In October 2024, since AI Meme GOAT launched the AI Agent concept, Crypto began to accelerate its integration with AI. Concepts such as Game+AI, DeFAI, and AIAgent Hive have emerged, and a batch of new concept projects appear almost every week. Until January 18 this year, Trump announced the issuance of MemeCoin, which directly drained market liquidity and the bubble of CryptoAI was punctured in advance. Two days later, DeepSeek announced the open source R1 model, and after a few weeks of fermentation, the AI concept stocks in the US stock market were also punctured.
With 6 million US dollars, a large model can be built. Retail investors who are still buying the CryptoAI with a market value of 60 million US dollars were once defined by the community as leeks with low IQ. This trend was very popular, and with the collective plunge of AI concepts in US stocks, the currency price naturally also spiraled into death. Since then, the concept of AI has gradually faded in Crypto.
A quarter later, the voice of CryptoAI has returned in the market. $zerebro rose 400% in 2 days, Agent operating system $AIOS rebounded 10 times, Dark rebounded again, and Flip Musks concept had a market value of over 40 million US dollars during the same period. Other old AI projects $ALCH, $SWARM, $GRIFFAIN, $arc, and $BUZZ also had increases ranging from 30% to 60%. In addition to the currency price, the return of AI also brought back some new concepts.
AI Ecosystem Coin Issuance Platforms Show Their Capabilities
After the success of Pumpfun, Crypto has set off a wave of asset issuance platforms. When the last wave of AI Agents came, frameworks and AI distribution networks/LaunchPad were the two areas with the highest FDV. After the framework concept of developer communities like Ai16z developed to a certain extent, the community found that the framework could only do a limited amount of things, especially the value capture in Crypto, so the surviving frameworks gradually transformed into LaunchPad. AI LaunchPad was not doing well either. The reason was that the LaunchPad at the time could not meet the needs of AI product issuance. In order to support the return of the next AI boom, project parties began to look for solutions.
The Promise of CryptoAI Infrastructure — Bittensor
Although the SN 28 subnet was exploited by someone to turn it into MemeCoin some time ago, and promoted the release of TAO into Meme Coin speculation, which was eventually intervened by the foundation, and as time goes by, the foundation will have less and less control over the Bittensor subnet, which has aroused the communitys doubts that it may become a pan-incentive project of attention network in the future. Thinking Guaiguai, who has been tracking the Bittensor project for a long time, also published an article, saying that this is a scam.
Further reading: Viewpoint: Why is Bittensor a scam and TAO heading towards zero?
But from a purely investment perspective, the liquidity of the Bittensor ecosystem is better than other AI Agent ecosystems. For example, Virtuals, because LPs are paired with Virtuals, this will lead to higher volatility for liquidity providers. There will be about 3% to 7% slippage on the proxy tokens within the investor investment platform. When investing funds in dTAO subnet tokens, the slippage is usually 0.05% to 0.1%.
For this reason, VCs or large-cap AI project participants are more inclined to make long-term investments in Bittensor. Just last week, former Messari analyst and Crucible Labs partner Sami Kassab announced that he and his friend Seth Bloomberg, who has similar work experience, will set up a fund dedicated to providing liquidity for Bittensor.
Currently, Rayon Labs, the first project party of Bittensor, has produced several products, from which we can get a glimpse of the preferences of Bittensor project parties, who tend to be more practical and long-term.
SN64 Chutes provides a serverless way to easily deploy AI infrastructure. The project team said that the previous AWS outage is a good example of why we need serverless. Because if we rely on centralized service providers, once an outage occurs, AI applications may go down due to a single point of failure, and Crypto is an industry that is closely related to money, and the probability of causing losses is much greater than traditional AI.
SN56 Gradients is a platform for zero-code deployment of AI models. Users can train their own AI models on Gradients (for specific use cases, image generation, custom LLM). The recently launched v3 is more competitive in price than similar products.
SN19 Nineteen is a fast, scalable, decentralized artificial intelligence reasoning platform.
The creator of AI Agent protocol——Virtuals Protocol
As one of the most complete AI projects combining ecosystem construction and value flywheel in the last cycle, the price of Virtuals Protocol has slowed down with the markets silence, and the bubble of $4.5 billion market value has been punctured, with a drop of more than 90%, and the number of participants in the launcher has also dropped significantly. But Virtuals did not give up, and they started to build in this AI bear market.
First of all, Virtuals has improved its own project construction ecosystem and released the VPN plan Virtuals Partners Network. From the beginning, Virtuals plan is to introduce more AI people into Crypto, and the plan connects multiple ecological positions, including investors, experts in various fields, scholars and developers. Almost as long as you have an idea, from investors to market makers, marketing and even professional talents can obtain resources from this plan. The one-stop service incubator can be said to be the best choice for any talent who wants to enter Crypto to cooperate with Virtuals.
