A 5,000-token prompt that turns GPT-4 into a personal tutor
Mr. Ranedeer is a carefully engineered system prompt, not code, that wrangles a large language model into structured teaching sessions with configurable depth and style.

What it does
Mr. Ranedeer is a single text prompt you paste into ChatGPT (or Claude) to make the model behave like a personalized tutor. It supports slash commands—/plan, /test, /config, /start, /continue, /language—that let you generate lesson plans, take quizzes, adjust settings, and resume cut-off outputs. The prompt itself is the product: there is no software to install, no API to call, no repository of code.
The interesting bit The prompt encodes a full pedagogical framework. You configure “depth” on a 10-level scale from elementary school to Ph.D, pick learning styles (visual, verbal, socratic, etc.), tone, reasoning framework, and language. The author even versioned the prompt like software—v2.7 clocks in at 5,376 tokens—and maintains a “reboot” branch when OpenAI’s model updates degrade behavior. It is essentially prompt engineering as product design.
Key highlights
- Requires ChatGPT Plus with GPT-4 Code Interpreter; explicitly warns that GPT-3.5 “will not be as effective and concise”
- Also tested on Claude-100k, suggesting the prompt is model-agnostic enough to transplant
- Includes a separate “Config Wizard” GPT for users who do not know what settings they want
- Maintains archived previous versions because “the quality of outputs may vary depending on how OpenAI updates GPT-4”
- Supports any language GPT-4 can generate, with a disclaimer that translation quality is unreliable
Caveats
- Discontinued: the README leads with this label, though the prompt and ChatGPT link still appear functional
- Output quality is explicitly unstable and tied to OpenAI’s opaque model updates; the author admits recent versions may be “degraded”
- The
/searchcommand requires plugins that are not detailed or guaranteed to persist
Verdict Worth a look if you are studying how to coerce consistent, structured behavior from large language models through prompt design alone. Not useful if you want a maintained codebase, an API, or any guarantee the experience works tomorrow.