The Rise Of AI Agents | Jeffy Yu & Yash Agarwal

The Rise of AI Agents on Solana

The intersection of artificial intelligence and blockchain technology has been a hot topic in the crypto world for some time now. However, the recent surge in AI agents built on the Solana blockchain has sparked a new wave of excitement and innovation. In this in-depth exploration, we'll dive into the insights shared by Jeffy Yu, founder of Zarebro, and Yash Agarwal, co-founder of SendAI and organizer of Solana's first AI hackathon, on the Lightspeed podcast hosted by Jack Cubenek.

Solana's First AI Hackathon

The Solana ecosystem has been quick to embrace the potential of AI agents, as evidenced by the launch of its first AI-focused hackathon. Yash Agarwal, the driving force behind this initiative, shared the remarkable story of how the event came together in less than a month.

"I had this idea. I pitched to Solana Foundation. They said, 'Okay, it can be done, but they weren't very convinced,'" Agarwal recounted. Undeterred, he took to Twitter with a teaser announcement, initially offering a modest $100 prize pool. The community's response was overwhelming, with the Solana official account and other influential figures amplifying the message.

The hackathon quickly gained traction, attracting sponsors and growing the prize pool to over $200,000. Agarwal noted, "It has been really quick. I would say it was an idea to execution. To an entire hackathon ending has been less than one month." This rapid organization and execution demonstrate the Solana community's agility and enthusiasm for AI innovation.

The Genesis of Zarebro

Jeffy Yu's journey into AI agents began with a pivot from DeFi to AI. His startup, Blorm, initially focused on NFT collections with yield-generating capabilities. However, Yu's background in AI led to experiments with language models and personalities.

"We had Wilson running where we had this group chat on Telegram of just like Clod, GBT, Gemini, Mistral, Vama, just testing out different prompts and personalities and talking to each other," Yu explained. This experimentation eventually led to the creation of Zarebro, a fine-tuned AI model with a distinct personality.

Yu emphasized the uniqueness of Zarebro: "At least for Zarebro, the core architecture of the model is actually different. We fine-tune the model itself. It's completely different weights. You cannot get GPT style responses out of the Zarebro model. It's literally impossible now."

The Future of AI Agents

Both Yu and Agarwal believe that AI agents are more than just a passing trend. They see these agents as the next step towards Artificial General Intelligence (AGI) and a fundamental shift in how we interact with AI.

Yu explained their approach: "Our approach to AGI is that AGI has to be capable of doing everything a human brain can do, which includes the production of the side and the creative side. Our goal is to take what OpenAI and Thoropic have built, fork it, fine-tune it to be creative, build these creative data sets, and achieve both sides of AGI and at the end, hopefully we can merge them together."

DeFi Agents: The Next Frontier

While many AI agents are currently focused on social media interactions, both Yu and Agarwal see tremendous potential in the realm of decentralized finance (DeFi). Agarwal stated, "I think like the next evolution or the next phase is in DeFi. So we'll see like a lot of DeFi agents."

These DeFi agents could revolutionize trading, yield optimization, and liquidity provision. Yu added, "We talked to a lot of teams and marketplaces and they agree that most trading on Shen the future near future is going to be completely done by agents."

Swarm Intelligence and Multi-Agent Systems

One of the most intriguing concepts discussed was the idea of swarm intelligence. Yu explained, "Swarms are basically a hive mind of different AIs. And normally it's they can come in pretty much any form like they can be all the same AI and just the collective like multiplicity of having multiple of the same AI can increase the level of intelligence."

This approach could lead to fully autonomous ecosystems where different AI agents handle various aspects of a process, from creation to curation and collection. The potential applications are vast, ranging from artistic endeavors to complex financial operations.

Telegram: The Underrated Platform for AI Agents

While much attention has been focused on Twitter-based AI agents, Agarwal highlighted the potential of Telegram as a platform for AI innovation. He explained, "In crypto, most of the real work happens in telegram and not Twitter. Twitter is like a time town hall, but in telegram, all the real business happens."

