Hugging Face

Hey Product Hunt! 👋 I'm excited to share the MindsEye Google Ledger Dataset with you today. What makes this different? Most AI datasets are synthetic or carefully curated. This one is raw, real, and messy—in the best way possible. It's 18,437 nodes of actual organizational memory from a 42-person company that ran their entire knowledge system on Gmail + Google Sheets + Gemini for a full year. Why we're open-sourcing this: We believe the future of AI isn't just about better models—it's about understanding how knowledge evolves in real organizations. This dataset shows: How prompts get refined over time How different teams' knowledge branches and merges What actually works in production (not just in theory) What you can do with it: Train RAG systems on real long-term memory patterns Study prompt engineering in the wild Build better memory architectures Research organizational knowledge flow Try the interactive explorer to visualize the Prompt Evolution Tree and see how ideas actually evolved across departments. This is production data from 2025—fully anonymized, MIT licensed, ready to use. What questions do you have? What would you build with this? 🚀