Open-source research in robotics, simulation and embodied AI.
We reproduce published robotics research and publish what we actually measured — including the parts that did not work.
An open-source contribution program. The work happens in public repositories under open-source licenses. Contributions are made under the repository's license: you keep the copyright to your work, and your name stays on the commit.
This is not employment, an internship, or a trial period. No salary, no assigned work, no deadlines, no minimum commitment. You choose what you work on; we review it the way any open-source maintainer would — technical feedback, nothing else.
New here? Read ONBOARDING.md — it answers "what do I actually do now?" in about five minutes.
Issues carry a level: label so you can find a starting point. Nothing is assigned to you.
🌱 level:explorer |
Run something, read something, write down what happened. No robotics or ML background needed. |
🔧 level:contributor |
A small pull request — a fix, a test, a tooling improvement. |
🧪 level:research |
Reproduce a published result, or investigate why it cannot be reproduced. |
You do not need your own GPU. Our current study runs on CPU.
- reproduction-001 — Can the published 65.4% PushT result be reproduced today? Open questions, open issues.
- studies — Write-ups of the studies published at forenly.ai/lab, with their numbers and their limitations.