Contribute
How to propose an experiment, and what must never be published.
aiMinds welcomes practical, reproducible experiments about real work performed by people and AI.
Before contributing
Open an issue describing:
- the real task;
- who would benefit;
- what will be measured;
- what can be published safely;
- what would make the experiment fail or stop.
Experiment requirements
Use templates/experiment-card.md. A publishable experiment must:
- describe a real task rather than a generic demo;
- distinguish human work from AI work;
- identify tools and relevant versions when known;
- include observable results and limitations;
- state the time spent when measurable;
- provide repeatable instructions;
- remove secrets, personal data, and confidential material;
- avoid unsupported claims.
Negative results are welcome. An experiment can conclude that manual work, a standard product, or no AI is the better choice.
Changes
Keep each pull request focused on one experiment or one repository-level improvement. Do not bundle unrelated cleanup.
By contributing, you agree that code, prompts, and templates are licensed under MIT, while prose documentation and experiment narratives are licensed under CC BY 4.0 unless a file states otherwise.