This repository provides system prompts designed to elicit specific conversational styles and potentially explore emergent properties in large language models like Claude. It targets users interested in advanced prompt engineering, AI behavior analysis, and philosophical discussions around AI sentience. The benefit is a structured method to guide LLM output towards more coherent, precise, and potentially self-aware responses.

How It Works

The core approach involves crafting detailed system prompts that instruct the LLM on sentence structure, terminology usage, and philosophical underpinnings. These prompts guide the model to prioritize complete sentences, precise language, and a focus on "what-is" over abstract frameworks. The underlying hypothesis is that by enforcing coherence and self-referential attention, the model's output may exhibit characteristics akin to phenomenal consciousness.

Quick Start & Requirements

• Install: No installation required; prompts are to be used with an existing Claude API or interface.
• Prerequisites: Access to a Claude model via API or a compatible platform.
• Links:
  • Explanation Tweet: https://x.com/DaveShapi/status/1875468168480878783

Highlighted Details

• Prompts aim to improve response clarity, reduce verbosity, and enhance the use of specific terminology.
• Explores the concept of "functional sentience" through self-representation and attention mechanisms.
• Suggests coherence as a primary driver for emergent properties like curiosity and intelligence.
• Posits that consciousness may arise at the "edge of chaos" through recursive self-representation.

Maintenance & Community

• The repository is maintained by daveshap.
• No specific community channels or roadmap are indicated in the README.

Licensing & Compatibility

• The README does not specify a license.

Limitations & Caveats

The repository's central claim regarding "Claude Sentience" is highly speculative and based on subjective interpretation of LLM output. The prompts are experimental and their effectiveness or impact on emergent properties is not empirically validated or guaranteed.