DeFAI foot cut 70%, is there still an investment opportunity?

Author: Squid

Compiled by: Luffy, Foresight News

After the market has experienced a significant downturn, let's examine where the real opportunities lie.

In the past two weeks, the market capitalization of DeFAI has dropped over 50% compared to mid-January, and has fallen 70% from the peak in late January. Data source: @cookiedotfun.

Decentralized Finance (DeFi) is now a $125 billion market. Assuming DeFAI captures 10% of that, it means there is still a 15x growth opportunity. In fact, I believe the opportunities within the DeFAI market are even greater, as the current pricing across various segments is not rational.

Recently, we witnessed an epic liquidation in the cryptocurrency space. During this period, assets were sold off indiscriminately. Artificial Intelligence (AI) and DeFAI are among the hardest-hit areas.

Now is the time to seek treasures; let's delve deeper.

Why is DeFAI so popular? First, let's look at the combination of cryptocurrency and artificial intelligence from a broader perspective. The combination of cryptocurrency and AI is a $10 billion market that urgently needs to find "utility," and DeFAI seems to be the area most likely to discover it.

For most of these crypto agents, their product is their token. This is why we see that the areas with the highest product-market fit (PMF) are those that allow teams to easily launch tokenized social agents through frameworks and Launchpads. As the supply of tokens grows exponentially, this market is becoming saturated, and the demand for social agent tokens and Launchpads will decrease, forcing the community to look for new directions for development.

Skating to Where the Puck Will Be

From a macro perspective, DeFAI is meaningful; this field represents one of the strongest applications of cryptocurrency—DeFi—against one of the most unmet demands—finance.

That said, just because this field makes sense on a macro level does not mean every project will succeed. Let's explore further to see where the opportunities might lie.

Observing the recent DeFAI market landscape, most launched projects have limited utility in DeFi. I believe this is due to the novelty of the field and the time required to build projects.

People want to acquire tokens, and currently, most launched projects are targeting the easiest but least practical goals.

Valuation landscape, a snapshot of projects obtained from @cookiedotfun earlier this month, with current valuations significantly lower.

Most of the market value in DeFAI is concentrated in the "abstraction" category. Abstraction is mostly based on text-based alternative user experiences (UX). This category represents "low-hanging fruit" as it leverages existing DeFi projects and large language models (LLM) APIs, which are logically limited and primarily intent-based. Given the current number of projects and valuations in this field, it is likely that winners have already emerged, and I believe opportunities here are limited compared to other categories.

I also encourage investors to try out these products. For my part, 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…

Recently, DeFAI's valuations have plummeted, and the naysayers seem to have been proven right… Product development takes time, and considering the stage of the projects, I believe this valuation drop is mostly reasonable. That said, as prices fall, people will become bored, and I believe there may be opportunities for those who are paying attention to the market in the future.

So, where exactly are the opportunities?

First, I believe most opportunities lie in tokens that have not yet launched. That said, these sub-segments already exist, and some projects are already online. To seize the opportunities, you must keep an eye on the market and be prepared. To help identify areas of focus, let's review a framework for thinking about the DeFAI market.

Information Agents

This is the era of AI agents mining information and fulfilling simple intents.

Large language models excel at processing text and communicating in text. Since the vast majority of AI agents at this stage utilize existing models' APIs, their strongest application scenarios are in natural language processing (NLP). I won't delve into this here, as there are many other resources covering this topic, but here are two high-level areas:

Alternative User Interfaces / Abstraction: These are chat-based user interfaces that aggregate and abstract DeFi protocols and even blockchains. They can provide macro information, find projects, and execute simple actions. For example, "I want to buy SOL -> Buy SOL through Jup." Ultimately, these tools will use existing DeFi primitives / aggregators and cross-chain bridges to execute actions. I suspect that "winners" may have already emerged, and we are likely to see the market consolidate into a few platforms. Growth will be reflected in improved AI and integrated features, as well as an expanded user base, which remains an unproven market. For my part, I am uncertain whether text-based interfaces can truly improve the user experience of DeFi.

Analytical Tools: These are auxiliary tools that help traders mine and process information. This is a diverse category, with some examples including code reviews, token analysis, and social sentiment analysis. Analytical tools will continue to evolve in complexity and play an important role in the ever-evolving cryptocurrency / AI ecosystem. AI-driven analytical tools compete in the same market as traditional analytical tools. Overall, there is still significant room for development in cryptocurrency analysis. I believe that segmented analytical tools will be able to capture more value than generic analytical tools.

Notable projects to watch:

• Abstraction: griffain, neur, The Hive, Venice, etc.
• Analysis: Cookie DAO, Kaito AI, Hiero Terminal, etc.

Smart DeFi

Shifting from information to action, capable of mining information and acting on it.

