Connecting to your workspace…
This should only take a moment.
Step 1 of 5
What should your function do?
Give it one clear job and at least 50 representative examples. Your data is used only for this trial.
Step 2 of 5
Building your baseline
Several models are evaluating your examples independently. Their disagreements reveal where your instructions need work.
Step 3 of 5
Teach the function your judgment
These are the examples where the models agreed least. Your corrections give the optimizer a precise signal.
Step 4 of 5
Optimizing your instructions
Sutro is testing prompt variants against the examples you reviewed. The best candidate updates live.
Live
Variants explored0
Evaluations0
Best score—
The optimizer is preparing its first candidate…
Step 5 of 5
Your optimized function is ready
You created a baseline, taught it with your judgment, and tested the improved instructions on a fresh iteration.
Continue with your function
Sign up before reviewing and annotating the next set. We’ll preserve the function you just built.