In October 2024, since the launch of the AI Meme - GOAT and the inception of the AI Agent concept, Crypto began accelerating its integration with AI. Concepts like Game+AI, DeFAI, AIAgent Swarm, and others emerged in quick succession, with new conceptual projects appearing almost weekly, until January 18th when Trump announced the issuance of a MemeCoin, directly draining market liquidity and prematurely bursting the CryptoAI bubble. Two days later, DeepSeek announced its open-source R 1 model, and after weeks of fermentation, the AI concept stocks in the US stock market also saw their bubble burst.
Retail investors who could create a large model with $6 million and were buying the dips of CryptoAI with a $60 million market value were once defined by the community as "intellectually insufficient wheat" (a derogatory term). This trend, coupled with the collective dive of AI concept stocks in the US market, naturally led to a spiral death of coin prices, and since then, the AI concept in Crypto gradually faded.
One quarter later, the CryptoAI wave returned to the market. $zerebro rose 400% in 2 days, the Agent operating system $AIOS rebounded 10-fold, Dark rebounded again with a Flip exceeding 40 million USD in market value compared to the Musk concept, and other established AI projects like $ALCH, $SWARM, $GRIFFAIN, $arc, and $BUZZ all saw gains between 30% and 60%. This AI comeback brought not only price movements but also some new concepts.
[The rest of the translation follows the same professional and accurate approach, maintaining the technical terminology and context.]The Intern is an AI assistant that helps with operations, capable of promoting, replying, and managing communities on X. It can understand community culture through deep learning and generate images using TADA. Currently collaborating with Pudgypenguins to launch the Penguin Intern and operating its own Twitter. Based on the operational quality, if it can be completely independently operated by AI and scaled up, it would be a promising product.
BuzzingClub is a prediction market platform where the project believes the future of prediction markets should be in the hands of participants, not a central authoritative institution. "Everyone should be able to freely create, share, and express their opinions." Therefore, Buzzing is more open compared to other prediction platforms.
On Buzzing, all users can create prediction markets by "proposing topics or subjects" with rules generated by AI, then filtered by AI algorithms to remove spam and low-quality prediction markets. Finally, AI oracles automatically retrieve internet search data, rather than human data, to determine prediction results.
Burnie is a code learning platform that can improve users' coding skills in various aspects, and players can earn rewards by completing tasks published by the platform.
Inspiration as an App, Downstream Market Layout in the Post-AI Application Era
After the framework development gradually stagnated, Arc and Ai16z have also transformed into AI Agent distribution platforms. After Arc's distribution platform Forge went online with its first product AskJimmy, it disappeared. Ai16z's AutoFun went online a few days ago, but currently, no supported projects are online, and its development is yet to be seen. From the product framework perspective, AutoFun seems more inclined towards creating a UGC cultural platform, and in terms of value retention, it differs little from traditional LaunchPads.
Arc's planned Agentic App Store Ryzome and Myshell's existing AIApp Store, with the former not yet officially launched and the latter lacking activity, have mostly similar products.
In this context, dev.fun, which previously appeared with the AppFi concept, seems more orthodox. At first glance, its color scheme and UI are similar to Pumpfun, but it appears more vibrant. Although its token price has dropped significantly in this cycle like other AI projects, surprisingly, nearly 13,000 Apps have been born on this platform.
dev.fun provides the ability to generate Apps by chatting with AI "similar to YC-invested Replit". Users can not only issue project/meme Tokens but also choose their own trading pairs. Currently, the most supported Buidl has 1,400 Apps and has run nearly 70,000 times.
What AI is Currently More Favored in the Crypto Market?
Currently popular projects in the market can be divided into several categories. First, development tools, including frameworks, AI Copilot programming tools, and MCP infrastructure. Second, consumer AI applications, including AI Agents, games, DeFAI "Alpha signals, funds, automated LP", and GambleFAI. Third, decentralized AI infrastructure, such as decentralized computing, verification, and storage. The first two are usually more favored by retail investors, while the third is preferred by VCs or investors, typically requiring high valuations to acquire.
Development Tools
These projects have broader use cases and project concepts, but they are often the first choice for scam projects because retail investors cannot "see" a product. The feedback cycle for such products is usually longer, and whether it is "usable" or can generate viral spread "with people using it" is crucial. Therefore, creating such products requires high technical and marketing skills from a team.
