Three months ago, I saw a post on a forum: "Day 60 of unemployment, savings running out. Can anyone help me start a side hustle?"
Hundreds of replies. Food delivery. Video editing. Content creation. All had one thing in common: you make money when you're working; you lose when you're sleeping.
Zhang Hao, a guy I met, was in exactly this position three months ago. Today, he runs an AI system he built himself—landing 4-5 freelance projects a month, earning over ¥12,000. And here's the key: this system works while he sleeps.
He's not some tech wizard. Three months ago, he didn't even know what GitHub was.
{
Starting Point: Knowing Nothing Except That AI Could Make Money
Zhang Hao used to work in real estate. The market tanked. Months without a single deal.
He saw a short video claiming "AI coding requires no programming skills." Half-skeptical, he gave it a shot. The first week was chaos—just setting up the environment took two days. By the second week, using Codex, he built a tiny web scraper that automatically searched for jobs matching his criteria and sent them to his WeChat.
"It was a mess of code, but it actually ran," he said. "At that moment I realized—this isn't a toy. This is leverage."
From then on, he learned something every night, tried something every day. When Codex hit a wall, he asked the community, then fed the answers back to Codex. After a month, he took on a small freelance project: an auto-reply bot for a pet store.
{
Month Two: From "It Works" to "It Pays"
The pet store project earned him ¥800. Not much—but Zhang Hao noticed a pattern. He'd write the core logic in Codex, then set up Hermes with scheduled tasks to run the program 24/7. The client woke up every morning to a report of all last night's auto-replies.
And here's the thing: the framework was reusable. The pet store's auto-reply, a tutoring center's class reminders, a small company's automated report generation—same core logic, different business parameters.
He took on three projects in the second month, each ¥1,500–2,000. Less than ¥5,000 total. But he saw something more important: once these programs were set up, they didn't need him. The AI ran in the background. He just checked in occasionally and communicated client requirements.
Month Three: The System Ran While He Slept
By month three, Zhang Hao had built his own "revenue pipeline":
Codex for writing code, quickly producing demos for clients. Hermes for scheduling, letting the Agent execute tasks on a timetable. A Kaihe AIBOX at home, plugged in and connected, running 7×24 without stopping.
He took on a mid-sized project: an inventory alert system for three chain bubble tea shops. Quoted ¥8,000.
How? Codex wrote the data collection and alert logic. Hermes set up scheduled runs at 8 AM and 5 PM daily, with results sent to the owner's WeChat. The whole system ran on Kaihe AIBOX—no cloud server needed, no worry about midnight disconnections.
This project took him about 30 hours. The rest of the time, the system ran on its own.
You sleep. Someone else does the work. That's what an Agent means.
It's Not That AI Is Powerful—It's That "24/7 Never Stops" Is Powerful
Zhang Hao isn't from a technical background. The tools he used—Codex, Hermes, WeChat—any college student could learn.
The difference: most people stop once the code runs. "The AI works" is good enough. He went one step further—making the AI not just run, but run itself, run on schedule, run when you're not there.
Codex lets you skip writing code manually. Hermes lets you skip clicking "run." Kaihe AIBOX lets you skip renting cloud servers, paying monthly fees, worrying about "how do I keep this running when my laptop is off."
Put these three together, and you have a digital employee working for you 24 hours a day. It costs less than a month of takeout.
Zhang Hao isn't exceptionally smart, nor exceptionally hardworking. He just thought one step further.
Want to learn more about Kaihe AIBOX? Contact us at [email protected]
KaiheAIBOX #Codex #Hermes #AIAutomation #AISideHustle #Agent #24x7
📖 Glossary
AI Box (also known as Agent Computer / Agent PC), is a dedicated local hardware device that runs AI Agents. Pre-installed with an AI agent management system, plug-and-play, running 24/7. Users can remotely command AI to work via Discord, Slack, Telegram, WhatsApp, and more.