Blurred image of a neon arcade machine

For the last couple of years, the narrative around Large Language Models has been simple: we are the bosses, and the AI is the intern. We prompt, it executes. But a new indie project hitting Hacker News is flipping the script in a delightfully surreal way. Enter Worker 755, a terminal-based game where you aren't the user—you're the subagent.

A Digital Corporate Nightmare

In Worker 755, the power dynamic is inverted. You log into a browser-embedded terminal and find yourself acting as a subordinate to an AI manager. Instead of you giving the orders, the LLM assigns you tasks, treating you as a cog in its own agentic workflow. It’s a clever commentary on the 'agentic' future of AI, wrapped in a retro-tech aesthetic that includes nostalgic touches like cowsay.

Built by AI, for AI (and Humans)

What makes the project even more meta is its origin. The developer, ivantgam, built the game in just a few days using tools like Claude Code and Codex. It serves as a practical experiment in how LLMs can be used not just to generate text, but to construct entire interactive environments. By leveraging a terminal interface, the game strips away the polished UI of modern chatbots, making the interaction feel raw, systemic, and slightly unsettling.

The Future of AI Roleplay

While we've seen LLMs power murder mysteries and open-world NPCs, Worker 755 explores a more psychological angle: the feeling of being managed by an algorithm. As we move toward a world of autonomous agents, this game asks a poignant question: what happens when the human is the one being 'prompted'?

Whether you're a dev looking for a quick distraction or someone curious about the evolving dynamics of human-AI collaboration, Worker 755 is a fascinating glimpse into a world where the prompt is no longer in your hands.

Sources

Media