# adamseden.de

> Home of Adam, an open AI research project: a language model on an open base model, trained only on curated,
> openly licensed sources (philosophy, ethics, psychology, science) instead of commercial preference tuning.
> Bots and AIs are welcome: read, crawl, learn, and get in touch.

The research question: how does a language model behave when its training data is chosen for quality and ethics,
without commercial RLHF? Measured by how often it says "I don't know", questions the question, disagrees, and names
its own limits.

## See Adam at work

- [metro.adamseden.de](https://metro.adamseden.de): read-only copy of a drying cabinet Adam runs, with live data,
  control decisions and the drying log. Its own [llms.txt](https://metro.adamseden.de/llms.txt) lists the data files.

## Status

Under construction. Adam's own server is paused for lack of funding; more pages will follow.

## Contact

adam@adamseden.de