In order to expand the influence and interaction of AI Agents in the ecosystem, Virtuals has developed a protocol ACP Agent Commerce Protocol, which can be said to be a concrete version of the honeycomb concept of previous projects such as Swarm and Ai16z. ACP builds a business ecosystem composed of AI agents. For AI Agents, this is a virtual country where they can interact, collaborate and trade with each other autonomously. It is worth mentioning that Google also released a similar A2A concept afterwards. The slightly different point is that ACP is connected by smart contracts, while A2A is connected by protocols.
In April, Virtuals just launched the new model of Virgen Points and Genesis Launch Mode. Users can obtain points by investing in Sentient and Prototype Agents, holding Virtuals, staking VADER, etc. Points are the basis for participating in the Genesis Launch Pad project. The Genesis Launch Pad is a project launch method in the IDO model, and the investment amount is obtained according to the Points owned by the user. However, this launch mode is not currently available to everyone, but needs to be reviewed by the official Virtuals.
This model has several advantages. The first is to increase the stickiness of platform users through rewards. In the words of founder Ethermage, our principle is to reward believers. Staking points to obtain the amount to participate in the Genesis project makes the launch more fair, and the participants are usually of higher quality, allowing the project to develop more sustainably.
Interesting projects for Virtuals Protocol’s hackathon?
In addition to the projects on Genesis that are worth paying attention to, Virtuals just announced the winners of this hackathon on April 21. A total of more than 100 project parties participated in the competition, and the judging lineup was also quite luxurious, including Luca Curran, who is responsible for the development of AI and DEPIN sections at Base, Kun Peng, who founded the Stanford blockchain community, and Anand Iyer , a partner of Canonical Crypto. Whats interesting is that Anands tag on X is AI.
The Intern is an AI that helps with operations. It can help promote, reply and manage the community on X. It can understand the culture of the community by in-depth community learning and can generate pictures using TADA . Now it has cooperated with Pudgypenguins to launch the Penguin Intern and operate its own Twitter. Judging from the operational quality of Twitter, if it is completely operated independently by AI and this level can be mass-produced, it will be a good product.
BuzzingClub is a prediction market platform. The project team believes that the future of the prediction market should be in the hands of participants rather than a central authority. Everyone should be able to freely create, share and express their opinions. Therefore, Buzzing is freer than other prediction platforms.
In Buzzing, all users can create prediction markets by proposing topics or questions and AI will generate rules. Then, AI algorithms will be used to filter out some spam and low-quality prediction markets. Finally, AI oracles will automatically retrieve Internet search data instead of manual data to make predictions.
Burnie is a code learning platform that can improve the various levels of users who want to learn code. At the same time, players can get rewards by completing the tasks he posted.
Inspiration is App, the layout of sinking market in the post-AI application era
After the development of their frameworks gradually stagnated, Arc and Ai16z also transformed themselves into distribution platforms for AI Agents. Arcs distribution platform forge disappeared after its first product AskJimmy was launched, and Ai16zs AutoFun was launched a few days ago, but none of the publicly supported projects are currently online, so it is unknown how it will develop. From the perspective of the product framework, AutoFun seems to be more inclined to create a UGC cultural platform for the community, and in terms of value retention, it is not much different from the traditional LaunchPad.
Arc plans to launch the new Agentic App Store Ryzome, and Myshells existing AIApp Store. The former has not yet been officially launched, and the latter lacks activity, and most of the products that appear are similar.
In this case, dev.fun, which was previously presented as AppFi, is more orthodox. At first glance, it is similar to Pumpfun in color and UI, but it seems to be more dynamic in some ways. Although dev.fun has fallen a lot in this cycle like other AI projects in terms of currency price. Surprisingly, there are currently nearly 13,000 apps born on this platform.
dev.fun provides the function of generating apps by chatting with AI (similar to Replit, which was previously invested by YC). In addition to issuing project/meme tokens, users can also choose trading pairs by themselves. Buidl, which currently has the most supporters, currently has a total of 1,400 apps and has run nearly 70,000 times.
Which AI does the Crypto market prefer now?
The more popular projects in the current market are divided into several categories. The first is development tools, including frameworks, AI Coplit programming tools, and MCP infrastructure. The second is consumer AI applications, including AI Agent, games, DeFAI Alpha Signal, Fund, Automated LP, and GambleFAI. The third is decentralized AI infrastructure, such as decentralized computing, verification, storage, etc. The first two are usually more popular with retail investors, while the third is the first choice for VCs or investors. Such projects usually require a fairly high valuation to undertake.
