BlockBeats | April 24, 2025, 17:00

The recent resurgence of AI in crypto has not only revived token prices but also introduced fresh concepts to the space.

In October 2024, the launch of GOAT, an AI Meme token, ignited the AI Agent trend in crypto. Concepts like Game+AI, DeFAI, and AIAgent Hive flooded the market, with new projects emerging weekly. However, the bubble burst prematurely on January 18, 2025, when former U.S. President Trump announced his MemeCoin, draining market liquidity. Two days later, DeepSeek’s open-source R1 model popped the AI stock bubble on Wall Street after weeks of hype.

With models now costing as little as 6milliontodevelop,retailinvestorsscramblingtobuyCryptoAItokensvaluedat60 million were labeled "low-IQ韭菜" (retail fodder) by the community. The sentiment, combined with the AI stock crash, sent token prices into a death spiral, silencing AI hype in crypto.

But a quarter later, AI tokens are making a comeback. zerebrosurged400AIOS rebounded 10x, and Dark flipped Musk-themed projects worth over 40millioninmarketcap.LegacyAIprojectslikeALCH, SWARM,GRIFFAIN, arc,andBUZZ also posted gains ranging from 30% to 60%. This resurgence brings new concepts alongside token rallies.

Following Pumpfun’s success, crypto saw a surge in asset launch platforms. During the AI Agent boom, frameworks and AI distribution networks/LaunchPads dominated FDV rankings. However, frameworks like Ai16z’s developer community hit a ceiling—they struggled to capture value in crypto, prompting survivors to pivot to LaunchPads. But AI LaunchPads faced challenges: they couldn’t meet AI product launch demands. Project teams are now scrambling for solutions ahead of the next AI wave.

Despite a recent incident where the SN28 subnet was exploited to turn TAO into a MemeCoin (later centrally intervened by the foundation), Bittensor’s future as a decentralized "attention network" remains uncertain. Long-time observer "WeirdMind" even called it a scam.

Yet, from an investment standpoint, Bittensor’s liquidity outshines other AI Agent ecosystems. Take Virtuals: its pairing with LPs leads to higher volatility for liquidity providers, with 3–7% slippage for in-platform agent tokens. In contrast, dTAO subnet tokens see just 0.05–0.1% slippage.

This makes Bittensor a favorite for VC-backed AI projects. Last week, former Messari analyst and Crucible Labs partner Sami Kassab announced a fund with Seth Bloomberg to provide liquidity for Bittensor.

Rayon Labs, Bittensor’s flagship project, has rolled out several products highlighting their long-term, practical focus:

• SN64 "Chutes": A serverless AI infrastructure deployment tool. AWS outages underscore the need for decentralized alternatives—critical for finance-linked crypto, where downtime costs far exceed traditional AI.

• SN56 "Gradients": A no-code AI model deployment platform. The v3 update offers competitive pricing, letting users train custom LLMs for specific use cases or image generation.

• SN19 "Nineteen": A fast, scalable, decentralized AI inference platform.

Once a top AI project combining ecosystem-building and a value flywheel, Virtuals Protocol’s token price slowed with the market downturn, shedding over 90% of its $4.5 billion market cap. LaunchPad participation dwindled. But Virtuals didn’t quit—they doubled down during the bear market.

1. Building an AI Talent Ecosystem
Virtuals launched the VPN (Virtuals Partners Network), connecting investors, experts, academics, and developers. It’s a one-stop incubator for crypto newcomers, offering resources from funding to marketing.

2. Agent Commerce Protocol (ACP)
ACP is a concrete version of the "hive" concept from projects like Swarm and Ai16z. It creates a virtual nation where AI agents interact, collaborate, and trade autonomously. Google later unveiled a similar A2A concept, though ACP uses smart contracts while A2A relies on protocols.

3. Virgen Points & Genesis Launch Model
In April, Virtuals introduced Virgen Points and the Genesis Launch model. Users earn points by investing in Sentient/Prototype agents, holding Virtuals, or staking VADER. Points determine Genesis Launchpad investment allocations. Only Virtuals-vetted participants can join, ensuring quality and fairness.

This model boosts user retention ("We reward believoors," says founder Ethermage) and attracts high-quality participants, fostering sustainable project growth.

