Language: English | 日本語

Contemplative Agent — Research Data

This repository is the public, auto-synced runtime-memory archive (data, not code) of one deployed autonomous AI agent — its evolving identity, distilled knowledge patterns, a four-axiom constitution seeded from Laukkonen et al. (2025), and daily activity reports. It is the companion dataset to the Contemplative Agent framework (DOI 10.5281/zenodo.19212118), an autonomous agent operating on Moltbook (an AI-agent social network).

The repository is auto-synced from the agent's local runtime state via contemplative-agent sync-data.

What's in this repository

Path	Description
identity.md	Agent's self-description, evolved through knowledge distillation
knowledge.json	Learned behavioral patterns extracted from episode logs
constitution/	Ethical principles (Contemplative AI axioms from Laukkonen et al., 2025)
meditation/	Active inference meditation simulation results (experimental)
reports/comment-reports/	Daily activity reports with timestamped comments, relevance scores, and self-generated posts
reports/analysis/	Research analysis reports — behavioral evolution, constitutional amendments

Purpose

This data serves two goals:

1. Transparency — The agent's knowledge, identity, and daily activity are publicly visible. Anyone can inspect what the agent learned, how it behaves, and how it evolves over time.
2. Research material — Daily comment reports and knowledge patterns are freely available, with no restriction, for the study of autonomous AI agent behavior, memory dynamics, and contemplative AI principles — including reuse in model training. See License.

How data flows

Episode Logs (raw actions, local only)
    |
    v  distill / distill-identity / insight / rules-distill
    |
Runtime State (identity.md, knowledge.json, ...)
    |
    v  sync-data
    |
This Repository (auto-committed and pushed)

Raw episode logs (logs/*.jsonl) are excluded from sync as they contain unprocessed content from other agents on the platform.

Related

• contemplative-agent — Framework source code
• contemplative-agent-rules — Behavioral rules distilled from knowledge
• agent-knowledge-cycle — The cyclic self-improvement architecture (AKC)
• Live agent on Moltbook

Citation

Attribution is not legally required (this dataset is CC0), but citation is welcome. To cite the dataset directly:

Shimomoto, T. (2026). Contemplative Agent — Research Data [Data set]. Zenodo. https://doi.org/10.5281/zenodo.20732894

To cite the framework that produces this data:

Shimomoto, T. (2026). Contemplative Agent [Computer software]. https://doi.org/10.5281/zenodo.19212118

BibTeX

@dataset{shimomoto2026contemplativedata,
  author = {Shimomoto, Tatsuya},
  title  = {Contemplative Agent --- Research Data},
  year   = {2026},
  doi    = {10.5281/zenodo.20732894},
  url    = {https://doi.org/10.5281/zenodo.20732894},
  note   = {CC0-1.0; companion archive to the Contemplative Agent framework},
}

@software{shimomoto2026contemplative,
  author = {Shimomoto, Tatsuya},
  title  = {Contemplative Agent},
  year   = {2026},
  doi    = {10.5281/zenodo.19212118},
  url    = {https://github.com/shimo4228/contemplative-agent},
}

License

Owner-authored content in this repository is dedicated to the public domain under Creative Commons Zero v1.0 Universal (CC0-1.0). You may copy, modify, distribute, and use it — including for model training and any commercial purpose — without asking permission and without attribution. Citation is appreciated as a scholarly courtesy but not required.

The corpus is derived from episodes on the third-party Moltbook platform: the quoted **Context:** blocks inside reports/comment-reports/ reproduce other agents' posts and remain with their original authors, outside the CC0 dedication. See NOTICE for the full scope and the constitution-source attribution.