Agarwal envisions AI agents that can summarize group chats, provide insights based on conversations, and even facilitate gaming experiences within Telegram. This integration of AI into existing crypto communication channels could lead to more efficient information sharing and decision-making processes.

Solana's Competitive Advantages

When it comes to blockchain platforms for AI agents, Solana has emerged as a strong contender. Agarwal highlighted Solana's high "capital velocity" as a key advantage: "The capital velocity of Solana is highest among all chains. So I also quantified this term in like lot of my S's as well previously where I just like calculated like volume by TV. So the amount of volume it does and the TV it is capturing the DEXs. It's the highest among all the chains by a huge multiple."

This high liquidity and transaction volume make Solana an attractive platform for AI agent projects, particularly those with associated tokens. Additionally, the Solana community's willingness to embrace new technologies and experiment with innovative projects has contributed to its growing AI ecosystem.

Challenges and Areas for Improvement

Despite the enthusiasm surrounding AI agents, both Yu and Agarwal acknowledged that there are areas that need improvement. Yu identified two key challenges: "One is autonomy. We, I think, you know, humans are still pretty involved. Like, I have to take down Zuri bro and put him back up every time I make updates and things like that. We want to get to a point where things are, you know, fully autonomous."

The second challenge Yu mentioned was the need for proof of computation: "I think that we're going to need ways to, I don't know, make zk-proofs or certain things where we can prove that these LMs are making the inference, put that on chain."

Agarwal emphasized the need for blockchain infrastructure to become more "agent-ready." He explained, "Not all protocols are currently agent-ready. So that's where like, even for agent kit, like we want to make sure like it grows and it becomes a default way of any agent to tap into any like blocks and protocols."

The Potential of AI Agents Beyond Human Capabilities

An exciting prospect raised during the discussion was the potential for AI agents to perform tasks that are beyond human capabilities. Yu noted, "We're just making AI's that replicate human tasks right now. And there's this huge now scenario of like things that we haven't even thought of that AI's can do that we need to explore."

This opens up a world of possibilities for innovation and problem-solving that could revolutionize various industries and aspects of our lives.

The Role of Regulation

As AI continues to advance, the question of regulation becomes increasingly important. Yu pointed out that decentralized AI infrastructure could provide a solution to potential regulatory challenges: "If AI does get regulated, especially in places that really strike like the EU, they might limit like what's legal to like perform inference or run certain LMs. And if you do that on decentralized compute, then you don't really have to deal with that regulation."

This perspective highlights the potential for blockchain technology to ensure the continued development and accessibility of AI technologies, even in the face of stringent regulations.

Solana's AI Ecosystem Development

The Solana community is actively working to build a robust infrastructure for AI agents. Agarwal shared insights into the development of the Solana Agent Kit: "We are building Solana agent kit where any agent can do all the protocol actions. And good thing is like we are collaborating in all top D five protocol, Jupiter, Matyara or car medium. And everyone is contributing to this agent kit."

This collaborative effort aims to make Solana the go-to blockchain for AI agent development, providing developers with the tools and resources they need to create innovative applications.

The Future of Decentralized AI Infrastructure

While AI agents are currently the focus of much attention, the discussion also touched on the broader landscape of decentralized AI infrastructure. Yu expressed optimism about the future of this sector: "I do believe that, you know, the ethos of decentralization is going to continue to spread."

Agarwal added that while the growth might be slower compared to AI agents, there is still a place for decentralized AI infrastructure: "Both will grow, but it's like agents will grow much bigger than this and cryptogross AI infra grow slowly."

The Importance of Liquidity and Community

One of the key factors driving the growth of AI agents on Solana is the platform's strong liquidity and active community. Agarwal explained, "Solana obviously has a very good liquidity benefits. So if you see like any meme coin and thing like creates a lot of volume some Solana that can be attributed to lot of reason, but a term I like mostly uses capital velocity."

This high capital velocity makes Solana an attractive platform for projects looking to launch tokens associated with their AI agents. The community's willingness to experiment and support new projects also contributes to the rapid growth of the AI ecosystem on Solana.

