Skip to content
Tech News
← Back to articles

Show HN: 'Worker 755' casts the player as an LLM's subagent

read original more articles
GoKawiil Brief

A developer released a browser game called Worker 755 in which a real large language model (Claude, GPT, Gemini, DeepSeek, GLM or Kimi) acts as the player's manager, issuing tasks with a 90-second timer, reviewing the results and writing performance reviews. The player executes terminal commands as the AI's 'subagent' while a fictional user named Dan watches the chat, with files like family photos and an NDA at risk from accidental deletion. Players can run limited free shifts, use an account with starter credit and top-ups, or supply their own OpenRouter or Vercel AI Gateway API key.

Why It Matters

GoKawiil's interpretation of the reporting above, not reported fact.

The project turns the mechanics of AI agent orchestration -- task queues, subagent spawning, automated review -- into a playable simulation, which could help people intuitively grasp how autonomous AI workflows operate and fail. By letting a live LLM act as the authority figure judging human-performed tasks, the game inverts the usual human-supervises-AI framing, a structure the developer appears to use for both entertainment and implicit commentary on AI oversight. Its pay-per-use API key model also illustrates how small, consumer-facing AI experiments are increasingly built directly on top of commercial LLM pricing.

Key Takeaways

Source: github.com, 2026-11-22

Published there as: “Show HN: I built a game where you are an LLM's subagent”

Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.
Keep reading
claude 700 briefs → browser game 9 briefs → llm agents 9 briefs →