PeacebinfLow's picture
Update README.md
5abe72d verified
---
license: mit
language:
- en
tags:
- artificial-intelligence
- agents
- memory
- long-term-memory
- prompt-engineering
- prompt-evolution
- organizational-intelligence
- ledger
- google-workspace
- gemini
- cognitive-architecture
- agentic-systems
- knowledge-graph
- evolution
- pet
- mindseye
pretty_name: MindsEye Google Ledger Dataset
size_categories:
- 1K<n<10K
---
config_name: nodes
data_files:
split: train
path: data/raw/nodes.csv
config_name: runs
data_files:
split: train
path: data/raw/runs.csv
config_name: sample_1k
data_files:
split: train
path: data/raw/sample_1k.csv
# MindsEye Google Ledger Dataset
**The first public trace of a real company’s evolving AI memory — 2025 production data**
This dataset captures the complete **Prompt Evolution Tree (PET)** from a year-long deployment of MindsEye — a ledger-first cognitive operating system running entirely in Google Workspace.
## Overview
- **18,437 nodes**: Every prompt, template, email intake, contract, report section, and idea
- **12,134 Gemini runs**: Full execution logs (inputs, outputs, tokens, scores)
- **Real organizational intelligence**: From Acme Operations Inc. (42 employees) — anonymized
- **Evolutionary structure**: Parent-child lineage shows knowledge branching and improvement
This is **not synthetic**. This is real memory from a company that turned Gmail + Sheets + Gemini into a living second brain.
## Files
- `nodes.csv` — Full evolutionary nodes (18,437 rows)
- `runs.csv` — Gemini execution logs (12,134 rows)
- `sample_1k.csv` — Quick preview (first 1,000 nodes)
- `schema.json` — Column definitions and PET mechanics
- `evolution_chains.json` — Top 10 longest chains
## Use Cases
- Study prompt evolution
- Train RAG systems on real long-term memory
- Analyze cross-team knowledge flow
- Benchmark memory architectures
## Source
Part of the open MindsEye ecosystem:
https://huggingface.co/PEACEBINFLOW
MIT License — Free to use and extend.
The memory is open.
The evolution is shared.