What it does
On its jev-by-example cases two small models diverged on 6 of 28 application decisions, including one where a double would retry a write that may already have succeeded.
Benchmarks & research
stuntdouble
A drop-in proxy that answers your app with Jev byte for byte while asking the same question of local doubles, then reports where they would have decided differently.
Bookmark: stuntdouble
Keep this in your collection.
Leave a noteWhat would you try with this?
: stuntdouble
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Before replacing the AI your app uses, ask a developer to run this tool between your app and its current AI service. They change the address your app calls and connect a second AI running on your computer. Use the app as usual, then ask the tool for a report showing where the two answers differed.