2025 is the 'Year of the Agent'
The release of Gemini 2.0 marked a critical shot towards the new world of AI Agents.
Today, Google has officially released its most powerful AI large model to date, Gemini 2.0, distinctly positioned as 'AI model for the agentic era.' New features include native image generation and audio output, with support for native calls to Google Search, Maps, Lens, and other tools.
Google claims that 2025 will be the 'Year of the Agent', and the financial, payment, and data aspects involved with Agents are a perfect match with blockchain. This is a revolution even more exciting than DeFi.
AI agents are the focus of mainstream development in the coming years, and agents in the crypto circle are the most natural combination with AI, allowing proxies to directly decentralize using currency.
This can be understood as:
$goat=BTC pure meme
$virtual =eth [its $game= L2]
$ai 16z =sol [its $ELIZA= JUP or ray, not accurate but easy to understand.] Both have an explosive number of developers and are very active on GitHub.
The AI Agent track will see projects worth 10 billion to 50 billion emerge. Currently, it seems:
(1) agent platforms or frameworks, virtual games, ai16z, ELIZA, etc.
(2) Leading meme types, $goat——this one is particularly interesting, Binance does not list it on spot, but if you believe that "trend is the ultimate force", then just stick to doing the right thing and wait.
In any case, please focus on the AI Agent track and welcome the arrival of this epic bull market's main rising wave.
1. AI MEME, only considering leading coins like $GOAT that have a narrative origin, lacks logic. An early meme in a track, although primarily speculative, plays an important role in educating the market and cultivating user awareness, much like DOGE did for the popularization of cryptocurrency.
2. AI Agent infrastructure is very important, and in the long term, it is underestimated. For instance, early frameworks like $ELIZA have rich GitHub open-source libraries and a very wide application coverage. Although it is indeed very early and could be outcompeted by stronger frameworks at any moment, its absolute advantages of open-source sharing and first-mover advantage serve as a barrier. It is likely that better frameworks will emerge on top of it, perhaps affecting its growth potential but definitely not eliminating it. If we treat ELIZA as the EVM of the AI Agent era, there will inevitably be new frameworks such as EVM++, EVM Compatible, etc. that will supplement it.
3. AI Agent asset issuance, custody, and trading are currently still at the narrative level of meme-like asset issuance. It is essential to lock in platform value and not be misled by the 'junk assets' on the platform. The logic is simple: AI Agents are essentially a process of value landing based on a crypto incentive mechanism within an AI scenario. The early infrastructure, applications, protocols, etc. are still immature, so how can there be so many super applications at the AIXBT level? The convenience of issuing tokens can bring what? Realization? In the end, it only creates a bubble trap of hype for asset issuance platforms like VIRTUAL, VVAIFU, etc. Therefore, do not overly pursue AI Agent super applications unless there is absolutely distinct information; otherwise, when you see the market value surge, the cost-effectiveness won't be high, while locking in innovative platform infrastructure is different.
4. There is still a large blank space in the construction of AI Agent infrastructure. For example, early frameworks like ELIZA only solved the dialogue mechanism between LLM large models and AI Agents, and were applied in scenarios like Twitter, Discord, etc., but more often they are just matching web2 information. When will there be a version of ELIZA++ that can bridge the ELIZA framework with web3 transaction scenarios based on public chain paradigms? It is worth looking forward to; also, platforms for AI Agent asset issuance like Virtual and VVAIFU can issue assets quickly, but they still lack greatly in areas like decentralized custody of assets and AMM-style intention-based trading. When will the previously hotly debated AI Agent + TEE private key management approach mature? Will asset issuance platforms extend new gameplay in trading, and will there be more interesting trading methods and incentive models emerging?
Also, can there be some chains that establish interaction standards for AI Agents, becoming the Cosmos of the AI Agent era, providing standard interoperability capabilities? Or some chains that build permanent memory modular storage for AI Agents, constructing the necessary DA capabilities? There are also AI Agent visualization tools, execution engines, modular combination packaging services, etc. With these technologies and market-driven supporting infrastructure, will you still be anxious about a pile of incomprehensible MEME?
That’s it. As for where the various layer1, high-performance chains, cross-chain infrastructure, chain abstraction, DA combination chains, layer2, Restaking, and so on, built on the smart contract paradigm, will go, apart from the original infralay application PMF delivery realization issues, one breakthrough point is the new concept of 'riding the AI Agent narrative.'
Therefore, any chain that can jump out to provide communication protocols, DA, application development, etc. for AI Agents deserves attention. Of course, there will definitely be chains that are purely concept-driven, but they are still better than those content with stagnation.
More practical strategies, specially prepared for spot enthusiasts and bull market hoarders! Click on my avatar to follow and get my bull market strategy deployment, shared for free with all fans. You need references, I need fans, to be the most generous bull market blogger on the internet!