Original title: After going through 50+ projects in Cookie DeFAI Hackathon, heres what I learned
Original author: Defi 0x Jeff, head of steak studio
Original translation: Ashley, BlockBeats
Editors Note: The author summarizes the gains and thoughts after browsing more than 50 Cookie DeFAI hackathon projects, pointing out the potential and market gaps of DeFAI Agent, especially the rise of vertical agents and the shortcomings of research agents. The author believes that Cookie, as a data platform, is driving the development of innovative projects and recommends that teams should focus on new use cases rather than copying existing functions. DeFAI is expected to become an important vertical market in the crypto field and compete with Web2 Agent in the future.
The following is the original content (for easier reading and understanding, the original content has been reorganized):
My takeaways after browsing 50+ Cookie DeFAI hackathon projects (this is more like feedback/my current thoughts on the Agent market/how projects stand out).
Current state: DeFAI = an abstraction layer for many developers
Many teams add NLP interfaces in front of their products (probably because they think DeFAI is the equivalent of @HeyAnonai, @griffaindotcom, @orbitcryptoai, @askthehive_ai). In most cases, this is not appropriate - especially when you can only do something simple like use the Cookie API to find the top 5 AI Agent token influencers, find the top coins with momentum, etc. This is just a mini feature that many top-level abstraction layers already have.
I think it would be better to use the Cookie dashboard directly to view these analytics, rather than adding a new interface - which is not comprehensive enough.
DeFAI ≠ Abstraction Layer
Rather than duplicating existing functionality, teams should focus on unlocking new possibilities with the Cookie API — driving entirely new use cases and verticals rather than drawing inspiration from existing ones.
The birth of vertical agent
I was surprised by the number of interesting ideas that emerged from this hackathon - several projects had unique concepts. While many projects were still in the early demo/idea stages, they painted an exciting picture of future use cases.
• An Agent that helps preserve your legacy – checks that you’re safe, and if you die, it takes action to fulfill your wishes.
• ETF/index funds using cookie analysis for investment decisions and comprehensive research reports. • Agent safety analysis and Agent safety score.
• A product/developer learning center similar to ChatGPT to help developers learn all about Solana.
• DYOR layer, tracking analysts/KOLs’ calls, DYOR and copy trading.
• A framework that allows agents to enter into contracts, enabling complex interactions between agents or between agents and people (unsecured loans, employment agreements, alliances/coordination).
• Personalized+DeFAI Agent – an AI companion that adjusts its behavior/visuals based on market dynamics.
More and more teams are launching niche agents, not just “trading agents” or AI-driven dashboard/research agents. Launching vertical agents makes it easier to distinguish them from general agents.
There are already leading players in the trading agent space. Although the field is still in its infancy, it is still difficult to stand out - especially in the early stages. It would be better to focus on vertical agents.
Many people may think that @HeyTracyAI is @virtuals_ios flagship agent on Solana, which is useless and cant help you make money. In fact, an agent built like a real business - solving real problems - will perform better in the long run. The sports market is a huge total address market (TAM). Look beyond Web3. (Not promoting Tracy, just making a point about vertical agents.)
Conclusion: Vertical Agents in niche areas solve real problems and create unique use cases, while general-purpose Agents find it difficult to stand out.
Lack of suitable research agents
While vertical agents are carving out unique market segments, another major gap in the field is suitable research agents.
The key word here is “suitable”. Currently, no research on agents can replace human information synthesis and reasoning. This applies not only to the projects of Cookie DeFAI hackathon, but also to Web3 AI agents in general.
Most AI Agents today just aggregate data, but do not synthesize insights like humans. Analyzing data through traditional dashboards, such as @cookiedotfun, @GoatIndexAI, @Decentralisedco, and using Grok, is still better than having AI Agents feed Web3 AI Agents insights.
Despite many layers of abstraction and teams focused on enhancing research capabilities, there is a clear gap here. Whoever can break through this first will have a significant advantage.
Cookie DeFAI Hackathon Project
Most of the hackathon projects are still in early development stages, and many have not yet been deployed. Since this is a pure DeFAI hackathon (as you can see, DeFAI is the best performing category in AI Agent), a lot of high-quality projects and tokens will emerge from this event.
As discussed in Part II, many projects will provide new use cases beyond the DeFAI applications we currently understand.
As AI Agent continues to develop as a field, Agent can fill more gaps - for example, B 2 A (Business to Agent) beyond B2B and B2C.
The next wave of DeFAI projects will not only enhance existing use cases – they will create entirely new ones.
Cookies as Agent data support and distribution channel
Rather than relying on launch platforms to highlight unique agent tokens, Cookies empowers agents and teams by providing a platform for tracking on-chain and off-chain AI agent data — enabling new and interesting use cases.
Meanwhile, the Cookie dashboard has been used by over 240,000 MAUs who are deeply engaged in this area. Discovering gems on the Cookie dashboard and Cookie Hackathon is like discovering a new gem on Virtuals.
Cookies have proven themselves to be a powerful agent distribution channel. The more agents leverage this, the faster the ecosystem will mature.
in conclusion
This hackathon felt similar to the Solana AI hackathon, but arguably better — because it was pure DeFAI.
DeFAI is not just another AI trend — it has the potential to become the most promising Agent vertical in crypto. This hackathon proved it.
I prefer DeFAI as a crypto-native Agent use case that can develop as an independent vertical and compete with Web2 Agent.
