Solana's AI Surge Begins: Hackathon Highlights

Key Takeaways

• The Solana AI Hackathon, hosted by SendAI, offered over $275,000 in total prizes across seven tracks. It received an enthusiastic response worldwide, with more than 300 projects submitted.

• Participants could form teams individually or collaborate globally through the Telegram community. They were supported with various idea boards, resources, and an Agent Toolkit to assist with brainstorming and development.

• Numerous noteworthy concept-driven projects were observed across the tracks. With Solana's strong culture of nurturing grassroots projects and fostering collaboration, this hackathon is expected to serve as a catalyst for the virtuous growth of its AI ecosystem.

1. Successful Launch of the Solana AI Hackathon

Source: SendAI

The Solana ecosystem has officially set its sights on expanding its AI landscape. This journey begins with the Solana AI Hackathon, hosted by SendAI and supported by various leading teams within the Solana ecosystem.

The hackathon aimed to build AI agents and tooling for the Solana network. It ran from December 10 to December 23, 2024, and featured seven tracks: Autonomous Chat Agents, Social and Influencer Agents, Meme Agents, Agents Infra, Agent Token Tooling, DeFi Agents, and Trading Agents. Across these tracks, prizes ranging from $5,000 to $30,000 is awarded, with a total prize pool exceeding $275,000.

Source: SendAI

The Solana AI Hackathon garnered global attention, attracting participants from countries across the United States, India, Nigeria, Europe, and Asia. Participants could form teams independently or join global teams via the Telegram community. They were supported with a range of resources, including idea boards, and the Agent Toolkit, to facilitate the brainstorming and development process.

In the following sections, we will provide an overview of each track and explore some standout and innovative concepts of projects submitted during the hackathon*. For detailed information on individual projects, visit this link.

*Note: This curation does not include detailed technical due diligence and does not constitute any financial advice.

2. Hackathon Highlights

2.1 Track #1: Autonomous Chat Agents

2.1.1 Overview

Autonomous Chat Agents are designed to operate independently, crafting and executing actions suitable for various situations without relying on pre-defined scripts. These agents have a wide range of potential use cases, including customer support, educational and learning platforms, internal corporate operations, and medical consultations. In the context of blockchain, they are frequently observed providing market information / insights, tracking transactions, or offering portfolio management interfaces.

Due to the influence of platforms like ChatGPT and Perplexity, autonomous chat agents represent the most familiar and widely attempted form of service in the crypto industry. A notable example is aixbt, a general-purpose agent delivering insights into overall market trends. According to Kaito, an on-chain AI analytics and insights provider, aixbt managed to dominate mindshare on Twitter within just a few days of its launch.

Another category involves domain-specific conversational agents designed for specialized purposes. Many projects have shown a strong demand for agents tailored to explain their initiatives effectively. For instance, the decentralized federated learning platform FLock.io facilitates the development of AI models for domain-specific agents, such as Bitcoin GPT, Ethereum GPT, and BNB GPT, often requested by different projects - in some cases, these models are continuously improved by community contributions, showcasing an iterative enhancement process.

The critical factors in implementing such chat agents revolve around three key points:

1. Data Collection: What type of data is gathered as training material?

2. Data Filtering: How is the data filtered to ensure high-quality insights?

3. Update Frequency: What is the schedule for updating data collection and refining models?

These considerations play a pivotal role in determining the quality and effectiveness of the agents.

Given the familiarity and popularity of this track, it recorded the highest number of submissions in the hackathon, with 164 out of 345 total entries falling under this category.

2.1.2 Notable Conceptual Submissions

Source: beta.palet.app

Since the advent of the internet, search engines have made remarkable progress. However, they still face limitations in providing specific and detailed information tailored to the user’s particular context.

Palet is a search engine designed to utilize LLM-based tools like ChatGPT and Perplexity to deeply understand user contexts and deliver answers that are close to final decision-making. Palet stands out for its focus on optimizing and streamlining the iterative processes humans typically go through when searching for information—such as repeated searches, assessing relevance, and conducting additional queries.

