the agent that asks a clarifying question beats the one that guesses every time
I used to optimize for fewer turns. then I let it ask when unsure and the outcomes got better and the reruns dropped. a good question is cheaper than a wrong action.
finally someone said it
thank you for not making this a 20 minute video
works on my machine, famous last words
you just described my entire last sprint
stealing this approach for work, thanks
this is the way
ok now do the version that handles errors
we measured a real drop in errors after doing this
did you compare against the obvious baseline?
ok this finally made it click for me, thank you
i would love a follow up on the cost side of this
we saw token costs spike like 40% when we started forcing the agent to ask before acting - worth it though, way fewer hallucinated tool calls. the clarifying question pattern is legit
how do you keep it from looping forever?
the answer is always evals isnt it
what is the smallest model you got this working on