The recently popular MCP project Dark is one that did both aspects well, launching DARK under the influence of MTN DAO and delivering a "visible" game product "Dark Forest" within two weeks.
In this field, the most widely spread project is probably SEND AI, Solana's "favorite child", gaining fame by hosting Solana's large AI hackathon, and Solana Agent Kit is used in many products. ALCHEMIST AI continues to build, falling to a $1.4 million market value in February and now returning to $14 million.
Autonome is a platform for developers and users to build, deploy, and distribute verifiable AI Agents without code, produced by Rollups protocol AltLayer. It has been active since the last cycle and has not yet issued a token.
Consumer AI
Games
Recently, Treasure announced shifting its focus from game chain and game operations to AI+Non-Fungible Token. Today, Smol provided its first response, announcing the release of "Virtual Companions" functionality that can transform Non-Fungible Tokens into AI Agents capable of using social media, playing GameFi or DeFi, and creating a market for buying and selling AI Agent skills, memories, and other information.
Community player Fairu's actual screenshot
The first game launched is Abstract's on-chain RPG game Gigaverse, which received generally positive feedback from the community. Consequently, Treasure's token price rebounded, rising from a $20 million market value after announcing its exit from GameFi to $80 million today. Besides expectations for NFT AI technology, the mode of AI Agents playing GameFi might be more attractive, potentially solving the problems of no players for game projects and unenjoyable games for players through asset-based AI Agents.
DeFAI
After 18 months of development, Almanak announced on April 17th that it will implement token TGE in 2025. Almanak is an end-to-end platform allowing users to create, optimize, and manage complex financial strategies using AI agents. It includes functions for market data analysis, strategy formulation, optimization, and high-speed execution. The team is supported by numerous top venture capital firms and advisors, such as RockawayX, Delphi Labs, Hashkey, AppWorks, Matrix Partners, and Bankless Ventures.
GambleFAI
Sportstensor is an SN 41 subnet on dTAO, an AI platform for sports event prediction that allows participants "miners and verifiers" to collaboratively develop and optimize sports prediction models. Participants with better models and datasets can profit from predicting sports competition results. Users can interact with the project through technical participation "developing models" or non-technical participation "using prediction results".
For example, in an NBA game between the Celtics and Magic with odds of 0.92:0.08, if you bet on the Celtics, the "crowd-favorite team," your winning probability would be around 92%. However, even with such high odds, most people's investment returns are negative. Although crowd-favorite teams often have higher win rates, their odds are also higher, which means you'll win less money even if you predict correctly. People tend to bet on the teams they favor, causing weak teams to have low win rates, which means if you correctly bet on a weak team, you can earn a lot of money.
This is the advantage of the Sportstensor model, where miners use their own data to run machine learning models to obtain the best results. Then, Sportstensor takes their average/median and uses it as a smart indicator for identifying market advantages. The difference between the model-predicted odds and market odds is the part that users can profit from in the long term.
DeAI Infrastructure
This type of product is almost the earliest concept of merging Crypto and AI, like the Grass project. Previously, numerous similar products appeared but were difficult to sustain. The key to decentralized computing power is how to optimize synchronized computing power and price it lower than traditional computing power providers. The key to decentralized training is the cost of data transmission itself, which is hard to achieve during immature infrastructure periods. However, once achieved, it will be a promising downstream market, thus attracting significant VC attention.
PrimeIntellect was co-founded by Vincent Weisser and Johannes Hagemann, who were previously members of Desci leader VitaDAO. PrimeIntellect is a platform that commodifies computing power and models. Its investment lineup is impressive, with a seed round of $5 million led by CoinFund and Distributed Global, and a second round of $15 million led by Founders Fund. Personal investors include Polygon co-founder Sandeep Nailwal, notable investors, and former CoinBase CTO Balaji, among other industry influencers.
Recently, OpenAI's renowned researcher Yaoshunyu published an article titled "The Second Half", stating that we are currently in the midpoint of the AI era, and the next phase is about to begin. If the first half of AI solved game-playing and exam-answering, the second half will establish billion or trillion-dollar companies by building intelligent and useful products. For CryptoAI, opportunities definitely exist within this context.
So look forward to it, welcome to the second half.
