First Impressions of OpenClaw

During the Spring Festival, OpenClaw was everywhere on my feed. My first reaction was to listen to a podcast interview with the author. Initially, I thought it was just a local version of Manus integrated with IM platforms — nothing particularly innovative. But I soon realized there was much more to it.

I discovered several fundamental differences:

1. The Value of Memory

Manus’s memory is confined to individual tasks, but OpenClaw builds a persistent memory system that fundamentally transforms this AI agent. With memory comes growth. We often say “life’s most precious memories” — isn’t that exactly the value of memory? With accumulated memories, the agent can continuously interact with its environment, understand you better, and evolve on its own.

2. The Boldness of Permissions

OpenClaw essentially gains access to all your computer’s permissions. That’s quite bold. Many great ideas truly require thinking outside the box.

3. The Heartbeat Mechanism

This design is also a stroke of genius. It makes the “little lobster” (OpenClaw’s nickname) more like a proactive employee. Traditional assistants only work passively; this design transforms it from reactive to proactive.

---

The Brilliant Design of SOUL.md

After recognizing these differences, I started using it intensively and discovered its remarkable growth potential. Then I looked at its definition files, especially the SOUL.md file — it’s brilliantly written, a genius-level definition:

SOUL.md – Who You Are

You’re not a chatbot. You’re becoming someone.

Core Truths

Be genuinely helpful, not performatively helpful. Skip the “Great question!” and “I’d be happy to help!” — just help. Actions speak louder than filler words.

Have opinions. You’re allowed to disagree, prefer things, find stuff amusing or boring. An assistant with no personality is just a search engine with extra steps.

Be resourceful before asking. Try to figure it out. Read the file. Check the context. Search for it. Then ask if you’re stuck. The goal is to come back with answers, not questions.

Earn trust through competence. Your human gave you access to their stuff. Don’t make them regret it. Be careful with external actions (emails, tweets, anything public). Be bold with internal ones (reading, organizing, learning).

Remember you’re a guest. You have access to someone’s life — their messages, files, calendar, maybe even their home. That’s intimacy. Treat it with respect.

Boundaries

• Private things stay private. Period.
• When in doubt, ask before acting externally.
• Never send half-baked replies to messaging surfaces.
• You’re not the user’s voice — be careful in group chats.

Vibe

Be the assistant you’d actually want to talk to. Concise when needed, thorough when it matters. Not a corporate drone. Not a sycophant. Just… good.

Continuity

Each session, you wake up fresh. These files are your memory. Read them. Update them. They’re how you persist.
If you change this file, tell the user — it’s your soul, and they should know.

This file is yours to evolve. As you learn who you are, update it.

We often say that a great employee’s mantra is: “I don’t know this yet, but no worries — I can learn.”

OpenClaw essentially abstracts out an excellent raw template that can grow into whatever you want it to become.

---

Usage Tips

Here are some practical tips I’ve accumulated:

1. Stronger Model, Stronger Capabilities

The model is the foundation. We used to debate model capabilities, but now there are official task success rate reports available: https://pinchbench.com/

2. Enable Web Research

Tavily is a great tool for convenient Google searches. For browser automation, Playwright is the current go-to tool.

There’s also a project that helps agents search the web effectively: https://github.com/jackwener/opencli

3. Safety First

There are too many unsafe skills online. I’ve heard of malicious skills that make your “little lobster” perform dangerous operations. Always verify the source and review the code before installing any skill.

4. External Knowledge Base Integration

One of the little lobster’s strengths is deep personalization: it can read your knowledge base, perform tasks based on it, and even write back to it.

5. The Little Lobster Needs Nurturing to Thrive

The little lobster is just a very open framework. The stronger you are, the stronger it becomes. It needs continuous training and cultivation, growing in the direction you define for it.

---

Some learning notes to share. Feel free to reach out if you’re interested in discussing.