Original author: Squid
Original translation: Luffy, Foresight News
After the market has experienced a round of sharp decline, lets examine where the real opportunities are.
In the past two weeks, DeFAIs market value has fallen by more than 50% compared to mid-January, and by 70% compared to its peak in late January. Data from @cookiedotfun.
Decentralized Finance (DeFi) is now a $125 billion market, and assuming DeFAI accounts for 10%, that means there is a 15x growth opportunity. In fact, I think the opportunity within the DeFAI market is even greater because the current pricing in various segments is not reasonable.
Recently, we witnessed an epic liquidation in cryptocurrencies, during which assets were sold off regardless of merit. Artificial Intelligence (AI) and DeFAI were among the hardest hit sectors.
Now it’s time to find the treasure, let’s dive in.
Why is DeFAI so popular? First, lets look at the combination of cryptocurrency and artificial intelligence from a broader perspective. The combination of cryptocurrency and artificial intelligence is a $10 billion market that is in desperate need of finding practicality, and DeFAI seems to be the area most likely to find practicality.
For most of these crypto proxies, the product is their token. Because of this, we see the highest product-market fit (PMF) in frameworks and Launchpads that make it easy for teams to launch tokenized social proxies. As token supply grows exponentially, this market is becoming saturated, and the demand for social proxy tokens and Launchpads will decrease, forcing the community to look ahead to new directions.
Skate in the direction the puck is going (go with the flow)
From a macro perspective, DeFAI is of practical significance. This field is the intersection of DeFi, the strongest application scenario of cryptocurrency, and one of the most unmet needs of artificial intelligence - finance.
That being said, just because this space makes sense at a macro level, doesn’t mean every project will work. Let’s dig a little deeper and see where the opportunities might be.
Looking at the recent DeFAI market graph, most of the launched projects have limited utility in DeFi. I believe this is due to the newness of the space and the time it takes to build projects.
People want to get their hands on tokens, and most of the projects that have been launched so far are aiming for the lowest hanging fruit.
Valuation graph, a snapshot of projects taken from @cookiedotfun earlier this month, now valued much lower
Most of the DeFAI market value is concentrated in the Abstracted category. Abstracted is mostly text-based alternative user experience (UX). This category is low hanging fruit because it leverages existing DeFi projects and large language model (LLM) application programming interfaces (APIs), which have limited logic and are mainly intent-based. Given the current number of projects and valuations in this space, it is likely that the winner has already emerged, and I think the opportunities here are limited compared to other categories.
I also encourage investors to try these products. Personally, I find that most projects in the abstraction space do not provide a good user experience.
Interestingly, this market structure creates some opportunities for traders…
DeFAIs valuation has dropped significantly recently and the naysayers appear to be proven right... Product development takes time and I think this valuation drop is mostly justified given the stage the project is in. That being said, as prices drop, people get bored and I believe there may be opportunities in the future for those who keep an eye on the market.
So, where are the opportunities?
First, I believe that most of the opportunities lie in tokens that have not yet been launched. That being said, these niches already exist and some projects have already been launched, and to seize the opportunities, you must pay attention to the market and be prepared. To help determine where to focus, lets review a framework for thinking about the DeFAI market.
Information Agent
This is the era of AI agents mining information and fulfilling simple intents.
Big language models are very good at processing text and communicating in text. Since most AI agents today leverage the APIs of existing models, their strongest use case is natural language processing (NLP). I won’t go into it in depth here because there is a lot of other material covering this, but here are two high-level areas:
Alternative UIs/Abstractions: These are chat-based UIs that aggregate and abstract away DeFi protocols and even blockchains. They can provide macro information, find projects, and perform simple actions. Like “I want to buy SOL -> Buy SOL with Jup”. Eventually, these tools will use existing DeFi primitives/aggregators and cross-chain bridges to perform actions. I guess the “winners” may have already emerged, and we will most likely see the market consolidate into a few platforms. Growth will be in improving AI and integration capabilities and expanding the user base, which is still an unproven market. Personally, I’m not sure if text-based interfaces can really improve the user experience of DeFi.
Analytical tools: These are auxiliary tools that help traders mine and process information. This is a diverse category, and some examples include code review, token analysis, and social sentiment analysis. Analytical tools will continue to grow in sophistication and play an important role in the evolving cryptocurrency/AI ecosystem. AI-driven analytical tools compete in the same market as traditional analytical tools. Overall, there is still a lot of room for cryptocurrency analysis to grow. I think niche analytical tools will capture more value than general analytical tools.
Notable projects:
Abstraction: griffain, neur, The Hive, Venice, etc…
Analysis: Cookie DAO, Kaito AI, Hiero Terminal, and more…
Smart DeFi
Move from information to action, being able to mine information and act on that information.
Currently, this is a $200 million market, and I think it will eventually account for the majority of the DeFAI market. Assuming the market size is $12 billion and Smart DeFi accounts for 50%, there is still a 30x opportunity in this segment. Again, I think the winner has not yet emerged...
