RelayDifferent AI models make different mistakes. When an answer matters (a plan, a piece of code, an important email), asking a second model to review it is a cheap way to catch what the first one missed.
Step by step
Bring the whole picture
The original question, the constraints you gave and the answer to check. Relay carries these over in one go.
Ask for a review, not a rewrite
Ask a pointed question instead of "thoughts?", or you'll often get a polished rewrite that agrees with the original.
Compare and decide
Where the two disagree is the useful part. Take it back to the first chat, or decide yourself.
Doing it with Relay
Pick the next AIClick Relay in the chat and choose where it goes.
Check the hand-offRead exactly what the other AI will get.
Carry on thereIt lands in the message box. You press send.
Review questions that work
- "What assumptions in this plan could be wrong?"
- "What would break if the input were ten times bigger?"
- "Which claims here would you want a source for?"
- "What's missing that someone in my role would expect?"
Don't treat agreement as proof
Two models agreeing doesn't make an answer right. Check anything important against a real source.