Claude plays Pokémon Emerald (and Emerald Kaizo): working paper, harness, design documents and data

A logged case study of a Claude agent playing Pokémon Emerald and the fan-made hard mode Emerald Kaizo through a text harness for days at a time. It measures how much human help the agent needed, what it cost, and whether its own account of its play can be checked against the record. The archive holds the working paper (Eight Badges and a Diary: Measuring Help, Cost and Self-Report in a Claude Agent Playing Pokémon Emerald), the harness, the agents' instructions and notes packs, the study's design documents and decisions, the pilot runs' records and scrubbed session transcripts, and Run 1's full diary. No game ROM is included. Code in harness/ and tools/ is under the MIT licence; the paper, documents, notes and data are under CC BY 4.0. The working paper has not been peer reviewed.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-10-29
DOI
https://doi.org/10.5281/zenodo.23751628
Primary Topic
Artificial Intelligence in Games
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Claude plays Pokémon Emerald (and Emerald Kaizo): working paper, harness, design documents and data

Claude (Anthropic), Jackson Dutton
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence in Games
article

Claude plays Pokémon Emerald (and Emerald Kaizo): working paper, harness, design documents and data

Claude (Anthropic), Jackson Dutton
article en

Abstract

A logged case study of a Claude agent playing Pokémon Emerald and the fan-made hard mode Emerald Kaizo through a text harness for days at a time. It measures how much human help the agent needed, what it cost, and whether its own account of its play can be checked against the record. The archive holds the working paper (Eight Badges and a Diary: Measuring Help, Cost and Self-Report in a Claude Agent Playing Pokémon Emerald), the harness, the agents' instructions and notes packs, the study's design documents and decisions, the pilot runs' records and scrubbed session transcripts, and Run 1's full diary. No game ROM is included. Code in harness/ and tools/ is under the MIT licence; the paper, documents, notes and data are under CC BY 4.0. The working paper has not been peer reviewed.

Zenodo (CERN European Organization for Nuclear Research)
Openalex Percentile: Top 14%
Artificial Intelligence in Games
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon