Cluster common errors and exemplars across the submission batch
Packaged know-how that tells an agent how to do a job well.
You might say…
“The same misunderstanding is showing up in 18 different ways across 30 submissions. I need to see the pattern, not re-read every one.”
What it does
Reads across a reviewed batch of submissions and clusters them into recurring errors, shared misconceptions, and standout strong responses, producing a synthesis note.
Trigger: Use as the post-review sense-making step after a cohort batch is fully reviewed, to identify what's systematic before deciding how to address it.
I/O: Batch of reviewed submissions with notes or scores → clustered error patterns, shared misconceptions, and exemplar responses
Recognise the problem?
The primitives are the commodity part. The fastest next step is a conversation about composing them into something that works for you.
Start a conversation