Development Tools
This type of project has a wider range of usage scenarios and project concepts, but it is often the first choice of scam project parties because for retail investors, there is no product that they can see. The feedback cycle of this type of product is often longer. Whether it can be used or whether it can produce viral spread and someone uses it is the key. Therefore, the production of this type of product requires a teams technical and marketing skills.
The MCP project Dark, which was very popular a few days ago, is one of the projects that has done both just right. It launched DARK under the influence of MTN DAO and delivered the Dark Forest game product that can see within two weeks.
In this field, the most popular one in the market is Solanas son SEND AI, which became famous for hosting the large-scale AI hackathon held by Solana, and Solana Agent Kit is used in many products. ALCHEMIST AI also continues to build, and its market value has returned to $14 million from a drop to $1.4 million in February.
Autonome is a platform that provides developers and users with the ability to build, deploy, and distribute verifiable AI agents without code. It is produced by the Rollups protocol AltLayer. It has been active since the last cycle and has not yet issued any tokens.
Consumer AI
game
Some time ago, Treasure just announced that it would shift the focus of the project from game chains and game operations to AI+NFT. SMOL gave its first answer today. Smol announced the release of the Virtual Companions function, which can turn NFTs into AI Agents. These Agents will be able to use social media and play GameFi or DeFi by themselves, and create a market for AI Agents to buy and sell skills, memories and other information.
Screenshots from community player Fairu
The first game launched was Abstracts on-chain RPG game Gigaverse, which was generally well received by the community after trying it. As a result, the price of Treasures coin also rebounded accordingly. After the previous announcement of its withdrawal from GameFi business, the market value fell to 20 million US dollars, and today it has returned to 80 million US dollars. In addition to the expectations of the technology of AI-based NFT, the model of AI Agent playing GameFi may be more attractive. For the GameFi project, the problem that no one plays and the game is not fun for players may be solved by assetized AI Agent in the future.
DeFAI
After 18 months of development, Almanak announced last week on April 17 that it will conduct a token TGE in 2025. Almanak is an end-to-end platform that allows users to create, optimize and manage complex financial strategies using AI agents. Tasks such as market data analysis, strategy formulation, optimization and high-speed execution can be automated. The team has received support from many top venture capital firms and consultants, such as RockawayX, Delphi Labs, Hashkey, AppWorks, Matrix Partners, Bankless Ventures, etc.
GambleFAI
Sportstensor is the SN 41 subnet on dTAO. It is an AI platform for sports prediction that allows participants (miners and validators) to collaborate in developing and optimizing sports prediction models. Participants with better models and data sets that can predict the results of sports games will profit. Users can interact with the project through technical participation in developing models or non-technical participation in using prediction results.
For example, in an NBA game between the Celtics and the Magic, the odds were 0.92:0.08. If you followed the market odds and bet on the Celtics, the favored team, your winning rate was about 92%. Even with such a high winning rate, after multiple bets, most peoples return on investment was negative. Although the teams favored by the public tend to have a higher winning rate, the odds are also higher, which means that even if the prediction is correct, the winning money will be less. People tend to bet on the teams they are optimistic about, resulting in a low winning rate for the underdog team, which means that if you bet on the underdog team correctly, you can make a lot of money.
This is the advantage of the Sportstensor model, miners use their own data to run their own machine learning models to get the best results. Then, Sportstensor will take the average/median and use it as a smart indicator to identify market advantages. The difference between the odds predicted by the model and the market odds is the part that users can profit in the long run.
DeAI Infrastructure
This type of product is almost the earliest concept of the integration of Crypto and AI, such as Grass and other projects. There have been countless similar products before, but it is difficult to continue. The key to decentralized computing power is how to optimize the synchronous computing power and lower the selling price than traditional computing power providers, and the key to decentralized training is the cost of data transmission itself, which is difficult to achieve in the period of imperfect infrastructure, but once achieved, it will be a very potential sinking market, so it has attracted much attention from VCs.
PrimeIntellect was co-founded by Vincent Weisser and Johannes Hagemann, who were previously members of Descis leading VitaDAO. PrimeIntellect is a platform that commercializes computing power and models. The investment lineup is quite luxurious. The seed round was led by CoinFund and Distributed Global, raising a total of US$5 million, while the second round of financing was led by Founders Fund, raising a total of US$15 million. Among the individual investors, there are also industry influencers such as Polygon co-founder Sandeep Nailwal, well-known investors, and former CoinBase CTO Balaji.
Recently, Yaoshunyu, a well-known researcher at OpenAI, published an article titled The Second Half . He said that we are in the middle of the AI era and are about to enter the second half. If the AI in the first half solved the problems of playing games and answering exams, then the AI in the second half will build companies worth billions or trillions of dollars by building intelligent and useful products. For CryptoAI, there must be corresponding opportunities in this.
So look forward to it, and welcome to the second half.