Virtuals’ April 21 hackathon featured 100+ projects judged by luminaries like Base’s LucaCurran (AI/DEPIN lead), Stanford Blockchain’s KunPeng, and Canonical Crypto’s Anand Iyer (X handle: "AI").

1. The Intern
An AI operations assistant that promotes, replies, and manages communities on X. It learns community culture and generates images via TADA. Partnered with Pudgypenguins, it runs its own Twitter. If fully AI-operated and scalable, it could be a game-changer.

2. BuzzingClub
A decentralized prediction market platform where users create markets (topics) with AI-generated rules. AI filters spam and low-quality markets, while an AI oracle aggregates internet data (not human input) for predictions.

3. Burnie
A coding learning platform where users earn rewards by completing tasks.

After framework growth stalled, Arc and Ai16z pivoted to AI Agent distribution platforms. Arc’s Forge launched quietly after AskJimmy, while Ai16z’s AutoFun debuted days ago with no live projects yet. AutoFun leans toward a UGC-focused community platform, resembling traditional LaunchPads in value retention.

Meanwhile, Arc’s upcoming Agentic App Store Ryzome and Myshell’s AIApp Store lack traction, with similar offerings. Dev.fun, an early AppFi player, stands out with a Pumpfun-like UI but more energy. Despite a token price slump, it hosts ~13,000 apps.

Dev.fun lets users generate apps via AI chat (like YC-backed Replit) and issue project/Meme tokens with custom trading pairs. Buidl, the most-supported app, has 1,400 apps and ~70,000 runs.

The market favors three categories:

1. Development Tools: Frameworks, AI Copilot-style coding tools, and MCP infrastructure.

2. Consumer AI Apps: AI agents, games, DeFAI (alpha signals, funds, automated LPs), and GambleFAI.

3. Decentralized AI Infrastructure: Computing, validation, and storage.

Retail investors prefer the first two, while VCs favor the third, which demands high valuations.

Development Tools
These projects have broad use cases but attract scams due to their intangible nature. Success hinges on usability and viral adoption. Dark, an MCP project, nailed this by launching the playable "Dark Forest" game within two weeks under MTN DAO’s influence.

Notable players include Solana’s "golden child" SEND AI (known for hosting Solana’s AI hackathon) and ALCHEMIST AI, which rebounded from a 1.4millionto14 million market cap.

Autonome
A no-code platform for building, deploying, and distributing verifiable AI agents, backed by Rollups protocol AltLayer. No token yet.

Smol
Treasure shifted focus from gaming to AI+NFTs, launching "Virtual Companions" to turn NFTs into AI agents that use social media, play GameFi/DeFi, and trade skills/memories.

The first game, Abstract’s Gigaverse (an on-chain RPG), earned praise, boosting Treasure’s token price from 20millionto80 million. The AI agent-driven GameFi model could solve player retention and fun factor issues.

Almanak
After 18 months, Almanak announced a 2025 TGE for its end-to-end platform automating financial strategy creation, optimization, and execution. Backed by VCs like RockawayX, Delphi Labs, and Hashkey.

Sportstensor
A dTAO SN41 subnet for sports prediction markets. "Miners/validators" collaborate on models; those with better predictions profit. The platform highlights market inefficiencies—e.g., betting on underdogs in NBA games (like the Celtics vs. Magic) can yield high returns if models outperform odds.

Early crypto-AI hybrids like Grass struggled to sustain decentralized compute or training due to infrastructure costs. Yet, if solved, they’d tap a massive VC-backed market.

PrimeIntellect
Founded by Desci’s VitaDAO veterans Vincent Weisser and Johannes Hagemann, PrimeIntellect commoditizes compute and models. It raised 5M(CoinFund,DistributedGlobal)inseedand15M (Founders Fund) in Series A, with backers like Polygon’s Sandeep Nailwal and Coinbase’s ex-CTO Balaji.

OpenAI researcher Yaoshunyu’s essay "The Second Half" argues we’re midway through the AI era. If the first half solved gaming and exam questions, the next will build intelligent products worth billions or trillions. CryptoAI will play its part.

Welcome to the second half.

