Compare two versions of your AI feature

Paste up to 10 examples with an output from each version (two prompts, two models, before and after a change), choose how to score them, and see which examples got better and which got worse. No sign-up, no SDK. Demo runs are deleted after an hour.

1. Paste examples

Up to 10 examples.

Rows from a spreadsheet, CSV, or a JSON array

First line is the header: input, output_a, output_b, and optionally expected. Copying cells from Google Sheets or Excel pastes tabs, which works.

Name for output A

Name for output B

Experiment name (optional)

2. Choose how to score

Each check runs on output A and output B for every example.