a traditional, publicly traded company

    - Better to look at FDV as “the total market cap, including the unissued shares”

        - When you look at the market cap of a company, you kind of ignore unissued shares because you assume that those shares will only be issued in a way that will be valuable to all of the shareholders

    - FDV includes all tokens that have not been spent

        - There is a distinct difference between tokens locked in a treasury vs. tokens locked by VC investments - but this distinction is not made by the FDV number

        - Investors should essentially ignore the tokens held by a protocol’s treasury - assume that these tokens will be spent in a way that increases the market cap of a project

 

**Responsibility of a Protocol’s Treasury 🪙 (28:54 - 31:40)**

 

- If everyone believed that the treasury would malinvest its funds, the token holders should vote to abolish the treasury - token holders ultimately control the treasury

- By in large, token holders have not voted to get rid of the treasury, signifying that they believe the treasuries are important

- Treasuries are necessary because give DAOs the ability to fund new projects, give grants, etc. - when these outlets are not there, token holders reliably vote to create them

 

**Intersection of Crypto and AI ⚔️ (31:40 - 36:37)**

 

- Historically skeptical about the intersection of crypto and AI

    - Crypto will enable / make some things in AI but the right way to back these projects may not be through direct investment

    - Many of these projects will be nonsense - this is what happens when any trend gets popular

- Must be able to think about what makes sense and what the technology can enable

    - Privacy-preserving machine learning, decentralized inference, decentralized training are a few areas that are likely all hype and not feasible at the moment

 

- There are places in AI that will create new features for crypto:

    - Large language models - great for data analysis, information aggregation, software development, auditing, etc.

 

**Network Equity Through Data Sharing 🤝 (36:38 - 42:08)**

 

- Any individual piece of data is tiny

    - LLMs are trained on trillions of words of text  , most of which is open-source content

    - Data is valuable in aggregate but individually, data is invaluable

- Individual data, especially from specialized users, does become more valuable when training a model for a specific task

    - Must think about your data per use case being valuable, not necessarily valuable in general

- Large-scale data labeling may be an opportunity to introduce AI to blockchain

    - Easy to coordinate buyers who need data labeling services and sellers who can provide the service

- In the short-term, it will be difficult to get everyone to agree on providing data in exchange for ownership of a platform

    - Different users have different incentives / goals

 

**Proof-of-Useful / Physical-Work 🛠️ (42:08 - 51:07)**

 

- These protocols make sense but it is not obvious what it will take to make them work

    - Will the entrepreneurs do the ground work to outcompete their centralized counterparts and make their project work?

 

- The hard part about Airbnb / Uber is not the software - there are competitors in nearly every country around the world

    - Uber is so good for other reasons outside their software - their edge comes back to their network effects (recruiting drivers / riders everywhere, strong incentives, reliability, attracting right customers, good support, etc.)

    - Businesses like Airbnb are more support businesses than software

 

- DAOs suck at support and customer service

- Even if decentralized protocols enter the gig economies and take over, what is to stop the DAOs from raising their take rates?

    - Token holders of the DAOs may vote to change their take rate to whatever the centralized counterpart was at before

    - The only reason why DAOs would charge less than centralized companies is because they are competing with other services

- In reality, if service providers of a decentralized service platform were given token incentives, they would likely sell the tokens immediately as a source of extra income

 

**Rapid-Fire Round 💨 (51:08 -55:24)** 

 

**What is more important for a project’s success: better tech or stronger network effects? (51:15 - 51:27)**

 

- Stronger network effects

 

**If you had $10 million to invest in the 24 hours, what would you do with it? (51:28 - 52:01)**

 

- Pass

 

**Last ChatGPT Prompt? (52:02 - 52:48)**

 

- Posted a blog post from a company into GPT-4 to prove that it could write in better English than the authors

 

**Non-Web3 Hot Take? (52:49 - 53:31)**

 

- Effective Altruism - Haseeb is an Effective Altruist and still is after the collapse of FTX

- Has not changed mind about anything regarding the philosophy behind Effective Altruism

 

**Advice for Other Podcast Hosts? (53:33 - 54:27)**

 

- “Be yourself, don’t try to emulate other podcasts.”

- “Get as far away from the center of the distribution, if you want to attract a good audience.”

 

**Question’s Mehdi Should Have Asked? (54:28 - 55:24)**

 

- “If you weren’t investing in crypto, what would you be investing in?”

- “I would definitely be investing in AI right now”