agent-limit-retry is a Claude Code plugin that manages usage limits by tracking reset times, handing off work before limits hit, and automatically resuming interrupted tasks after resets. It works locally without proxies or telemetry, handling both interactive sessions and headless jobs across 5-hour and weekly quota windows.
TailStory, a pet health record app built by two people, runs on €60–200/month. The author reveals that AI inference is cheap (a tenth of a cent per call), and the real costs come from storage, databases, and idle servers—not from AI itself. Four key decisions kept costs down: combining operations, filtering with code before models, caching non-real-time data, and per-user rate limits instead of global ones.