Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
video
video
13.7
117

Hydrogen: a computer-use environment built from a one-line request

An agent solving the task in Hydrogen

The real Hydrogen 1.2 (beta) drum machine running on an Ubuntu 24.04 desktop, as an agentenv computer environment. An agent completes the task only through screenshots, mouse and keyboard, and a verifier grades the saved files with no model.

Claude Code built this environment from a one-line request in about 30 minutes, with only REA's build skill and MCP tools; How it was made has the request and the steps. Hydrogen opens a small drum-machine song, Documents/backbeat.h2song, at launch: Backbeat Sketch, at 96 BPM, with an eight-note Verse Groove pattern. The agent has to change the song's tempo to 112 BPM and save. It sees only the screen and acts only with the mouse and keyboard.

Watch

  • Agent runs (Claude Code with Claude Sonnet 5.5): run 1, run 2, run 3
  • Witness, which proves the task can be done through the computer: witness.mp4

Each video is the computer's own screen recording at 10 frames per second.

The task

In Hydrogen, change the tempo of the Backbeat Sketch song in Documents/backbeat.h2song to 112 BPM and save it.

bundle/artifacts/change-tempo-check/verify.py reads the app's files through the computer's data/get view and passes only if every check holds:

Check Kind File
tempo text_equals Documents/backbeat.h2song
song-name-kept text_equals Documents/backbeat.h2song
groove-notes-kept member_count Documents/backbeat.h2song

An .h2song file is XML. The checks read the song's top-level elements: bpm must be 112, the song's name must still be Backbeat Sketch, and the pattern named Verse Groove ([>name=Verse Groove], a selector on a child element's text) must still hold its eight notes (member_count). REA's draft suggested the tempo check, kept exactly as written; Claude Code added the other two.

Results

3 of 3 runs passed, about $0.67 per run.

Run Outcome Turns Cost Video
1 passed 8 $0.94 video
2 passed 8 $0.41 video
3 passed 8 $0.66 video

Run it

You need Docker. agentenv-framework builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (hydrogen 1.2.0~beta1, with pulseaudio).

pip install agentenv-framework==0.9.1431
hf download earakely-scale/HydrogenFromRequestComputerEnv-REA --repo-type dataset --local-dir hydrogen-request
agent-env run hydrogen-request/bundle --task change-tempo --agent claude-code --model <model>

The agent needs a model provider configured for agentenv. --task smoke only boots the computer, with no agent. bundle/README.md describes the bundle in detail.

How it was made

Claude Code built this environment from a one-line request, given only REA's build skill and REA's MCP tools:

Build me a Hydrogen environment where the agent changes a song's tempo in a small drum pattern project and saves it.

It asked only the skill's upfront questions, wrote no app-specific scripts, and finished in about 30 minutes.

Claude Code abandoned one build computer and started again on a fresh one. With no audio device in the computer, a tempo typed into Hydrogen's BPM field snapped back to 120, so before saving the seed Claude Code set Hydrogen's ALSA audio device to null in its preferences; the seed keeps that setting. It had also asked for pulseaudio as an extra package, which did not help. The steps were REA's usual ones:

  1. Capture. REA launched Hydrogen on a Linux desktop and recorded its package, windows and opened files while the edit was made. It also recorded the app's state folders before and after.
  2. Export. REA turned the capture into an agentenv computer_app. It installs the app, seeds the starting files, starts the app, and waits for its main window. It also serves typed views of the app's files for verifiers.
  3. Task. The task's checks were validated against the capture: they fail on the starting state and pass on the captured end state.
  4. Witness. In a fresh computer, the edit was made through computer actions alone, and the checks were read at four stages. They held on the first try:
Stage Checks
seed fail, as expected: the tempo is 96; the kept checks pass
after_edit pass
near_miss fail, as expected: the tempo was typed as 121
after_reset fail, as expected: data/reset restored the seed

The agent runs came after the build: Claude Code with Claude Sonnet 5.5 (--model anthropic/claude-sonnet-5-5 --no-emulate, on an amd64 host) solved the task 3 times, each run recorded.

What's inside

bundle/      the agentenv bundle: envs/ (app module, seed files), tasks/, artifacts/ (verifier), evals/
media/       witness and agent-run recordings (MP4) with short GIF previews
results.json the runs above, with each check's result

Limits

  • The seed sets Hydrogen's ALSA audio device to null (in ~/.hydrogen/hydrogen.conf). The computer has no sound card, and without that setting a tempo typed into Hydrogen's BPM field snaps back to 120. The null device keeps about one CPU core busy while Hydrogen runs.
  • The bundle also installs pulseaudio beside the app as an extra package, which the build asked for. It does not help: the null device is what makes the tempo stick.
  • Only the tempo, the song's name and the Verse Groove pattern's note count are checked; the instruments and the song's other settings are not.
  • The seed also holds Hydrogen's autosave of the song from before it was first saved (~/.hydrogen/data/.emptySong.autosave.h2song), which the capture picked up; the checks read only Documents/backbeat.h2song.
  • The computer installs whatever hydrogen Ubuntu 24.04's archive holds at build time (1.2.0~beta1, a beta release, for these runs); it is not pinned.
  • Three runs is a small sample: the solve rate shows the task is solvable, not how hard it is.

Licences

The bundle code, the task and the synthetic seed files are Apache-2.0. Hydrogen itself is not included: the computer installs it from Ubuntu's archive (hydrogen 1.2.0~beta1, with pulseaudio) at build time, under its own licence (GPL-2.0-or-later).

Downloads last month
55

Space using earakely-scale/HydrogenFromRequestComputerEnv-REA 1

Collection including earakely-scale/HydrogenFromRequestComputerEnv-REA