1 min lesson
Two versions, always
Match each case in "Two versions, always" to the signal and response that fit it.
Step 1 of 2
Two versions, alwaysthe speed read is part of the test
Cursor cares about running good-enough research fast. Show that you can dial rigor up and down on purpose by presenting the same question at two scopes.
Pull existing accept/revert telemetry on multi-file Agent edits today
5 moderated sessions with developers already in the funnel
One-page readout with a directional recommendation and a flagged confidence level
Goal: unblock a roadmap call this week, not settle the question forever
Behavioral analysis across both segments with a clean denominator
12–16 moderated sessions to saturation on the why
A short in-product survey to size how widespread the trust gap is
Triangulated readout: where the three signals agree and where they fight
The scrappy version is not the rigorous version with steps deleted. It is a deliberately different study with a different claim strength. Say out loud what you are choosing not to learn in 48 hours and name the risk you are accepting by skipping it.