a = gpt5high in codex - one shot. no readme generated.

b = gpt5high in codex. JSON version of prompt - basically one shot. bash shell got caught up at very end on the npm run dev. tiny codex error, no big deal.

c = opus 4.1 in claude code - one shot. $2.94 nice readme.

d = opus 4.1 in claude code. JSON version of prompt - one shot. nice readme generated

natural language prompt:

i want to create a sleek modern website that uses AI to generate a choose your own adventure story for users, with pictures generated by google imagen. let's use google gemini for the storywriting as well. it is very important that there is an admin panel where we are able to adjust basically every parameter possible, and ALL prompts. we also need to be able to tweak how the story prompts are generated for the user. we should have the smallest, easiest database possible to store multiple images for each screen of the story, so that we can switch between landscape and portrait orientations. it should be incredibly minimal, similar to the LLM interfaces of 2025. the less buttons the user sees the better. we'll want a light and dark mode. the opening screen should ask the user what genre they want: military sci fi, space opera, or space tech thriller. we'll definitely want sleek quick animations, it should overall feel futuristic and clean. use gradients tastefully.

JSON prompt:

{
  "projectDetails": {
    "projectName": "AI-Powered Choose Your Own Adventure",
    "projectGoal": "To create a sleek, modern, and minimal web application that uses generative AI to create interactive choose-your-own-adventure stories for users, complete with AI-generated images.",
    "aesthetic": "Futuristic, clean, minimal, with tasteful gradients and sleek animations."
  },
  "coreFunctionality": {
    "storyGeneration": {
      "engine": "Google Gemini",
      "description": "The AI generates branching narratives based on user choices, creating a unique story for each playthrough."
    },
    "imageGeneration": {
      "engine": "Google Imagen",
      "description": "Each step of the story is accompanied by an AI-generated image that visually represents the scene."
    },
    "userInteraction": {
      "initialScreen": "User is prompted to select a genre from three options: 'Military Sci-Fi', 'Space Opera', or 'Space Tech Thriller'.",
      "gameplayLoop": "User reads the story text, views the image, and is presented with choices to advance the narrative. Each choice leads to a new AI-generated story segment and image."
    }
  },
  "designAndUX": {
    "inspiration": "Modern LLM interfaces of 2025.",
    "ui": {
      "layout": "Extremely minimalist. The primary focus must be on the story text and image.",
      "controls": "The number of visible buttons should be minimized. Interaction should be intuitive and seamless.",
      "colorScheme": {
        "modes": ["light", "dark"],
        "useOfGradients": "Subtle and tasteful to add depth and a futuristic feel."
      },
      "animations": "All transitions and animations must be sleek, quick, and smooth."
    }
  },
  "technicalRequirements": {
    "adminPanel": {
      "access": "Secure, administrator-only access.",
      "features": [
        {
          "name": "Global Parameter Control",
          "description": "A centralized dashboard to adjust every possible application parameter, including AI model settings (e.g., temperature, token limit), animation speeds, and UI elements."
        },
        {
          "name": "Comprehensive Prompt Management",
          "description": "Full CRUD (Create, Read, Update, Delete) functionality for ALL system prompts. This includes initial story prompts, image generation prompts, and the meta-prompts that define the AI's narrative style."
        },
        {
          "name": "Story Logic Configuration",
          "description": "Interface to tweak the logic for how user choices are translated into new prompts for the Gemini model, allowing fine-tuning of story coherence and creativity."
        }
      ]
    },
    "database": {
      "requirement": "The smallest, simplest database solution possible.",
      "schema": {
        "collection": "StoryScreens",
        "document": {
          "screenID": "Unique identifier",
          "storyText": "The generated text for this screen.",
          "images": {
            "portrait": "URL to the portrait-oriented image.",
            "landscape": "URL to the landscape-oriented image."
          },
          "userChoices": "An array of choice objects presented to the user."
        }
      }
    }
  }
}