Keep pulling the thread on Claire Vo.
For security, OpenClaw agents should be set up with their own dedicated email and calendar accounts and then be granted delegated access, similar to onboarding a human executive assistant.
Claire Vo considers OpenClaw to be the most mind-blowing and important AI experience she has had since the launch of ChatGPT.
Claire Vo manages context window limitations in OpenClaw by creating multiple, specialized agents, each dedicated to a specific domain like work or family, to prevent context overload.
For security and separation of concerns, Claire Vo runs her family-focused OpenClaw agent ("Finn") on a physically separate machine from her work-related agents.
Claire Vo uses an OpenClaw agent named "Sam" as an automated Sales Development Representative that scans her CRM daily, uses Exa to research prospects, and sends them introductory emails.
The OpenClaw agent "Sam" replaced a human contractor who was previously paid to work 10 hours per week on sales tasks for Claire Vo's company.
Claire Vo predicts that within a few years, the primary users of websites will be AI agents, not humans.
A powerful technique for debugging OpenClaw is to use a separate coding agent like Claude Code on the same machine to diagnose and fix configuration issues.
NVIDIA founder Jensen Huang has stated that every company needs an "OpenClaw strategy," referring to it as "the new computer."
OpenClaw is the fastest-growing open-source project in history.
Claire Vo recommends using high-end models like Claude 3 Opus, Claude 3 Sonnet, and GPT-4 with OpenClaw for better performance and security.
The maintainers of OpenClaw, led by Peter, have implemented significant work to harden the platform against security risks like prompt injection.