Behind Hermes Agent flooding the screens, Web3 forces are infiltrating the AI battlefield

Author: Nancy, PANews

Recently, a dark horse suddenly emerged on GitHub’s star ranking list. The open-source autonomous AI agent framework Hermes Agent, which has been only a little over a month old, has quickly accumulated over 70k stars. Projects reaching this star scale are less than 0.0014% in the historical rankings.

Last weekend, Hermes Agent dominated online discussions, with many developers abandoning “Lobster” to switch to “Hermes.” This hot new star is backed by the Web3 community.

In fact, Web3 is becoming an influential force of innovation and output within the AI ecosystem.

Behind the AI hit products, there are a group of Web3 veterans

As AI rapidly explodes, a wave of talent migration from the Web3 world is emerging.

Moltbook, an AI agent social network derived from OpenClaw, once caused a phenomenon outside the circle, and was eventually acquired by tech giant Meta. Its developer, Matt Schlicht, has a strong background in crypto and is a serial entrepreneur.

The team behind the popular Hermes Agent is Nous Research, an open-source AI research organization. The project originated from an AI community on Discord. CEO Jeffrey Quesnelle previously served as the lead engineer for Eden Network, an Ethereum MEV infrastructure project, giving Nous a Web3 DNA from the start.

Alex Atallah, founder of OpenRouter, the world’s largest AI model aggregation platform, was a co-founder of OpenSea. In 2022, before the NFT boom fully subsided, he chose to leave the Web3 track and pivot to AI entrepreneurship.

Behind these rapidly emerging products, there are entrepreneurs with crypto industry experience. As the crypto market enters a downturn, Web3 talent naturally flows into the AI field, which offers new narratives and capital enthusiasm. The volatility of crypto cycles also enhances the resilience and iteration capabilities of industry practitioners. Of course, during the previous Web3 boom, the sector was a hub for global capital and talent, attracting many excellent developers. Now, as AI becomes a more explosive and long-term promising direction, some of these people are shifting toward AI.

However, fundamentally, the ability of these products to achieve rapid cold starts and community explosive growth mainly stems from technological innovation and genuine market demand. Web3 experience is more of an embellishment rather than the core driving force.

Additionally, Web3 talent is also demonstrating advantages in AI security investigation. In March, AI giant Anthropic experienced a major security incident when approximately 512k lines of TypeScript code and over 1,900 files from its programming tool Claude Code were leaked unexpectedly. The first discoverer was Chaofan Shou, a researcher from blockchain security firm Fuzzland and also an engineer at Solayer Labs. Moreover, blockchain security agencies like SlowMist frequently monitor vulnerabilities and risks in AI systems.

Compared to traditional developers, Web3 security practitioners are more sensitive to attack surfaces, vulnerabilities, and abnormal behaviors. The bounty hunter culture within the industry has cultivated high-level security researchers. This capability is gradually being transferred to AI security auditing and risk detection, injecting greater resilience into the AI ecosystem.

Web3 community, the ideal early adopters of AI

The Web3 community is naturally an ideal early user base for AI as a productivity tool.

Take OpenClaw as an example. Before truly breaking out, the crypto community had already participated early, becoming one of the first groups of users and dissemination nodes. Although market hype once troubled developer Peter Steinberger, the enthusiasm and engagement of crypto users in the early stages were undeniable, including offline lobster events organized mainly by Web3 practitioners across various regions.

Another example is Worldcoin, co-founded by OpenAI CEO Sam Altman, which aims to solve the “Witch Attack” problem in the AI era by combining biometrics, zero-knowledge proofs (ZK), and blockchain to verify real identities. However, without incentives, ordinary users might lack motivation to participate. Based on this, Altman proposed the concept of universal basic income (UBI) in the AI era, providing periodic rewards to verified users in the form of frictionless tokens worldwide.

From these cases, compared to ordinary people, Web3 users tend to be more optimistic about technology and early adopters. They are generally willing to accept higher learning costs and tolerate uncertainties in experimental products. When many products are still immature, they are already willing to participate in testing, feedback, and even secondary dissemination. This trait aligns well with the core user profile needed for Agent’s transition into an economic tool.

In terms of usage habits, Web3 users are already accustomed to complex on-chain interactions, such as wallet signatures, transaction execution, smart contract calls, and multi-protocol operations. Having crossed a high learning curve, they are more capable of learning and more receptive to prompts engineering, token consumption, authorization mechanisms, or permission management in Agent.

Meanwhile, the crypto sector’s inherent financial incentives and speculative nature make information dissemination and user growth easier to exponentially amplify, further strengthening the spread of attention economy. Coupled with Web3 users’ high activity on social platforms like X, Discord, and Telegram—communities with high information dissemination efficiency—new products or narratives can quickly spread, ferment, and expand within short cycles.

Although Web3 users are among the dense experimental early adopters of AI products, long-term retention still depends on the product itself.

From ecosystem to infrastructure, Web3 is “invading” AI

Web3’s methods are also “invading” AI.

The rapid expansion path of Nous clearly carries Web3 characteristics. Starting in 2024, Nous introduced a DePIN mechanism, incentivizing users to contribute idle GPU and computing power for model training. This token airdrop expectation directly stimulated community participation and accelerated product cold start.

Funding scale and lineup further boosted community airdrop expectations. So far, Nous has raised over $70 million through seed and Series A rounds, with the Series A valuation reaching $1 billion in tokens. The investor lineup is highly crypto-oriented, including Paradigm, Delphi Ventures, North Island, Solana co-founder Raj Gokal, and OpenSea founder Alex Atallah.

Today, AI competition has long surpassed mere technical rivalry. Crypto governance thinking and incentive design have expanded Nous’s user engagement and ecosystem capabilities, laying a foundation for Hermes Agent’s breakout. It’s worth noting that Nous has not yet explicitly launched a token, so airdrops remain uncertain.

Similar approaches are also seen in AI ecosystem tools like OpenRouter. As a platform that helps developers unify access to and select AI models, founder Alex Atallah’s initial logic was that if a model is already so cheap, then in the future, there will be thousands of models, each needing its own internet presence. OpenRouter functions similarly to how OpenSea served as a liquidity aggregator in the NFT space.

This Web3 approach of aggregating liquidity has benefited OpenRouter during the explosive growth of AI models. Currently, it has integrated over 300 models, with more than 5 million global users, and its annualized revenue has grown from about $10 million last October to over $50 million. In terms of funding, OpenRouter has completed a $40 million round and is pushing forward a new round of approximately $120 million, with a post-money valuation of around $1.3 billion.

The crypto world is also infiltrating AI infrastructure. Cryptocurrency mining was once a highly profitable “lay-and-earn” business, with miners earning daily profits through low electricity costs and high coin prices. Even NVIDIA, a major AI giant, made a fortune from GPU mining demand before officially entering the AI track.

Although Bitcoin halving has compressed profits, increased energy costs, and price volatility has challenged many miners’ survival, AI is becoming a new growth curve. Compared to building new data centers that often take years, crypto miners have unique advantages: large-scale power contracts, data center infrastructure, cooling systems, land, and power grids—resources that are scarce for AI data centers. As a result, leading mining companies like Bitdeer, Riot Platforms, and MARA Holdings are shifting toward AI/HPC (High-Performance Computing).

In the vibrant AI battlefield, from upstream computing power and midstream model frameworks to downstream application scenarios, the presence of Web3 is becoming increasingly frequent.