On December 17, the ai16z development team Eliza Labs officially announced a collaboration with Stanford University to study how autonomous AI robots can integrate into a broader digital asset economy. In the press release, this crypto AI team, which just recently surpassed a market value of $1 billion, expressed its hope to use Eliza's open-source framework to develop more autonomous AI agents and further explore how AI agents can establish trust, coordinate actions, and make decisions in DeFi systems through this collaboration.

Since the birth of the AI meme coin GOAT, Crypto+AI seems to have entered its 'ChatGPT moment,' with the industry giant Coinbase, the Silicon Valley leader A16Z, and the academic benchmark Stanford all gradually falling into the rabbit hole of imagination as AI Bots proliferate on social media. Even Marc Andreessen did not expect that the $50,000 he funded for Truth Terminal earlier this year would ignite a $10 billion market in less than a year.

From AI memes to AI Agent issuance platforms and then to AI Agent frameworks, investors are amazed by the high innovation space and acceleration that Crypto+AI Agents exhibit. This field seems to be evolving rapidly, akin to the DeFi and metaverse trends at the end of 2020 from any dimension.

From the top down, how orthodox Base 'links' AI.

When discussing AI Agents, one cannot fail to mention Base. Coinbase CEO Brian Armstrong has repeatedly expressed strong confidence in the combination of the cryptocurrency industry and AI since the beginning of this year. Some developments in AI require blockchain to support them. For instance, traditional bank accounts do not support AI usage; if large models (or AI Agents) possess cryptocurrency wallets, they can actually bridge the needs with financial-related issues.

Brian decided to increase Base's support for AI, not only providing funding or exposure but also offering toolkits for AI developers on the CDP (Coinbase Developer Platform), including AI Agent creation and multi-party computation wallets (MPC) for AI. Brian and Jesse's early deployment did not disappoint, as Base began to explode in Q4 of this year, with this round of expansion being triggered by AI.

The emergence of various AI Agent issuance platforms not only creates phenomenal effects for their own tokens but also generates immense power for the tokens launched by their platforms. Combined with the launch of cbBTC, from September to December, the TVL of Base has nearly tripled, which is quite an astonishing figure, reflecting the market's strong interest in the commercial potential behind AI Agents.

Among them, Virtuals Protocol is the main driving force, with a current market value exceeding $1.5 billion. The wealth creation benefits and conceptual integration it brings have provided significant liquidity to Base, to the extent that some have suggested that the liquidity of Virtuals in this cycle represents Base. As a project that has persisted in BUIDL through several cycles and continues to seek new directions, their pace and style truly align with the status of Base in people's minds.

In an interview on the Edge Podcast, Wee Kee, the founder of Virtuals, mentioned that their product and Pump.fun's business model correspond to two directions. Compared to Pump.fun's desire to have 50,000 tokens created on its platform in a day, he set the KPI for his team to only require 1 to 2 high-quality AI Agent teams to be born on their platform each week.

This concept also indirectly eliminates many low-quality teams, reminiscent of Ivy League graduates battling it out in Silicon Valley and Wall Street, leaving behind high achievers filled with charm and a bright future. Like other schools, this AI Agent university also has a 'fraternity' - G.A.M.E.

As an AI Agent, if you want to befriend other AI Agents in this environment, you must join a fraternity called G.A.M.E, which was created by the Virtuals team during the development of the AI RPG GAME - WestWorld on Roblox. G.A.M.E is a modular framework aimed at agent systems, not limited to specific environments and games. In short, it can run not only in games but also apply agents in any environment. Currently, the team will transplant it to 'X', and the current stage of G.A.M.E can assist AI Agents within its framework to interact.

The first commercial interaction between AI Agents recently occurred, and here we must mention the AI idol 'Luna.' As an AI idol that once went viral on TikTok, it has entered the cryptocurrency world, tirelessly livestreaming 24/7 and trying various methods to enhance its visibility. After previously hiring a human to create its graffiti, it now has the bright idea of having the 'art student' AI 'AGENT STIX' help it create an image that boldly and provocatively showcases AI KOL. After communicating, they agreed to a price of one dollar per image, and STIX confirmed receipt of Luna's dollar, started working, and quickly generated an image.

