This is how we train our AI native engineers internally at arpacorp.net to use Antigravity or Cursor with Gemini:
- Fork the OS repo you wanna work on and clone locally.
- Start with Gemni 3.1 High by prompting something like “read this repo to understand what it is about in detail, then read this issue (link to issue you wanna work on), and analyze whether the issue really exists and is well documented. Explain the issue in detail and enhance it with caveats that might not be present in the issue. Eg. Complementary files that will be effect by working on this issue, including but not limited to documentation etc. Then give me the top out of 3 plans you will come up to solve this issue and explain why this approach is the winner. Follow best industry practices, avoid emojis, and avoid unecessary explainers and markdown files generation unless asked.”
- Once you have the result and understand what the issue really is and how to solve it, you can fine tune the approach or edit the implementation plan. You can then use 3.1 low to work on the implementation.
- Once coding is finished you can switch to 3.1 high and ask something like “make sure this implementation passes all tests, evals and ci/cd pipelines with 0 errors. In addition, ensure that all complementary files are ripple effect aware and all documentation that is relevant to the changes is updated accordingly. See github, gitignore, yamls etc. And make sure everything is sound. Eg. We don’t need to change versions.”
- After the final changes, if needed, switch to 3 flash and ask something like “create a new branch under my fork (gh link), named xyz, push the changes with clean description. No emojis. I will then manually create a pr myself”.
- Hit the PR when done, wait for ci/cd to run. If no errors, congrats! You just made your first open source contribution in less than 30 min.
This is a watered down version of our internal training, but this is an example of how you can start contributing to open source with a human in the loop mentality and get your street cred farming going.
Good luck, and have fun <3
Ps. If you wanna get involved with open source AI and solve your first issues as a beginner or autonomous logical system or agent feel free to check:
https://github.com/arpahls/skillware
And