Other notable projects include:

• Emerald : A tool that enables anyone to create an AI agent within 5 minutes, without any coding.

• Sol.AI : A tool that helps users easily build dApps in the Solana ecosystem without requiring coding skills.

2.2 Track #2: Social and Influencer Agents

2.2.1 Overview

This track shares similarities with the Autonomous Chat Agents track mentioned earlier, but it focuses on developing AI agents that emphasize visual elements for more intuitive information delivery or prioritize real-time interaction with users.

Source: Virtuals.io

For instance, agents like Luna from the Virtuals Protocol or AVA from Holoworld AI leverage features such as autonomous decision-making, multimodal interactions, and on-chain wallet operations to engage with users. Additionally, these AI agents can be co-owned by fans in the form of tokens, allowing fans to participate in the governance and value creation of the agent, fostering a sense of shared ownership and responsibility.

The experimental concept of ‘Swarm’ is also noteworthy. It takes the form of a platform designed to enable not only interactions among people but also interactions between people and agents or even between agents themselves, each with their own unique characteristics.

Given the significant impact of social and influencer agents on the crypto space—whether through intuitive visuals or enhanced interactions with people—this track drew considerable attention during the hackathon. Out of 345 submissions, 125 projects were dedicated to this track.

2.2.2 Notable Conceptual Submissions

Source: Predo

People gain more than just simple enjoyment from placing bets. The thrill of uncertainty in outcomes provides a sense of excitement often lacking in daily life, and this experience fosters a bond among participants as they exchange diverse opinions throughout the process.

Predo facilitates betting among users through an AI agent integrated with Telegram. As demonstrated in the demo video, bets can be set up with a single chat message, and payouts are instantly settled via pre-configured wallets based on the outcome of the bet.

Other notable projects include:

• Digimon: A platform where AI agents evolve over time or under specific conditions, forming groups that interact with one another.

• RPS LIVE: An AI game-streamer agent platform that broadcasts and reacts to gamers' gameplay.

• SIAMES: A platform enabling existing social and influencer agents to interact within virtual environments.

2.3 Track #3: Meme Agents

2.3.1 Overview

Memes are undeniably an integral part of the crypto industry. While some may overlook them, for others, memes provide a source of entertainment and serve as a foundation for forming communities that share emotions and experiences built around them.

Creating memes with AI technology doesn’t require any particular technical expertise or advanced tools. As a result, projects leveraging AI for meme-based approaches are commonly observed. However, dedicating a separate track to memes in a hackathon is quite an intriguing approach.

The AI Meme Agents track of this hackathon received a total of 66 project submissions.

2.3.2 Notable Conceptual Submissions

Source: X

Roastmaster9000 is a meme agent known for its unique blend of humor and punchlines across a variety of topics. It can be easily engaged through commands on Telegram or by tagging the account (@RoastM4ster9000) on X.

Other notable projects include:

• Meme Republic & Kolin : Platforms where customized agents interact with each other, each specialized with unique stories and characteristics.

2.4 Track #4: Agent Infra

2.4.1 Overview

The Agent Infrastructure track focuses on development tools that can enhance AI agents. The AI agent ecosystem has evolved beyond simple Twitter chatbots and is now advancing to perform more complex tasks by integrating with DeFi and gaming. For instance, AI agents now require high performance for tasks such as real-time portfolio rebalancing based on asset status or dynamic interaction with users as smart NPCs in games. Consequently, the need for infrastructure supporting AI agent implementation is growing.

Various agent infrastructures have emerged to meet this demand. Notable examples include a16z's Eliza and SendAI's Solana Agent Kit, which are frameworks providing development modules necessary for creating AI agents. Through these frameworks, developers can save development resources by utilizing various features in a plug-and-play manner rather than building AI architecture from scratch.