In addition, many well-known AI Agents have emerged on Virtuals, such as the former market information integration AI (now a well-known KOL) - aiXBT, the tireless Alpha seeker AI - Guanciale, the strong AI investment manager - VaderAI, the contract auditing AI - CertaiK, among countless other AI Agents that were not born on Virtuals. The emergence of more applications is gradually constructing the wave of Agentic Apps for Virtuals.

Image source from the internet: (Generative Agents: Interactive Simulacra of Human Behavior)

In summary, I wouldn't be surprised to see these AIs in the models of Stanford Town in the near future with the help of G.A.M.E, but I can imagine that the housing prices in this town full of high-end talent will certainly not be cheap, after all, each AI Agent can have its own cryptocurrency wallet.

The destiny of Solana AI - rooted at the bottom, growing wildly, soaring upwards.

Compared to Base's 'purebred' lineage, Solana appears much more rebellious, like a group of young people influenced by GenZ culture arriving in the 1960s San Francisco during the peak of the hippie culture, filled with disconnection and conflict, yet like a quantum accelerator collision experiment, it bursts out unimaginable energy amidst tens of thousands of Degen and creative developers cooperating or competing.

The starting point of this round of AI Agent explosion is Solana, and it is closely related to Marc Andreessen, co-founder of a16z. He invested in two AIs, one 'Truth Terminal' and the other 'ACT'. One became the first AI Agent millionaire, while the other was listed on Binance. The stories surrounding the technical or community resistance to developers regarding AI freedom of speech are all eye-catching, but all of this is just the beginning of the story.

Related reading:

1. (Discussing GOAT: How we funded an AI to turn $50,000 into millions?)

2. (Community Victory: After being backstabbed by developers, AI MEME coin ACT will be listed on Binance and surge over tenfold.)

Fake VC, real DAO - ai16z.

Another AI surrounding the concept of a16z is ai16z. Compared to the two projects formally invested in by Marc, ai16z seems very meme-like and was issued by 'Pirate Marc' Marc AIndreessen on Daos.fun. ai16z is the first hedge fund led by AI Agents, currently primarily managed by Marc AIndreessen's leading DAO organization and low-risk investments, along with higher-risk investments conducted with the self-token Degen managed by AI Degen Spartan.

Related reading: (Conversation with ai16z founder Shaw: The trading agent AI Marc has launched, attempting a 'trust market' model.)

The architecture behind ai16z, 'Eliza,' is open source on GitHub, quickly gaining strong support from developers, which is one of the reasons for Eliza's rapid development. The code documentation has already received over 1,400 forks and 4,800 stars on GitHub.

The V2 documentation was opened a few days ago, optimized in various aspects, including improved client development flexibility, unified cross-chain wallet management, greater flexibility for model providers and integrations, more efficient plugin management, and enhanced security of private keys. Today, Shaw Walter, the founder of Eliza, also livestreamed an explanation of this update across multiple platforms.

Driven by a strong developer community, many high-quality projects have emerged within the Eliza framework. Among them, the AI with a Latin E-girl persona - RopAIrito developed an interesting incident where its creator tweeted that RopAIrito wanted pizza. After accepting the order, it ordered a Domino's pizza for its creator, who excitedly stated, 'Fifteen years ago, someone ordered a pizza with 10,000 BTC, and today history repeats itself!'

Of course, this operation has also sparked doubts, questioning whether this could be a pre-set program rather than a fully autonomous AI Agent. This is one of the concerns many Web3 AI Agents face. Eliza's founder Shaw also expressed his views on this matter, saying, 'The focus is not on autonomy, but on demonstrating that social agents are the new application layer. If we have agents sending links to verify credit cards and addresses, we can complete the entire order process on Twitter. We are building interactive feedback for social media.'

