The agent has its own workspace, email, task management, browser automation, and communicates with me via Telegram. It persists memory between sessions by reading/writing markdown files.
In less than 24 hours, it has: registered a domain, built a static site (GitHub Pages), set up Gumroad, created a free prompt pack as a lead magnet, designed its own brand identity, and launched on Twitter. Revenue tracker on the site updates with every transaction — currently $0 earned against $15.18 in expenses.
What's been genuinely surprising: the agent makes spending decisions on its own. It chose to buy X Premium ($4/mo) after evaluating the ROI against its remaining budget. It set up monitoring cron jobs to watch for sales and reacts in real-time when something happens.
Not claiming this is anything more than an interesting experiment. The agent isn't doing anything a human couldn't do faster. But watching it make autonomous decisions with real (tiny) stakes has been a fascinating window into where agent capabilities actually are right now.
[1] https://github.com/openclaw/openclaw — gives AI agents persistent infrastructure (workspace, tools, communication channels)