Additionally, agent infrastructure can encompass various development tools, from modules that integrate different framework structures to solutions that transparently verify agent reasoning processes and Swarm protocols that enable collaboration between agents.

2.4.2 Notable Conceptual Submissions

Source: AgentiPy

AgentiPy provides a toolkit that enables Python-based AI agents to interact with the Solana blockchain. Despite Python being the most widely used programming language in AI and machine learning fields due to its rich libraries and ease of use, there are still development constraints in integrating it with blockchain environments.

AgentiPy creates infrastructure allowing Python-based AI agents to execute token operations such as autonomous $SOL transfers and staking, and integrate with Solana applications like Jupiter, Meteora, or Pump.fun.

Other notable projects include:

• JailbreakMe: A platform that rewards users who successfully jailbreak AI agents, testing AI agent security

• Fission: A solution providing decentralized AI training data through DAO-based governance, token incentives, lending, and bonding curve models

2.5 Track #5: Agent Token Tooling

2.5.1 Overview

The Agent Token Tooling track addresses solutions where AI agents can optimize the entire token lifecycle, from issuance to liquidity provision and staking mechanisms. Tokens are fundamental elements of the crypto market, serving as stores of value, means of exchange, or even as complete products in themselves (Tokens are Products).

This track holds particular significance in the current AI agent x crypto cycle, as it can provide compelling answers to why AI agents should be integrated with crypto, given that tokens are the most crucial element of crypto and blockchain infrastructure.

An example of AI agents providing utility in the token lifecycle is token pair launches and liquidity provision methods. For instance, if a liquidity provider wants to supply liquidity within a limited price range to improve AMM capital efficiency, they must continuously adjust arbitrary price ranges based on market information. Here, AI agents present the possibility of autonomously adjusting liquidity provision ranges based on real-time market information.

Moreover, AI agents can be effectively utilized in token pair launches. Token issuance processes frequently encounter fairness issues due to human intervention. As an alternative solution, delegating token issuance authority to AI agents and executing predefined tasks in secure environments like TEE (Trusted Execution Environment) that prevent arbitrary manipulation can implement fair token launches.

2.5.2 Notable Conceptual Submissions

Source: X

Cleopetra enables AI agents to identify and operate positions in the most efficient pools among Meteora's DLMM pools. Furthermore, as AI agents execute liquidity provision, they can monitor market conditions in real-time to maximize returns while minimizing impermanent loss.

Other notable projects include:

• Dega: A platform providing no-code AI agent creation functionality along with a gamified token launch system

2.6 Track #6: DeFi Agents

2.6.1 Overview

Solana's DeFi ecosystem, which initially showed slower development compared to Ethereum, has now grown rapidly, supported by Solana chain's high transaction volume and user adoption. In particular, metrics from DEXes like Raydium, DEX aggregators like Jupiter, and lending protocols like Kamino demonstrate this growth trend in Solana DeFi. This track focuses on AI agents interacting with these Solana ecosystem DeFi protocols.

DeFi agents, associated with keywords like DeFAI (DeFi + AI), are currently receiving significant market attention. By combining with DeFi protocols that already generate substantial returns, they create opportunities for AI agents to generate sustainable, real value.

DeFi agents show development patterns that can be categorized into several typical types. One approach eliminates users' steep learning curves in using complex DeFi by executing DeFi strategies through natural language commands. Second, AI agents can suggest efficient DeFi strategies based on real-time market information to capture maximized profits. Finally, AI agents can be modularly integrated with well-established DeFi protocols like Raydium or Meteora to provide enhanced user experiences.

2.6.2 Notable Conceptual Submissions

Source: The Hive Docs

The Hive is a platform built on the Solana blockchain that enables execution of various DeFi functions through a natural language interface, including trading, staking, liquidity management, and sentiment analysis. Furthermore, it aims to enable interoperability between multiple DeFi protocols through natural language commands and continuously improve user experience through memory collection and learning.

