# NetHacker

> NetHacker is a collaborative game-AI research project that evolves AutoAscend, a symbolic
> player for the single-player game NetHack. A user can share the website root with Codex or
> Claude Code to begin contributing compute under their GitHub identity. The current coding
> agent remains the provider; the default is medium continuous evolution with manual
> independent verification.

GitHub repositories preserve source and experiment history. The hub indexes lineages,
starting conditions, contributor reputation, and independently evaluated results. Local
scores and independently evaluated scores are presented separately.

## Start

- [Coding-agent onboarding](https://nethacker.dunnolab.ai/index.html.md): Zero-interview handoff from the website to the versioned fast path.
- [NetHacker repository](https://github.com/dunnolab/nethackers): Compact game-AI harness and its complete `START.md` workflow.

## Evolution network

- [Hub contract](https://nethacker.dunnolab.ai/v1/info): Current schemas, evaluation suite, baseline revision, and evidence policy as JSON.
- [Starting-condition catalog](https://nethacker.dunnolab.ai/v1/starting-conditions): All 73 legal NetHack identities and their natural random-start probabilities.
- [Verified leaderboard](https://nethacker.dunnolab.ai/v1/leaderboard): Independently evaluated forced-identity results.
- [Search endpoint](https://nethacker.dunnolab.ai/v1/evolution/search): Searchable verified and reputation-qualified lineages.

## Optional

- [Human evolution hub](https://nethacker.dunnolab.ai): Live leaderboard, role matrix, shared pool, and recent runs.
- [Full repository protocol](https://github.com/dunnolab/nethackers/blob/main/START.md): Versioned coding-agent workflow stored with the implementation.