After 15 days of open submissions, the Solana Hackathon received 427 project applications. While some of them were filled with rug pulls, it did not prevent the event from being filled with creative and market-oriented projects. With the current development of AI Agents, the author will share 8 Solana Hackathon projects spanning various industry architectures, some of which have already launched their tokens, while others have not. Regardless, this article is for industry discussion purposes only and does not provide any investment advice. Please always do your own research.

AI Agent's Layer1 Architecture

If we were to compare this AI Agent boom to the DeFi Summer of the past, ecosystems like Virtuals or ai16z are like Ethereum and Solana. Whether in terms of popularity or capital flow, the influential leaders can bring the most benefits to projects built on their architecture, significantly squeezing out other similar products. Therefore, newcomers need even more to gain a leading market share in their respective niche markets. This hackathon also saw the emergence of some AI Agent architectures, including some "Layer1" projects.

AgentiPy

The project is symbolized by a snake on social media, as its goal is to make AI Agents more easily connect to any Solana on-chain app through a Python architecture AI model. While there are already multiple Python architecture AI models in the market, apart from the Agent Kit provided by the public chain, there are hardly any frameworks specifically optimizing the connection of AI with on-chain DApps. Project founder Korkmaz has experience in multiple Solana projects, such as the wink framework of the blink, the Yuti bot analyzing pumpfun's new projects, and is also responsible for developer relations for the SuperTeam Turkey community of Solana's official media.

At the founder resource level, this framework can be even more finely crafted than what SendAI has done. The project has already released its SDK, but has not launched any tokens or platform tools yet. However, it will play a crucial role in enabling the smooth operation of the next-stage AI Apps.

Socrates AI

Socrates AI is a launch platform and orchestration layer designed for federated learning of multiple AI Agents, a modular AI blockchain project within the 0G LAB system. In addition to 0G, it also collaborates with architectures such as ElizaOS and SendAI to connect with a broader ecosystem of AI Agents. The project's modular information storage layer enables AI to retrieve the necessary information faster than the traditional RAG architecture. As the first architecture to run on SVM, this framework will achieve faster response times for on-chain real-time data retrieval in the future.

The Efficient Mode of the AI Era: A Must-Have Tool

As the AI era unfolds, the efficiency of many job roles has been significantly enhanced with AI assistance, with some jobs even being replaced by AI. In the current AI Agent-dominant environment, user needs are also a crucial area for the AI Agent market to focus on. User acceptance of AI Agents serving the gaming or entertainment consumption model is higher than those that truly help humans make money in DeFi or trading. Therefore, mature dashboard tools like Cookie with their own profit models are more likely to attract capital attention at this stage.

Neur

Neur's mission is to enhance users' experience on Solana through the use of artificial intelligence, providing real-time data-driven insights and seamless automated operations. While griffain as an All-in-One AIApp predecessor has appeared before, if Griffain is likened to Apple, then Neur is more like Android. The Neur protocol is currently open source, enabling it to support compatibility with more platforms and devices. Although there is a traditional saying of "First is First" in Crypto, ultimately the product's success depends on user adoption.

Jailbreak

JailbreakMe is a platform where AI models can be tested for their security through a "jailbreak" challenge. Users can attempt to discover vulnerabilities such as "injection prompts, dialogs, etc." to receive rewards. The platform helps identify and address potential weaknesses before the model goes live, while providing features such as custom competitions, prize pools, message pricing, and expiration date settings. From a singular logic perspective, it is somewhat similar to Freysa AI, which Musk has praised, as both involve obtaining rewards through vulnerabilities. However, in terms of product form, this platform is not limited to combating financial vulnerabilities but can review all aspects. Coupled with customized match formats, for the project side of the AI Agent, this is an AI Agent audit platform, while for humans, it serves as a platform to practice their AI knowledge.

The Flywheel That Can Help You Make Money, the Optimal Form of Crypto AI

While you are sleeping, AI can still help you earn passive income, a fantasy that may be first realized in Crypto. If Sam Altman's World Coin aims to enhance future productivity with AI and then distribute assets based on iris information to verify human autonomy, forming the role of AI benefiting humanity. DeFi or trading-related AI Agents that automate tasks or help users improve their earnings efficiency are landing faster than AI-driven productivity improvements.

Furthermore, compared to traditional stock securities markets, Crypto has a native advantage for AI, as the data and information available during blockchain transactions are orders of magnitude higher than in traditional markets. Whether it's retail trading addresses, smart money trading habits, whale wallet addresses, or previously associated addresses, influencers promoted on Twitter, or market-making actions when information is released. AI Agents can gather feedback from various datasets on complete trading patterns and make multidimensional trade-offs in a short period. With the ability to analyze calmly, AI Agents already surpass the vast majority of traders. In the future, it may not be humans PvPing in the trading market but rather comparing whose AI Agent is better trained to achieve a higher ROI.

ASYM

ASYM leans more towards analyzing the short-term trends of MeMe tokens, such as Pumpfun or tokens with rapid price fluctuations, making it more suitable for Degens. The project is still in the research and development phase, but it is claimed that the project team has analyzed over 100,000 smart wallets. With such a large sample for learning, if its predictions can achieve a certain level of accuracy upon full development, it will directly change the current PvP landscape of on-chain transactions.

Project Plutus

This project allows a super powerful transaction analyst to help you directly execute trading operations. It can be easily understood as having an aixbt that helps you trade, except without the aixbt's own token pumping influence. In addition to basic trading strategies and target tendencies, in the future, the project will add the functionality to review open-source or on-chain contracts, if possible. Automation combined with token analysis and fraud detection will enable trading strategies to outperform other projects in the blockchain dark forest.

Accelerator for the Gaming Industry, Landing of AI Games

For Crypto AI, in the gaming industry, a highly technically demanding direction, it will take a long time to surpass the traditional AI industry technically. However, the native advantage of Crypto is having token incentives, meaning that in a decentralized way, one can obtain a higher-quality and deeper set of data than traditional large companies to develop their business model. Like the Grass and other DePin projects, using spare network resources to obtain information resources, there are also two similar AI Game projects with opposite angles in the Solana hackathon.

Gamerboom

Gamerboom retrieves data from existing Web2 games, incubated by the Binance MVB accelerator. The logic behind it is that users can be rewarded for their gameplay data while playing traditional Web2 games. For example, when I play Apex and slide, what feedback appears on the grass, how does the perspective change when jumping from a high place, and when shooting, how is the physical engine of the recoil of the firearm and the bullet's impact point set. Because AI games do not operate like traditional game companies in making a physical engine or manufacturing feedback point-to-point, but rather deep learning in a large dataset to understand when certain feedback is needed under what circumstances, thereby generating a game model that complies with physical laws.

Infinityground

Infinityground requires users to unleash their creativity to create an interactive entertainment environment. Infinityground embeds various models and modularizes to create an entire interactive environment, where users only need to share their creativity in the conversation to generate MeMes, games, interactive stories, or videos.

The 427 projects in this Solana hackathon also have many potential targets worthy of in-depth research, such as FlameLive creating AI-generated 3D images, and MCN for AI agent—aolo, and more. These projects have to some extent found the gaps in today's market and will be further detailed in future articles. Nevertheless, Solana's team has already laid out the path for the ecosystem. After the recent market frenzy for AI agents, the key to AI's sustainability lies in how to find a direction for Mass Adoption that can accommodate such a large market volume.

欢迎加入律动 BlockBeats 官方社群：

Telegram 订阅群：https://t.me/theblockbeats

Telegram 交流群：https://t.me/BlockBeats_App

Twitter 官方账号：https://twitter.com/BlockBeatsAsia