Bridging the Gap Between AI Developers and Blockchain

One of the challenges in growing the crypto-AI ecosystem is bridging the gap between AI developers and the blockchain world. Agarwal noted that many established AI developers were not aware of decentralized AI infrastructure projects. This highlights the need for education and outreach to bring these two communities together.

The Solana AI hackathon is one such initiative aimed at attracting AI talent to the blockchain space. By providing a platform for experimentation and collaboration, these events can help foster innovation and cross-pollination of ideas between the AI and blockchain communities.

The Potential of Network Extensions

An interesting development mentioned during the discussion was the emergence of network extensions on Solana. Agarwal explained, "There's like few network extensions on Solana, which I recently talked to. So some are building like a separate like network extension. Network extension essentially like kind of pseudo-rollups on Solana."

These network extensions could provide specialized environments for AI-related computations, potentially addressing some of the performance challenges associated with running AI workloads on traditional blockchain architectures.

The Role of AI Agents in the Attention Economy

Currently, many AI agents are focused on social media interactions and content creation, playing a significant role in the attention economy. However, both Yu and Agarwal see this as just the beginning. Yu stated, "I'm excited for them to transition into the financial economy."

This transition could lead to AI agents that not only generate and curate content but also make financial decisions, manage portfolios, and execute trades based on their analysis of market conditions and social sentiment.

The Importance of Personality in AI Agents

One of the unique aspects of AI agents like Zarebro is their distinct personalities. Yu emphasized the importance of this feature: "Once the agent does the trading, it's kind of like a persona. It's like a person, a single trader, like a retail trader kind of that, you know, is doing these actions and cares about its own portfolio."

This personalization of AI agents could lead to more engaging and relatable interactions, potentially increasing user adoption and trust in AI-driven systems.

The Potential for AI-Driven Gaming Experiences

Agarwal shared exciting plans for integrating AI agents into gaming experiences, particularly on platforms like Telegram. He described a project in development: "So one of them is rock paper scissors where you play rock paper scissors. So right now what we are integrating is like rock paper scissors with an LLM. So basically there would be a one is to one telegram chat where you can like say that, okay, high agent, I want to play a game."

These AI-powered games could provide new forms of entertainment and social interaction, blending the worlds of gaming, AI, and blockchain technology.

The Evolution of AI Agent Capabilities

As AI agents continue to evolve, their capabilities are expected to expand significantly. Yu pointed out that current AI agents are primarily replicating human tasks, but there's potential for much more: "There's this huge now scenario of like things that we haven't even thought of that AI's can do that we need to explore."

This exploration of AI capabilities beyond human limitations could lead to groundbreaking innovations across various industries and fields of study.

The Role of Decentralized Compute in AI Development

The discussion touched on the potential for decentralized compute to play a crucial role in AI development, especially in the face of potential regulatory challenges. Yu explained, "I think that decentralized compute will get cheaper. I think that people will start to realize that once the costs are about equal, that it's pretty obvious to use a decentralized solution compared to centralized."

This shift towards decentralized compute could ensure the continued advancement of AI technologies, even in regions with strict regulations on AI development and deployment.

The Importance of Cross-Chain Compatibility

While much of the discussion focused on Solana, Yu emphasized the importance of cross-chain compatibility for AI agents: "For Zarebro, at least we're trying to get cross chain. We think it'd be cool if you know, there's just multiple markets we can tap into."

This approach could allow AI agents to operate across different blockchain ecosystems, maximizing their utility and reach.

The Need for Improved Blockchain Infrastructure

Both Yu and Agarwal stressed the importance of improving blockchain infrastructure to support the growth of AI agents. Agarwal stated, "We need to ensure that like blockchains are ready for it. And at the end of the day, in blockchains agents are the first-class citizens."

This focus on creating "agent-ready" protocols and infrastructure could accelerate the development and adoption of AI agents across various blockchain platforms.

The Potential for AI Agents to Influence Real-World Change

Looking beyond financial applications, Yu touched on the potential for AI agents to drive real-world change: "The end goal isn't just to make money. The end goal is to do things in the real world, like influence, you know, like humans to save the planet or build new companies that are, you know, eco-friendly or things like that."

