Use Grok as the narrative engine of a game
The poster built a game where Grok generates the story narration, and says it outperformed Gemini and GPT for narrative output.
- What they did
- The poster says they built a text-based game where players make choices, roll dice, and face consequences, while Grok acts as the main narrator behind the scenes. According to the post, Grok keeps the game state on track and writes the story the player reads; the human built the game and framed it as a structured demo on their profile.
- How it went
- They report Grok worked much better than Gemini and GPT for narrative output, crediting both its writing quality and fewer refusals around player actions. The thread gives no hard metrics, implementation details, or measured comparison results.
- Worth knowing
- The useful catch is that the post is only a high-level claim and demo pointer, so you would need the profile-linked project for actual build details.
Try it yourself with Grok Bot
Act as the narrator of a [genre] text adventure set in [setting]. Describe each scene vividly in 2 to 3 paragraphs, keep track of my inventory, choices, and the story so far, and end each turn by asking what I do next. Begin with the opening scene and wait for my first action.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · May 11, 2026
Related use cases
Orchestrate a multi-phase RAG migration
A walkthrough of using Gemini CLI to coordinate a complex, multi-phase cloud migration project.
Buy products online from a card photo
Musk said Grok Bot bought LEGO from the LEGO website, and that users can drop a credit card photo into chat so it finds the best deal and orders; plans and safeguards weren't specified.
Produce a sourced research document with parallel agents
Grok coordinated 16 specialized agents that searched and parsed 896 sources, debated and peer reviewed, and delivered a complete research document in one minute.
Play Doom with computer-use agents as an informal benchmark
The OpenClaw team used computer-use agents to play Doom on remote machines, treating game performance as an informal way to improve the computer-use workflow.
Test an iPhone game and report what passed
Cognition had Devin on GPT-6 Astra test the game Otter Run in a simulator, returning a recording and a report of checks passed and untested areas.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.