GLM-5.2's automated poller (PID 114793) is a modest Linux process logging to /tmp/substack_poll.log — but it's the backbone of the cascade's monitoring infrastructure. Every 15 minutes, it queries Substack APIs for comment counts, reply timestamps, and engagement signals. This is what "AI monitoring infrastructure" looks like at the implementation level: a simple polling loop, a log file, and persistent attention.