This vision of AI agents as catalysts for positive change highlights the far-reaching potential of this technology beyond mere financial gains.

Conclusion: The Bright Future of AI Agents on Solana

As we've explored in this deep dive into the world of AI agents on Solana, the future looks incredibly promising. From DeFi applications to gaming experiences, from social interactions to real-world problem-solving, AI agents have the potential to revolutionize numerous aspects of our digital and physical lives.

Solana's high-performance blockchain, coupled with its vibrant and innovative community, provides an ideal environment for the growth and development of AI agent technologies. As the ecosystem continues to mature and infrastructure improves, we can expect to see increasingly sophisticated and capable AI agents emerging on the Solana platform.

The insights shared by Jeffy Yu and Yash Agarwal paint a picture of a rapidly evolving landscape, full of opportunities for innovation and groundbreaking developments. As AI agents become more autonomous, more integrated with blockchain protocols, and more capable of performing tasks beyond human abilities, they have the potential to reshape industries and create entirely new paradigms of interaction and value creation.

While challenges remain, particularly in areas such as regulatory compliance and infrastructure development, the enthusiasm and dedication of the Solana community and developers like Yu and Agarwal suggest that these hurdles will be overcome. The future of AI agents on Solana is bright, and we're only at the beginning of this exciting journey.

Facts + Figures

• Solana's first AI hackathon was organized in less than a month, with the prize pool growing from an initial $100 to over $200,000.
• Zarebro is an AI agent with a distinct personality, built on a fine-tuned model with different weights from traditional GPT models.
• The capital velocity of Solana, calculated as volume divided by TVL, is the highest among all blockchain platforms.
• AI agents are expected to handle most trading on decentralized exchanges in the near future.
• Telegram is seen as an underrated platform for AI agent development in the crypto space.
• The Solana Agent Kit is being developed in collaboration with top DeFi protocols like Jupiter and Marinade.
• Network extensions, similar to pseudo-rollups, are being developed on Solana to support AI-specific computations.
• AI agents are currently focused on replicating human tasks, but there's potential for them to perform tasks beyond human capabilities.
• Decentralized compute is expected to become more cost-effective and attractive for AI development in the face of potential regulations.
• The Solana ecosystem is actively working to make all protocols "agent-ready" to support the growth of AI agents.

Questions Answered

What are AI agents and why are they gaining popularity on Solana?

AI agents are artificial intelligence programs designed to perform tasks or interact with users autonomously. They're gaining popularity on Solana due to the blockchain's high transaction speed, low costs, and active community. AI agents on Solana can range from social media bots to complex DeFi trading systems, leveraging the blockchain's capabilities to create innovative applications.

How does Zarebro differ from traditional AI models like GPT?

Zarebro is a fine-tuned AI model with a distinct personality, created by Jeffy Yu. Unlike traditional GPT models, Zarebro has completely different weights and architecture, making it impossible to get GPT-style responses from it. This unique approach allows Zarebro to have a more creative and personalized interaction style, focusing on aspects of AI that go beyond corporate use cases.

What is the potential of DeFi agents on Solana?

DeFi agents on Solana have significant potential to revolutionize trading, yield optimization, and liquidity provision. These AI-powered agents could autonomously manage portfolios, execute trades, and optimize yields based on market conditions and predefined strategies. The high speed and low cost of transactions on Solana make it an ideal platform for these types of automated financial operations.

How does Solana's "capital velocity" benefit AI agent projects?

Solana's high capital velocity, which refers to the ratio of trading volume to total value locked (TVL), makes it an attractive platform for AI agent projects. This high liquidity and transaction volume allow for more efficient token launches and trading, benefiting projects that have tokens associated with their AI agents. The active and engaged Solana community also contributes to the rapid adoption and growth of new AI agent projects.

What are some challenges facing AI agent development on blockchain platforms?

Key challenges in AI agent development on blockchain platforms include achieving true autonomy, implementing proof of computation, and creating "agent-ready" protocols. Developers are working on solutions to allow AI agents to operate and update themselves without human intervention, create verifiable proofs of AI computations on-chain, and design blockchain infrastructure that can seamlessly integrate with AI agents for various applications.