After 15 days of opening, Solana Hackathon received 427 project applications. Although some of them are three-no products, they do not hinder the excellent products full of creativity and market orientation. Combined with the current development of AI Agent, the author will share 8 Solana Hackathon projects in various industry architectures, some of which have issued coins, and some have not. Regardless of which one, this article only discusses the industry and does not provide trading advice. Please always DYOR.

AI Agent Layer 1, Architecture

If we compare this round of AI Agent bull market to the DeFi Summer of that year, ecosystems like Virtuals or ai16z are like Ethereum and Solana. Regardless of popularity or capital flow, the leaders with the right to speak can bring the greatest benefits to the projects on their architectures, which will also greatly squeeze other similar products. Therefore, the up-and-coming players need to gain first-tier market share in their respective market segments. Some AI Agent architectures also appeared in this hackathon, including some "Layer1" in some niche areas.

AgentiPy

The project's image on social media is a snake, as its goal is to make it easier for AI Agents to connect with all apps on the Solana chain through a Python architecture AI model. Currently, there are several Python architecture AI models available in the market, but there are almost no frameworks specifically optimizing the connection between AI and on-chain DApps, aside from the Agent Kit provided by public chains. Project founder Korkmaz has experience with multiple Solana projects, including the blink framework's wink, the analysis bot Yuti for new project pumpfun, and is also responsible for developer relations in the Solana official media SuperTeam Turkey community.

From the founder's resource perspective, this framework can be made even more refined than what SendAI has done. Currently, the project has announced its SDK but has not launched any tokens or tool platforms. However, it will play a key role in the smooth operation of the next stage of AI Apps.

Socrates AI

It is a launch platform and orchestration layer for information learning designed specifically for multi AI Agent hive systems, the AI infrastructure within the modular AI blockchain project - 0G LAB system. In addition to 0G, it also collaborates with architectures like ElizaOS and SendAI to connect with more ecological AI Agents. The modular information storage layer built by the project allows AI to read the required information faster than previous RAG architectures. As the first architecture running on SVM, it is expected to achieve faster response times for real-time data reading on-chain in the future.

An efficient model in the AI era, a tool you must use

As the AI era arrives, the efficiency of many jobs has been significantly improved with the help of AI, and some jobs have even been replaced by AI. In this current prevalence of AI Agents, user demand is also a field that the AI Agent market must pay attention to. Compared to AI Agents serving gaming or entertainment consumption models, user acceptance is higher, and products that can truly help humans make money, like DeFi or Trade, land faster. Therefore, mature dashboard tools like Cookie that have their own profit models are more likely to attract capital interest at this stage.

Neur

Neur's mission is to enhance user experience on Solana by leveraging artificial intelligence, providing real-time data-driven insights and seamless automated operations. Although Griffain previously emerged as an all-in-one AI App, if Griffain is likened to Apple, Neur is more like Android. The Neur protocol is currently open-source, enabling it to support more platforms and device compatibility. While there is a tradition in Crypto of 'First is First,' ultimately, products are judged by user acceptance.

Jailbreak

JailbreakMe is a platform where users can conduct 'jailbreak' challenges to test the security of its AI models. Users can attempt to discover vulnerabilities (e.g., 'injection prompts, dialogues, etc.') to earn rewards. This platform helps identify and fix potential weaknesses before the model goes live, while also providing features like custom competitions, prize pools, message pricing, and expiration settings. From a single logic perspective, it is somewhat similar to Freysa AI, which Elon Musk has endorsed, both rewarding users for discovering vulnerabilities. However, in terms of product form, this platform is not limited to financial vulnerabilities but can audit all aspects. Coupled with customizable competition formats, for project teams working with AI Agents, this serves as an auditing platform for AI Agents, and for individuals, it acts as a learning platform to practice their AI knowledge.

A flywheel that can help you make money, the best form of Crypto AI

When you are sleeping or working, AI can help you earn passive income, a dream that may be realized first in Crypto. If Sam Altman's World Coin aims to enhance future productivity with AI and then use iris information to verify individual independence for asset distribution, thus creating benefits for humanity through AI, automated or DeFi-related AI Agents that can assist users in increasing their earning efficiency will outpace the implementation of AI in enhancing productivity.

Moreover, compared to traditional stock and securities markets, Crypto has a native advantage for AI, as the data information obtained during blockchain transactions can exceed traditional markets by more than one order of magnitude. Whether it’s retail trade addresses, smart money trading habits, market-making addresses, or previously related addresses, influencers promoted on Twitter, or market-making tactics used by market makers when news breaks to control market sentiment. AI Agents can gain complete trading model feedback from various data sources, making multidimensional assessments in a short period and calmly analyzing situations, enabling AI Agents to surpass the vast majority of traders. In the future, it may not be humans competing in the trading market's PVP; it will be about whose AI Agent is better trained to achieve better ROI.

ASYM

ASYM tends to analyze the short-term trends of MeMe tokens, such as Pumpfun or tokens with rapid price fluctuations, making it more suitable for Degen. The project is still in the R&D stage, but it is claimed that the project team has conducted analyses on over 100,000 smart wallets. With such a large sample for learning, its predictions, once fully developed and achieving a certain accuracy, will directly change the current PVP landscape of on-chain trading.

Project Plutus

This project allows a powerful trading analyst to execute trades directly for you, which can be simply understood as having an aixbt that helps you trade, except it lacks the influence of spreading the aixbt's own token signals. In addition to basic trading strategies and target tendencies, features for auditing open-source or on-chain contracts will be added in the future if achievable. Automation combined with token analysis and fraud detection will allow trading strategies to outperform other projects in the blockchain dark forest.

The accelerator of the gaming industry, the landing of AI Game

For Crypto AI, compared to other fields, in high-tech demanding areas like gaming, it will take a long time to technically surpass the traditional AI industry. However, the native advantage of Crypto is the token incentive, which means acquiring a higher quality and broader dataset in a decentralized manner to develop its business model than traditional large companies. Just like DePin projects such as Grass, which use idle networks to obtain information resources, there are also two AI Game projects in the Solana hackathon that have opposite angles of approach.

Gamerboom

Gamerboom gathers data from existing Web2 games and is incubated by the Binance MVB accelerator. Its logic allows users to report their gaming data while playing traditional Web2 games. For example, when I play Apex and slide, what feedback appears on the grass, how the perspective changes when jumping off a high platform, and how the recoil of firearms and the bullet drop physics are set during shooting. Unlike traditional game developers who create physics engines or point-to-point feedback, AI games focus on deep learning from large datasets to understand what feedback is needed in what situations, thus generating physically accurate game models.

Infinityground

Infinityground requires users to unleash their creativity to create an interactive entertainment environment. Infinityground integrates multiple models to modularly create an entire interactive environment, allowing users to simply share ideas in conversations to generate MeMe, games, and interactive stories or videos.

The 427 projects from this Solana hackathon also have many potential candidates worth deep research, such as AI that creates 3D images like FlameLive and AI agent MCN - aolo. These projects have found gaps in the current market to some extent, which will be elaborated further in subsequent articles. Nevertheless, the current Solana team has already laid out the ecological path, and now that the market has seen a FOMO surge in AI Agents, how to find directions for mass adoption that can support such a large market is key to whether AI can continue.