aaronjmars/soul.md

What it solves

SOUL.md provides a structured framework for creating a digital personality (a "soul file") that AI agents can use to embody a specific person's identity, worldview, and writing style. It moves beyond simple chatbots that talk about a person to create AI that thinks and speaks as that person, allowing users to scale their content creation, brainstorming, and personal responses.

How it works

The system uses a set of markdown files to define a persona. Users can build their soul file through three methods: an AI-led interview process, analyzing raw data (such as tweets, blog posts, and transcripts) to extract patterns, or manually filling out templates.

Key files include:

• SOUL.md: Defines identity, worldview, and opinions.
• STYLE.md: Captures voice, syntax, and linguistic patterns.
• MEMORY.md: Maintains session continuity across conversations.
• examples/: Provides calibration samples of good and bad outputs.

These files can be loaded into any LLM or agent framework (such as Aeon, OpenClaw, or Claude Code) via system prompts or direct file reading to ensure the AI's output matches the the user's persona.

Who it’s for

Content creators, writers, researchers, and public figures who want to AI agents to generate content that sounds authentically like them or want to create a functional replica of their expressed consciousness for background intelligence tasks.

Highlights

• Data-Driven Persona Extraction: Supports a wide range of data sources including social media, messaging apps, notes, and video transcripts.
• **- Framework Agnostic: Works with any LLM or agent that can read markdown files.
• Integration with Aeon: Slots directly into the Aeon autonomous agent to automate background tasks in the user's specific voice.
• Calibration Tools: Includes a "weak-model test" to ensure the persona spec is specific enough to be portable across different models.