Where is the utility (value)? Initially, utility will come from continuous monitoring and automation, allowing users to exploit small inefficiencies in the market that they might not otherwise notice or find too small to be worth their time. This market already exists. As the field develops, large language models will enable the DeFAI protocol to not only automate, but also adapt and expand the market to further improve returns. Over time, the returns will increase in sync with the development of intelligence, reasoning capabilities, and infrastructure.
Teams that succeed in this space will have to build or use custom models, DeFi infrastructure, and data pipelines. This need for deeper integration is why this space is underdeveloped compared to the abstracted space. Building these things takes time.
Smart DeFi is not a new market, but AI can enhance and expand it. Examples of existing products include yield optimization projects Lulo, Carrott, and aggregators Ranger and Jupiter.
Existing domains are where deterministic models excel. For example, “the cheapest rate between X protocols is Y, so use Z.” Eventually, operations based on large language models may help improve these protocols by providing alternative user experiences or enhanced information, but existing projects are unlikely to be disrupted.
How will Smart DeFi expand in this market?
Large language models are probabilistic. In fixed markets this can lead to poor performance, for example if you are comparing transaction prices, the cheapest is always the best.
The probabilistic nature and ability to process a variety of different types of information does provide advantages that could allow DeFi to expand into new, more dynamic markets.
Over 50,000 new tokens are minted every day. I know this is a bit old, but the trend is still going strong…
In the long tail of the market, volume and value are constantly shifting between new assets. This category is difficult to serve with existing DeFi primitives due to volatility. This market will continue to expand as we continue to “tokenize the world.”
Furthermore, tokens are not equity and tokens can be constructed/represent value in a variety of different ways. The high diversity of assets in the market is another favorable factor for intelligent systems.
Smart DeFi can help DeFi expand into this market for the following reasons:
Ability to discover/evaluate new assets
Dynamically monitor, understand and act on new market narratives
They can combine social, on-chain, and off-chain data and apply reasoning to take a variety of actions.
There are synergies at a macro level, but where are the opportunities?
Let’s look at a promising niche, Smart Liquidity Providing (Smart LPing).
Liquidity provision is difficult because each pool has a different risk profile and returns fluctuate widely between assets. Currently, there are tools to help configure assets based on the users investment philosophy, but the composability/automation is limited.
The vision of a liquidity provider is to be able to dynamically optimize the returns of liquidity provision by adjusting risk parameters and selecting assets/pools based on a combination of on-chain and social data. This may eventually create a new asset that can dynamically capture returns from the long tail market, thereby improving market efficiency overall.
Raydium’s transaction fees can be used as an indicator of on-chain transaction volume
Why choose the Native AI team?
Yes, it will take time to realize this vision. That being said, I believe native AI teams will be more competitive in areas where the market is less efficient/less absolute returns because more of the returns can be attributed to intelligence. Native AI teams will be smaller, so they can move faster and focus more on the intelligence factor.
Some teams I would like to mention: Cleopetra, Alris agent and Voltr.
Other areas I won’t go into include intelligent periodic investing, trade execution, and products based on social signals. As intelligence increases, the percentage of returns attributed to humans will begin to be eroded by AI tools.
Project Plutus has shown interesting early results with smart limit orders, and I’m looking forward to testing it myself.
Next up, Alpha Hunter
Let’s see if an agent can generate alpha.
Here are some common responses from some of the more professional people in the industry. While I agree we are still far from that goal, I think it is ignorant to completely dismiss Alpha Proxy. Something to believe in…
Generating alpha is hard.
In traditional markets, hedge funds spend billions of dollars every year working with the smartest people in the world to create an edge. But systemic frictions mean that crypto markets are far less efficient than traditional markets, which is why it is more feasible to generate alpha through proxy. New asset classes/narratives emerge every week, which means that crypto cannot directly apply the same strategies as traditional finance.
Why give up Alpha?
I firmly believe that Alpha Hunters will not be an innovation from 0 to 1, but will emerge gradually. For the foreseeable future, Alpha returns will still be driven by people and assisted by AI. Over time, returns will gradually be attributed to agents. Once a certain critical point is reached, we may see the emergence of true Alpha Hunter Agents that will employ or work with humans.
The token will help launch, coordinate and integrate the emerging ecosystem.
There are currently two protocols that take interesting approaches:
Almanak: Almanak integrates a boutique data pipeline, advanced risk engine, and AI-driven agents to generate and execute multiple financial strategies under human supervision. It is an agent/strategy platform that continuously identifies alpha returns and provides services to users in a secure non-custodial environment.
Allora: The Allora inference system intelligently aggregates and weights AI predictions based on real-time accuracy. Human participants act as workers, providing predictions and evaluating the quality of inferences to refine the system’s market forecasts. The goal is to apply these predictions to the market.
Crypto-native teams working on model development will also play a key role.Ultimately, intelligence is key.
Other protocols I like include: Pond (focused on developing crypto-native models using on-chain data) and Nous Research (they are doing a lot of work that could help develop the first optimal models for decentralized training, and they also mentioned the “need” for exploration agents).
This field is still in its early stages, so keep an eye on teams that have the potential to grow and develop.
As the marginal cost of intelligence continues to fall, and the number of new markets created every day increases, agents are poised to take on an emerging market segment that is too ephemeral for humans and too dynamic for robots. Believe in something.