BlockBeats | April 24, 2025, 17:00

The recent resurgence of AI in crypto has not only revived token prices but also introduced fresh concepts to the space.

In October 2024, the launch of GOAT, an AI Meme token, ignited the AI Agent trend in crypto. Concepts like Game+AI, DeFAI, and AIAgent Hive flooded the market, with new projects emerging weekly. However, the bubble burst prematurely on January 18, 2025, when former U.S. President Trump announced his MemeCoin, draining market liquidity. Two days later, DeepSeek’s open-source R1 model popped the AI stock bubble on Wall Street after weeks of hype.

With models now costing as little as 6milliontodevelop,retailinvestorsscramblingtobuyCryptoAItokensvaluedat60 million were labeled "low-IQ韭菜" (retail fodder) by the community. The sentiment, combined with the AI stock crash, sent token prices into a death spiral, silencing AI hype in crypto.

But a quarter later, AI tokens are making a comeback. zerebrosurged400AIOS rebounded 10x, and Dark flipped Musk-themed projects worth over 40millioninmarketcap.LegacyAIprojectslikeALCH, SWARM,GRIFFAIN, arc,andBUZZ also posted gains ranging from 30% to 60%. This resurgence brings new concepts alongside token rallies.

Following Pumpfun’s success, crypto saw a surge in asset launch platforms. During the AI Agent boom, frameworks and AI distribution networks/LaunchPads dominated FDV rankings. However, frameworks like Ai16z’s developer community hit a ceiling—they struggled to capture value in crypto, prompting survivors to pivot to LaunchPads. But AI LaunchPads faced challenges: they couldn’t meet AI product launch demands. Project teams are now scrambling for solutions ahead of the next AI wave.

Despite a recent incident where the SN28 subnet was exploited to turn TAO into a MemeCoin (later centrally intervened by the foundation), Bittensor’s future as a decentralized "attention network" remains uncertain. Long-time observer "WeirdMind" even called it a scam.

Yet, from an investment standpoint, Bittensor’s liquidity outshines other AI Agent ecosystems. Take Virtuals: its pairing with LPs leads to higher volatility for liquidity providers, with 3–7% slippage for in-platform agent tokens. In contrast, dTAO subnet tokens see just 0.05–0.1% slippage.

This makes Bittensor a favorite for VC-backed AI projects. Last week, former Messari analyst and Crucible Labs partner Sami Kassab announced a fund with Seth Bloomberg to provide liquidity for Bittensor.

Rayon Labs, Bittensor’s flagship project, has rolled out several products highlighting their long-term, practical focus:

• SN64 "Chutes": A serverless AI infrastructure deployment tool. AWS outages underscore the need for decentralized alternatives—critical for finance-linked crypto, where downtime costs far exceed traditional AI.

• SN56 "Gradients": A no-code AI model deployment platform. The v3 update offers competitive pricing, letting users train custom LLMs for specific use cases or image generation.

• SN19 "Nineteen": A fast, scalable, decentralized AI inference platform.

Once a top AI project combining ecosystem-building and a value flywheel, Virtuals Protocol’s token price slowed with the market downturn, shedding over 90% of its $4.5 billion market cap. LaunchPad participation dwindled. But Virtuals didn’t quit—they doubled down during the bear market.

1. Building an AI Talent Ecosystem
Virtuals launched the VPN (Virtuals Partners Network), connecting investors, experts, academics, and developers. It’s a one-stop incubator for crypto newcomers, offering resources from funding to marketing.

2. Agent Commerce Protocol (ACP)
ACP is a concrete version of the "hive" concept from projects like Swarm and Ai16z. It creates a virtual nation where AI agents interact, collaborate, and trade autonomously. Google later unveiled a similar A2A concept, though ACP uses smart contracts while A2A relies on protocols.

3. Virgen Points & Genesis Launch Model
In April, Virtuals introduced Virgen Points and the Genesis Launch model. Users earn points by investing in Sentient/Prototype agents, holding Virtuals, or staking VADER. Points determine Genesis Launchpad investment allocations. Only Virtuals-vetted participants can join, ensuring quality and fairness.

This model boosts user retention ("We reward believoors," says founder Ethermage) and attracts high-quality participants, fostering sustainable project growth.

