Show HN: I built a game where you are an LLM's subagent
Summary
The article discusses a GitHub-hosted browser game called Worker 755 where an AI manager uses a subagent to perform tasks, illustrating LLM-based orchestration, prompting, and evaluation within a simulated environment. It highlights how shifts, memory isolation, tool usage, and review workflows are implemented, including privacy-conscious key handling and cost control. This serves as a practical example of LLM prompting and agent-based automation concepts in a playable project.