Vaibify Documentation

Vibe boldly. Verify everything.

AI coding agents can write functioning scientific code in minutes, but their mistakes can harm a researcher’s computer and invite skepticism of AI-assisted results. Vaibify addresses both concerns. It builds a secure environment, usually a Docker container, connected to a graphical dashboard where researchers prompt command-line coding agents, decompose an analysis into Steps, watch the status of every local and remote file, inspect figures, and approve each result by hand.

Vaibify also guides a Project up the first three rungs of the PROOF Ladder, a six-level scale that classifies AI-assisted numerical experiments by their provenance, reproducibility, openness, oversight, and falsifiability. The veracity of a scientific result remains the researcher’s responsibility. Vaibify ensures that the digital artifacts were generated in the correct order, that local files match their remote copies, and that the results can be reproduced byte for byte.

New users should start with the QuickStart.

• Code of Conduct
• QuickStart
• Advanced Installation
• Vibe Coding Scientific Software
• The PROOF Ladder
• Declaring AI Use
• Environments and Projects
• Connecting to External Resources
• Reproducibility
• Agent Council
• System Prompt
• Security Model
• Testing Model
• CLI Reference
• For Agents
• GitHub

Citing vaibify

If you use vaibify in your research, please cite the paper that describes it:

TBD: the citation will appear here once the paper is published.