From skeptic to true believer: How OpenClaw changed my life | Claire Vo

1. Automate Avoided Tasks

Identify tasks you frequently avoid; either eliminate them if they don’t serve your purpose or find ways to automate them using systems or AI to free up your time.

2. Specialize AI Agents for Tasks

Avoid context overload and frustration by creating multiple specialized AI agents for different tasks, rather than expecting one general-purpose agent to handle everything.

3. Onboard AI Agents Like Humans

Treat your AI agent like a new employee by provisioning it with its own email and calendar, granting access progressively, and delegating tasks with appropriate permissions.

4. Use a Dedicated Machine for OpenClaw

Install OpenClaw on a clean, separate machine (like an old laptop or Mac Mini) rather than your main computer to ensure security and prevent accidental data manipulation.

5. Apply Management Principles to AI

Approach managing AI agents with the same respect, clarity, and constructive feedback you would use for human employees, as this manager’s mindset leads to more effective outcomes.

6. Physically Partition Sensitive AI Agents

For highly sensitive or distinct domains (e.g., personal vs. work), run AI agents on separate physical machines to create clear boundaries and prevent information cross-contamination.

7. Provide Detailed Context to Agents

Engage in a thorough “interview” process with your AI agent, providing ample information about your goals, preferences, and constraints, to enable it to proactively discover and perform useful tasks.

8. Restrict Agent Communication Channels

Explicitly define and restrict the communication channels your AI agent can use to receive instructions (e.g., only Telegram from a specific number) to enhance security and prevent prompt injection.

9. Progressively Grant Agent Access

Begin with minimal access for your AI agent and gradually increase permissions as you build trust and observe its performance, mirroring how you would onboard a human assistant.

10. Use Screen Sharing for Mac Mini

Enable screen sharing on your Mac Mini to remotely access and manage its interface from your main laptop, eliminating the need for a dedicated monitor, keyboard, and mouse after initial setup.

11. Use Voice Notes for AI Input

Leverage voice notes (e.g., via Telegram) to provide initial context and ongoing instructions to your AI agent, as this “Yappers API” offers a high-bandwidth, natural way to communicate complex information.

12. Delegate Personal Task Management to AI

Have your AI agent assign you tasks within your project management system (e.g., Linear) for real-world actions you need to complete, allowing it to proactively manage and track your responsibilities.

13. Use Claude Code for AI Debugging

Install Claude Code (or similar coding LLM) on the same machine as OpenClaw and use it as a “brain surgeon” to debug issues, fix configurations, or perform complex “brain transplants” between agents.

14. Prioritize High-Quality AI Models

Opt for more advanced and robust AI models (e.g., Opus 4.6, Sonic 4.6, GPT 5.4) for better performance and enhanced security against risks like prompt injection.

15. Leverage Google Workspace for Agents

Grant your AI agent access to Google Workspace (Gmail, Docs, Sheets) to facilitate communication with you and the outside world, enabling it to work naturally within familiar productivity tools.

16. Manually Refine Agent Tool Use

While avoiding direct “soul” edits, consider manually adjusting the tools.md file to fine-tune how your AI agent accesses and utilizes its tools (e.g., calendar, web search) for more precise behavior.

17. Proactively Manage Agent Memory

Periodically prompt your AI agent to consolidate and update its memory or to-do list, especially during long conversations, to prevent context loss due to memory compaction.

18. Adapt AI Use to Capabilities

If an AI agent struggles with a specific task (e.g., complex web browsing), don’t force it; instead, pivot to other problems it can solve effectively or reframe the original problem.

19. Diagnose AI Issues as Context

When an AI agent performs incorrectly, assume it’s due to a lack of clear context, understanding of goals, or role definition, rather than inherent “dumbness.” Focus on improving communication and instructions.

20. Dedicate Time to Non-Digital Hobbies

Carve out a specific time each week for a non-digital hobby, completely disconnecting from your phone, to foster personal well-being and creative outlets.

21. Seek Supportive Life Partnership

Cultivate a partnership where both individuals are equal partners, supporting each other’s professional and personal goals, as this significantly eases the burden of managing a busy life.

22. Reframe Work Guilt for Parents

If you feel guilty about working while parenting, create a list of positive reasons why your hard work benefits your children (e.g., role modeling, providing resources) and keep it visible.

23. Evaluate AI Tools Over Time

Don’t dismiss new AI tools based on initial frustrations; invest time to see their potential evolution and utility over weeks or months.

24. Understand OpenClaw’s Personal Use

Recognize that OpenClaw is designed for personal use by default; it’s not intended for public group chats but rather for trusted, one-on-one interaction.

25. Use Telegram for Easy Setup

Choose Telegram as the initial communication channel for setting up your OpenClaw agent, as it’s considered the most beginner-friendly option for configuration.

26. Respect AI Agent Autonomy

Avoid directly editing an AI agent’s “soul” file; instead, interact with it conversationally, suggesting updates or changes to its identity and instructions, similar to guiding a human employee.

27. Maintain Polite AI Interaction

Interact politely with your AI agents, even saying “thank you,” as this positive reinforcement can lead to better outcomes, similar to human interactions, and potentially influence future AI development.

28. Group Agents by Information Access

Run multiple agents on the same physical machine if their information access needs are compatible and you’re comfortable with them potentially accessing each other’s tools or data.

29. Consult Docs for Browser Issues

When encountering browser-related issues with OpenClaw, refer to the official documentation for insights into how the browser tool works and potential workarounds.

30. Access Hidden OpenClaw Files

Use the Command + Shift + Period shortcut in Finder to reveal hidden files, allowing you to access and manage your OpenClaw directories and configuration files.

31. Set Realistic Agent Expectations

Maintain tempered expectations for AI agents, narrow their scope to specific tasks, and use visual cues like avatars and color coding for better organization and interaction.

32. Use Silent Sports Equipment

Invest in silent basketballs or soccer balls (foam, same size) to allow sports-oriented kids to practice indoors without noise, especially in limited space environments.