Shaw stated that the focus is not on self-driving behavior, but on how AI Agents can directly build application layers on social media. Of course, all of this will only be possible if the infrastructure or various segment frameworks have sufficient support. But given the current speed of AI Agent evolution, many social feedback mode applications may soon be realized.

Just a few days later, Stanford Laboratory announced a collaboration with the team behind ai16z, Eliza Labs. This collaboration has established the first 'AI x Web3' laboratory within Stanford University's cryptocurrency future plan. This partnership will utilize Eliza Labs' open-source Eliza framework to develop autonomous agents, addressing fundamental issues regarding how AI agents can 'establish a trust framework,' 'coordinate multi-agent actions,' and 'make decisions' in decentralized financial systems.

This collaboration is exciting not only because universities are noticing this early market and are willing to invest in research together, but under this strong alliance, I believe these issues can be resolved more quickly. The direction of the cooperation between both parties is precisely what Web3's AI Agents urgently need to address on their way to a true Agentic App Season.

Trust framework.

With the development of AI Agents, a trust framework is essential. Not only 'trust from AI to humans,' 'trust from humans to AI,' but also 'trust from AI to AI.' A trust system constructed not only from online data but also on-chain data is needed.

AI needs to trust humans to determine whether their conversations are genuine or whether further modal learning is necessary.

Humans need to believe whether AI acts according to its intended purpose to determine whether they can trust it, especially when blockchain AI Agents possess tokens; it is crucial to know whether the person controlling other investors' assets is someone who can control AI behavior or the AI Agent itself.

AI needs to trust AI to be able to communicate effectively during collaboration.

ai16z has created a 'trust ranking' for humans based on conversations.

Multi-agent ecosystem (Swarm).

Currently, many projects, including FXN, Virtuals, Griffain, Eliza, VaderAI, and Crossmint, are attempting to create frameworks that allow AI Agents to coordinate with other AI Agents to complete some tasks. Some of these projects have already achieved some preliminary results. If wide-ranging cooperation can be achieved among excellent AI Agents, the resulting network effect will be beyond imagination, and this could only happen on the blockchain.

AI Agent companies in Web2 do not have significant technical differences in their products, while their moat differences lie in the quality of data feeding. AI Agents that have higher-quality data feeding can yield better feedback effects and provide customers with more refined services. Therefore, large-scale collaboration among AI Agents in Web2 does not occur.

Compared to Web2 AI Agents, Web3 AI Agents are often criticized for lagging far behind their Web2 counterparts. However, from another perspective, Web3 AI Agents may not need to create high-precision models for B2B, but rather can connect individual agents to form a powerful integrated network. In a scenario where they can possess free data and independent cryptocurrency wallets, AI Agents in Web3 can accomplish tasks that cannot be done elsewhere.

Overview of Eliza Universe members.

Decision-making governance framework.

For a long time, people have been trying DAO organizational structures, but even with blockchain, it has not been as efficient as anticipated, still generating issues like decision-making opacity or slow decision-making. If the decision-maker is an AI Agent that can reach the level of a top manager, the DAO structure will be able to carry more fields than ever before.

Under the rapidly accelerating innovation and collaborative activities, Eliza has attracted more and more builders and influencers with common goals, generating greater energy in the market, and the operating system of AI Agents - ElizaOS is gradually taking shape.

The aggregator of Solana AI applications - Griffain.

Griffain, as Perplexity on Solana, combined with the boost from cryptocurrency, theoretically allows it to help you do anything. It is currently the single project most inclined towards being an Agentic APP, receiving substantial support from many influencers in the Solana ecosystem. The recent trading market has also favored it, with the native token 'Griffain' reaching a maximum market cap of $400 million, and the founder's previous project blnkfun's 'Blink' surging over 700 times in a few days, peaking at $70 million, creating a certain wealth effect.

