OpenClaw vs. Hermes Agent: The race to build AI assistants that never forget (thenewstack.io)

OpenClaw and Hermes Agent are two emerging open-source projects aiming to revolutionize AI assistants by addressing the common challenge of context loss between sessions. OpenClaw, developed by Peter Steinberger, has rapidly gained popularity with over 345,000 stars on GitHub since its inception, offering extensive integration with messaging platforms like WhatsApp, Telegram, and Discord. Designed for persistent operation, it allows users to interact with the assistant across various channels while maintaining continuity and context. However, it faced security issues, including a significant supply chain attack labeling it a "security nightmare". The adoption of VirusTotal for skill scanning and better security measures is an ongoing effort to enhance trust in its ecosystem. In contrast, Hermes Agent, from Nous Research, focuses on deep learning capabilities and memory persistence. It uses advanced techniques like full-text search with LLM summarization to recall past interactions and autonomously create skills for recurring tasks. The latest version supports multiple isolated instances for team workflows and promotes a self-training loop for continual improvement. Hermes emphasizes security with architectural constraints and has remained relatively free from major security incidents. Both projects highlight a shift towards long-lived AI assistants capable of evolving with user needs, indicating a future where such technologies could be integral to workflows, but also raising questions around knowledge ownership and security in this increasingly persistent environment.