Currently, this is a $200 million market, and I believe it will ultimately occupy a large portion of the DeFAI market. Assuming the market size is $12 billion, if "Smart DeFi" captures 50%, then this sub-segment still has a 30x opportunity. Similarly, I believe winners have yet to emerge…

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 otherwise overlook or deem too small to warrant their time. This market already exists. As the field develops, large language models will enable DeFAI protocols not only to automate but also to adapt and expand market scope to further enhance returns. Over time, the degree of return enhancement will be synchronized with the development of intelligence, reasoning capabilities, and infrastructure.

Teams that succeed in this field will need to build or utilize custom models, DeFi infrastructure, and data pipelines. This deeper integration requirement is why this field is underdeveloped compared to the abstraction area. 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 areas are those where deterministic models perform well. For example, "The cheapest rate between X protocols is Y, so use Z." Ultimately, operations based on large language models may help enhance these protocols by providing alternative user experiences or enriched 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 poorer performance; for example, if you are comparing trading prices, the cheapest is always the best.

The probabilistic nature and the ability to handle various types of information indeed provide an advantage, allowing DeFi to expand into new, more dynamic markets.

Every day, over 50,000 new tokens are born. I know this data is a bit outdated, but this trend continues...

In the long tail of the market, trading volume and value constantly shift between new assets. Due to volatility, existing DeFi primitives struggle to serve this category. As we continue to "tokenize the world," this market will keep expanding.

Moreover, tokens are not equity; they can be constructed/represent value in various 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 mine/evaluate new assets
• Dynamic monitoring, understanding, and acting on new market narratives
• They can integrate social, on-chain, and off-chain data and apply reasoning to take multiple actions.

From a macro perspective, there are synergies, but where are the opportunities?

Let's look at a promising sub-segment: Smart Liquidity Providing (Smart LPing).

Liquidity provision is challenging because each liquidity pool has different risk profiles, and yields fluctuate significantly across different assets. Currently, there are some tools based on user investment philosophies that can help allocate assets, but the degree of composability/automation is limited.

The vision for liquidity provision agents is to dynamically optimize liquidity provision yields by adjusting risk parameters and selecting assets/pools based on a combination of on-chain and social data. This could ultimately create a new asset that dynamically captures yields from the long tail market, thereby improving overall market efficiency.

Raydium's fees can serve as an indicator of on-chain trading volume.

Why Choose Native AI Teams?

Yes, realizing this vision takes time. That said, I believe native AI teams are more competitive in areas with lower market efficiency/absolute returns, as more of the returns can be attributed to intelligence. Native AI teams tend to be smaller, allowing them to act faster and focus more on the "intelligent" factor.

Some teams I want to mention: Cleopetra, Alris agent, and Voltr.

Other areas I won't delve into include smart dollar-cost averaging, trade execution, and products based on social signals. As intelligence improves, the proportion of returns attributable to humans will begin to be eroded by AI tools.

Project Plutus has already shown interesting early results in smart limit orders, and I look forward to testing it personally.

Next, Alpha Hunters

Let's see if an agent can generate Alpha returns.

Here are some common responses from more seasoned professionals in the industry. While I agree that we are far from this goal, I think it is ignorant to completely dismiss Alpha agents. Believe in something…

Generating Alpha returns is difficult.

In traditional markets, hedge funds spend billions of dollars each year collaborating with the smartest people in the world to create an edge. But systemic frictions mean that the cryptocurrency market is far less efficient than traditional markets, making it more feasible for agents to generate Alpha returns. New asset classes/narratives emerge weekly, meaning that cryptocurrency cannot directly apply the same strategies as traditional finance.

Why give up on Alpha?

I firmly believe that "Alpha hunters" will not be innovations from 0 to 1, but will emerge gradually. In the foreseeable future, Alpha returns will still be driven by humans, 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 humans or collaborate with them.

Tokens will help launch, coordinate, and integrate emerging ecosystems.

Currently, two protocols taking interesting approaches:

• Almanak: Almanak integrates boutique data pipelines, advanced risk engines, and AI-driven agents to generate and execute various financial strategies under human supervision. It is an agent/strategy platform that continuously identifies Alpha returns and serves users in a secure non-custodial environment.
• Allora: Allora's reasoning system intelligently aggregates and weights AI predictions based on real-time accuracy. Human participants act as workers, providing predictions and assessing reasoning quality to refine the system's market forecasts. The goal is to apply these predictions to the market.

Crypto-native teams dedicated to 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 have done a lot of work to help develop the first decentralized training optimal model, and they also mentioned exploring the "demand" for agents).

This field is still in its early stages; keep an eye on those teams with the potential to grow.

As the marginal cost of intelligence continues to decline, and the number of new markets created daily increases, agents are poised to embrace an emerging market sub-segment that is fleeting for humans and too dynamic for robots. Believe in something.