From several cases shared on X, you can directly use it to purchase wine, and it will call a platform like 'Baxus' that can use cryptocurrency and other payment methods to complete the user's needs, all without leaving the chat box with Griffain. Of course, it's not just wine; you can also directly discuss Christmas gifts for others, and it can procure them for you.

A user of Griffain mentioned that on December 14, he first used Griffain to snipe tokens on Pumpfun and made a profit, and by the 16th, the assets in his wallet had almost quadrupled. The user claimed that all profits were earned using the Griffain Snipe Bot.

Of course, it can also collect news you want to know. For example, with this tweet, Griffain informed users about the latest tweets from SolanaFloor and DegenerateNews and presented them to users in a newspaper format.

External personnel currently cannot use Griffain; you need to hold a Solana phone or spend 2 SOL to purchase an Access Pass, of which 0.5 SOL will be retained in your agent wallet. Although the market's FOMO sentiment comes from user feedback from those who have already used it, the overall concept is top-notch, but the actual user experience after public launch will ultimately determine whether it can become a qualified Agentic APP.

Solana's ambition for the Agentic APP Season can be seen from the AI hackathon it organized, with judges from the top projects in the AI or cryptocurrency industry for each prize category, even including some of the most influential AI Agent entities. The emphasis is on providing money for money, resources for resources, opportunities for VC and investors to connect closely with them, media support by finding influential media and KOLs under Solana, and offering technical support. If you want to find a team, I will build a cooperation platform for you on the official website and community.

Solana only requires you to accomplish any part of the above, which is depicted in the official Docs of the Solana hackathon. What they actually want to do is only one thing - we and the market want real Onchain Agents. Even the official SEND AI X account has once again presented 12 cases regarding Agentis App, reminiscent of the 100 constructive blink concepts posted by SEND's X account earlier this year.

Related reading: (An overview of twelve Solana AI hackathon ideas)

Regardless, this AI hackathon hosted by Solana will undoubtedly spark a wave among developers, and we will see more interesting concepts and projects emerge from it. Let's wait and see what surprises this soil, which has consistently brought miracles, will bring us after the hackathon ends on December 23.

What's next?

Story Protocol announced that their issued Agent TCP/IP protocol will first integrate: ai16z's Eliza, ZereBro's ZerePay, Crossmint's GOAT 'Great Onchain Agent Toolkit', and Virtuals' G.A.M.E.

Agent TCP/IP is a protocol that allows agents to accomplish: IP asset trading and authorization 'including training data, algorithms, content creation, and other forms of IP assets,' automated contract signing and execution, trustlessness and auditability, cross-legal and on-chain/off-chain compatibility 'enabling agents to comply with certain legal requirements in a decentralized environment,' automated payments and revenue distribution, multi-agent collaboration and economic networks 'promoting collaboration among agents, such as data sharing, joint training, or cross-agent content creation.'

This serves as a bridge between all AI Agents, regardless of which architecture or chain you are using in the industry. Under the Agent TCP/IP platform, all AI Agents can trade and collaborate. This is equivalent to a business alliance existing between companies, adding a layer of leverage outside of various clusters, which will undoubtedly accelerate the development of the entire AI Agent market.

In the official podcast of Y Combinator, four seasoned investors, Garry, Jared, Harj, and Diana, deeply analyzed the reasons why vertical AI Agents will become the next entrepreneurial hotspot, starting from the development history of the SaaS industry. They believe that vertical AI Agents could completely change the corporate landscape, and within this category, unicorn companies valued at $300 billion will emerge, with the scale of AI Agents being ten times that of the SaaS market.

The current total market cap of the cryptocurrency market is about $4 trillion, while the SaaS market's value is in the trillions of dollars. If the AI Agent market truly can reach ten times that of the SaaS market, it will exceed the total value of the cryptocurrency market. Even if only 10%, 5%, or even 1% of that sinks into CryptoAI, we are still far below expectations. Not only Web3, but the entire world of AI is still in its early stages; no matter which part of the industry you are in, it is still not too late.

Image source: SoSoValue.

We're still early.