Lab note 01
Autonomous AI Agents
Experimenting with OpenClaw and Hermes to create agents that can interact with tools, execute workflows and operate within controlled environments.
What I am exploring
How far an agent can go on its own when it has real tools: a shell, a browser, files and APIs. I run OpenClaw and Hermes and give them practical work rather than demos.
What I have learned
Agents are most useful with clear limits. Approval steps for risky actions, credentials kept out of the conversation and checks against real output matter more than a larger model. Reusable procedures that the agent writes down after a task make the next run faster and more consistent.