In a strange turn of events where the tool became the detective, a software developer discovered their computer had been hijacked by a cryptocurrency miner thanks to an unexpected alert from an AI coding assistant. The user, known online as UntilHEX, was utilizing Claude Code to assist in building a temperature monitoring application when the AI interrupted the workflow to report some alarming activity happening in the background of the system.
The AI agent pointed out that it had detected a process called XMRig running on the machine, explicitly noting that it hadn’t placed the software there itself. According to the logs shared by the agent, the malicious miner was aggressively consuming resources, hogging roughly twelve processor cores and hitting a staggering 1175 percent CPU usage. This massive drain on hardware explains why the laptop had become physically hot to the touch and sluggish during operation.
While most users rely on dedicated antivirus software to catch such threats, this instance highlights a new frontier where LLM-based agents can spot systemic anomalies through direct environment interaction. For the developer, who admitted their Mac had been performing poorly for some time, the revelation provided an immediate answer to why their device felt like it was melting under pressure.
The incident serves as a vivid reminder of how stealthy modern miners can be, often blending into system processes until they cause extreme thermal throttling. In this case, however, the very project being built—a temperature monitor—led to a discovery made not by code designed for security, but by an AI simply noticing that something wasn’t right with its host environment.