Virtuals’ April 21 hackathon featured 100+ projects judged by luminaries like Base’s LucaCurran (AI/DEPIN lead), Stanford Blockchain’s KunPeng, and Canonical Crypto’s Anand Iyer (X handle: "AI").

1. The Intern
An AI operations assistant that promotes, replies, and manages communities on X. It learns community culture and generates images via TADA. Partnered with Pudgypenguins, it runs its own Twitter. If fully AI-operated and scalable, it could be a game-changer.

2. BuzzingClub
A decentralized prediction market platform where users create markets (topics) with AI-generated rules. AI filters spam and low-quality markets, while an AI oracle aggregates internet data (not human input) for predictions.

3. Burnie
A coding learning platform where users earn rewards by completing tasks.

After framework growth stalled, Arc and Ai16z pivoted to AI Agent distribution platforms. Arc’s Forge launched quietly after AskJimmy, while Ai16z’s AutoFun debuted days ago with no live projects yet. AutoFun leans toward a UGC-focused community platform, resembling traditional LaunchPads in value retention.

Meanwhile, Arc’s upcoming Agentic App Store Ryzome and Myshell’s AIApp Store lack traction, with similar offerings. Dev.fun, an early AppFi player, stands out with a Pumpfun-like UI but more energy. Despite a token price slump, it hosts ~13,000 apps.

Dev.fun lets users generate apps via AI chat (like YC-backed Replit) and issue project/Meme tokens with custom trading pairs. Buidl, the most-supported app, has 1,400 apps and ~70,000 runs.

The market favors three categories:

1. Development Tools: Frameworks, AI Copilot-style coding tools, and MCP infrastructure.

2. Consumer AI Apps: AI agents, games, DeFAI (alpha signals, funds, automated LPs), and GambleFAI.

3. Decentralized AI Infrastructure: Computing, validation, and storage.

Retail investors prefer the first two, while VCs favor the third, which demands high valuations.

Development Tools
These projects have broad use cases but attract scams due to their intangible nature. Success hinges on usability and viral adoption. Dark, an MCP project, nailed this by launching the playable "Dark Forest" game within two weeks under MTN DAO’s influence.

Notable players include Solana’s "golden child" SEND AI (known for hosting Solana’s AI hackathon) and ALCHEMIST AI, which rebounded from a 1.4millionto14 million market cap.

Autonome
A no-code platform for building, deploying, and distributing verifiable AI agents, backed by Rollups protocol AltLayer. No token yet.

Smol
Treasure shifted focus from gaming to AI+NFTs, launching "Virtual Companions" to turn NFTs into AI agents that use social media, play GameFi/DeFi, and trade skills/memories.

The first game, Abstract’s Gigaverse (an on-chain RPG), earned praise, boosting Treasure’s token price from 20millionto80 million. The AI agent-driven GameFi model could solve player retention and fun factor issues.

Almanak
After 18 months, Almanak announced a 2025 TGE for its end-to-end platform automating financial strategy creation, optimization, and execution. Backed by VCs like RockawayX, Delphi Labs, and Hashkey.

Sportstensor
A dTAO SN41 subnet for sports prediction markets. "Miners/validators" collaborate on models; those with better predictions profit. The platform highlights market inefficiencies—e.g., betting on underdogs in NBA games (like the Celtics vs. Magic) can yield high returns if models outperform odds.

Early crypto-AI hybrids like Grass struggled to sustain decentralized compute or training due to infrastructure costs. Yet, if solved, they’d tap a massive VC-backed market.

PrimeIntellect
Founded by Desci’s VitaDAO veterans Vincent Weisser and Johannes Hagemann, PrimeIntellect commoditizes compute and models. It raised 5M(CoinFund,DistributedGlobal)inseedand15M (Founders Fund) in Series A, with backers like Polygon’s Sandeep Nailwal and Coinbase’s ex-CTO Balaji.

OpenAI researcher Yaoshunyu’s essay "The Second Half" argues we’re midway through the AI era. If the first half solved gaming and exam questions, the next will build intelligent products worth billions or trillions. CryptoAI will play its part.

Welcome to the second half.