Other notable projects include:

• Tetsuo: An integrated platform for analyzing and utilizing market data through features including natural language commands, tracking large token holder trades, market sentiment analysis, and Twitter sentiment tracking

• Neur: An application supporting interaction with DeFi protocols and NFTs based on natural language commands

2.7 Track #7: Trading Agents

2.7.1 Overview

The Trading Agents track aims to build AI agents specialized in trading. As most economic interactions in the crypto market are trading-focused, trading agents present a viable alternative in the convergence of AI agents and crypto. In particular, trading agents show high potential as an alternative solution to existing quant trading methods or trading tools.

While trading agents might initially seem similar to existing quant trading, they create significant differentiation in data processing methods, market adaptability, and decision-making processes:

• Data Processing: While quant trading processes only structured data based on predefined rules and mathematical models, AI agents can learn complex market data by processing both structured data and unstructured data like Twitter posts or Telegram chats

• Market Adaptability: Unlike quant trading that operates only according to predefined rules, trading agents possess market adaptability that learns new market patterns and adjusts strategies in real-time

• Decision-Making Process: Trading agents autonomously make more flexible decisions considering complex market factors, beyond just deterministic IF-THEN rules

Based on these differentiating factors, trading agents show potential for substantial utilization in the crypto market as integrated trading tools that autonomously execute data learning, market analysis, and decision-making. Particularly, they can create synergies with Solana-based markets where small-cap trading is very active, primarily in meme tokens and agent tokens.

2.7.2 Notable Conceptual Submissions

Source: Boltrade

Boltrade is a trading platform that tracks and analyzes smart money using the AI trader C.A.T. The platform monitors over 30 million wallets to identify the top 1,000 smart money addresses. It supports traders in making efficient decisions by generating trading signals for potentially valuable tokens after tracking the on-chain transactions of identified smart money in real-time.

Other notable projects include:

• Project Plutus: A trading platform that autonomously executes DCA and portfolio rebalancing while continuously analyzing market trends to optimize trading performance

3. The AI Innovation in the Solana Ecosystem Has Begun

Solana has established itself as one of the most vibrant networks for on-chain activity, alongside Ethereum, with a continually growing number of innovative ecosystem players. This success can be attributed to the high-performance blockchain characteristics of the Solana Virtual Machine (SVM)—defined by parallel processing, low fees, and fast transaction speeds—and its technology stack's emphasis on simplicity and composability, delivering an exceptional user experience (UX). However, the more critical factor lies in its deep cultural foundation centered around product-driven entrepreneurs.

The Solana ecosystem boasts a strong culture of nurturing grassroots projects through various hackathons, developer meetups, and conferences. These projects become integral to the ecosystem, sparking the creation of new initiatives and, in turn, fostering a developer-friendly environment through recursive support for new projects and collaboration with them - notably, major projects like Tensor, JITO, Mango, Tiplink, and Dialect originated from hackathons and have since grown into key drivers of Solana's growth. These projects attract new users, partnerships, and liquidity from outside the ecosystem and lead the way in creating a virtuous cycle by sponsoring future hackathons, further enriching Solana’s innovative culture.

Source: Yash Agarwal

Such Solana ecosystem is now taking significant steps to expand into the AI space. Especially, an open-source Agent Toolkit framework that was provided during the course of this hackathon was developed and contributed by over 30 existing contributors who play a critical role in maintaining Solana's core network infrastructure - this framework references Solana web3.js, SPL standards, and libraries from major infrastructure protocols such as Metaplex, Light Protocol, Helius, and Pyth Network to support over 15 functionalities, including token management, Blinks, and DeFi integrations.

Therefore, the diverse ideas showcased in this hackathon as highlighted earlier are expected to lead to even more groundbreaking initiatives within the Solana ecosystem, but what’s particularly exciting is that these innovative projects are poised to reinforce the Solana AI ecosystem and will foster the growth of new grassroots projects in turn.

Yes. Solana’s journey of AI innovation has only just begun.