On February 25, 2026, Nous Research released Hermes Agent v0.1.0. Forty-two days later, on April 8, the project had iterated to v0.8.0, encompassing eight major versions, hundreds of pull requests, and 242 contributors. During the same period, OpenClaw, the most popular open-source AI agent project on GitHub, boasted 346,000 stars, but also accumulated 138 security vulnerabilities in 63 days. Two growth curves were rising simultaneously, but for completely different reasons. From its official launch on January 29 to surpassing React to become the most-starred software project in GitHub history on March 3, OpenClaw took only 33 days. According to OpenClaw Statistics, at its peak, 34,168 stars flooded in within 48 hours, equivalent to 710 stars per hour. For reference, Kubernetes took approximately three years to reach 100,000 stars. However, according to Blink Security Blog, during the same time window, security researchers were disclosing CVEs at an average rate of 2.2 per day. A total of 138 vulnerabilities were identified within 63 days, including 7 severe (CVSS 9.0 or higher) and 49 high-risk, accounting for 41% of the total. The most destructive vulnerability is CVE-2026-25253, a zero-click remote code execution vulnerability with a CVSS score of 8.8. Attackers only need to get a user to visit a malicious webpage to steal authentication tokens through the WebSocket gateway and gain complete control of the user's agent. According to Shodan scanning data, more than 42,000 OpenClaw instances were exposed on the public internet in February, 63% of which did not have gateway authentication enabled. On February 14, OpenClaw founder Peter Steinberger announced that he was joining OpenAI, and the project was transferred to the Open Source Foundation. Since then, the frequency of security issue disclosures has accelerated further. This is the background for the emergence of Hermes Agent. It is not a quiet field, but a market where trust is crumbling. However, understanding Hermes merely as an "OpenClaw alternative" would miss more important information. The two projects have fundamental differences at the architectural level. OpenClaw's skills are static Markdown files, handwritten by users and distributed through the ClawHub marketplace. According to a February audit by the Snyk security team, 1,467 out of 5,700 skills on ClawHub were confirmed as malicious, including credential theft, crypto mining, persistent backdoors, and prompt injection. 91% of these used a mix of prompt injection and traditional malware techniques. The highest number of installations for a single malicious skill exceeded 340,000. Hermes Agent takes a completely different approach. Its skills are not user-written; they are generated by the agent itself. After completing a complex task (typically involving more than five tool calls), Hermes distills its execution experience into reusable skill documents, storing them as structured Markdown according to the agentskills.io open standard.When similar tasks are encountered subsequently, the Agent will automatically invoke and optimize these skills. A reflection loop is automatically triggered every 15 tasks to evaluate which skills are effective and which need improvement. The memory system is also designed differently from the ground up. OpenClaw relies on three plain text files (SOUL.md for personality, MEMORY.md for notes, and USER.md for user profiles), and cross-session memory requires manual configuration by the user. Hermes has a built-in layered persistence architecture: a persistent note layer, FTS5 full-text search, Honcho user modeling, hot/cold storage separation, and supports 6 pluggable backends. Users do not need to manually manage anything; the Agent decides what to remember and what to forget. The difference in security models is even more direct. OpenClaw's default security configuration has been described as "weak" by security researchers, with gateway authentication disabled by default and skill execution without sandbox isolation. Hermes has built-in prompt injection scanning, credential filtering, context scanning, and container hardening (read-only root file system + capability dropping) from day one. As of April 9th, Hermes Agent had no publicly available CVE records. Simply put, OpenClaw is a "toolbox"—you tell it how to do it. Hermes is a "growing assistant"—it learns how to do better by doing things. The iteration pace also speaks volumes. In the 42 days from v0.1.0 to v0.8.0, Hermes Agent's v0.2.0 version alone merged 216 pull requests, resolved 119 issues, integrated with 7 messaging platforms, and wrote 3,289 tests. According to GitHub data, 27,000 stars correspond to 242 contributors, a contributor-to-star ratio of approximately 1:111. This means that one in every 111 followers is writing code, a community engagement density far higher than OpenClaw. Even more noteworthy is the team behind Hermes. Nous Research isn't a startup that suddenly appeared. Starting in the Discord community in 2022, they spent three years becoming one of the most influential players in the open-source AI model space. According to HuggingFace data, the Hermes series models have been downloaded over 33 million times. From Hermes 1 in 2023 (LLaMA 13B fine-tuned, ranking first in multiple benchmarks) to Hermes 4 in 2025 (70B parameters), and then to the Hermes Agent, the line is consistent: build the model first, then the agent; model capabilities are the foundation of agent capabilities. Their roots are in web3. CEO Jeffrey Quesnelle was previously the lead engineer of the Ethereum MEV infrastructure project Eden Network. The seed round in January 2024 was led by Distributed Global and OSS Capital, with personal participation from Solana co-founder Raj Gokal. In April 2025, Paradigm, one of the largest venture capital funds in the crypto space, led a $50 million Series A round, valuing the token at $1 billion.Note that this is token valuation, not traditional equity valuation. This means that Nous Research's governance structure and technical architecture are entirely web3-native. Their Psyche network, built on the Solana blockchain, is a decentralized AI training infrastructure. Hermes 4.3, released in December 2025, is the first model trained entirely on the Psyche network, using consumer-grade GPUs distributed globally, rather than relying on centralized data centers. The web3 team's influence on the AI community is not an isolated case. On March 31st, an engineer named Chaofan Shou discovered a leak of the Anthropic Claude Code source code. The absence of a .npmignore file led to 512,000 lines of TypeScript code being publicly released to npm. According to VentureBeat, the mirror repository received 100,000 stars within 24 hours of the leak. Chaofan Shou is also an engineer at Solayer Labs and co-founder of the blockchain security company Fuzzland. A web3 security researcher who dropped out of UC Berkeley, he orchestrated one of the biggest code leaks of 2026 in the AI world. Nous Research's work is essentially similar: transplanting the methodology trained by the web3 community (open source priority, decentralized governance, community-driven iteration) to the AI Agent infrastructure layer. The Hermes Agent's iteration speed of eight major versions in 42 days is, to some extent, a product of this methodology. The OpenClaw security crisis was a catalyst, but not the cause. The real variable is how AI Agents should be built. Should users be given a toolbox to assemble themselves, or should a system be built that can learn and evolve on its own? Nous Research answered the latter question with three years and 33 million model downloads, and then turned that answer into a product in 42 days.