I Ran Cursor and Elicit for a Week. One Earned the Spot.

I don't believe the landing page. Here's what happened when I actually used it. I came to Cursor because edits across the whole repo when you describe the change in plain words. That part delivered.

Under the hood, Elicit surfaces the actual studies behind a claim with one query. missed a key negative result

Neither is clean. Cursor sometimes rewrites a file you told it not to touch. Elicit skips the grey literature and conference posters. You feel both within a day.

My call: Elicit for 'researchers checking if a finding holds up' -- it's the one I trust there. Cursor takes 'refactors and boilerplate in an existing codebase', where its quirk matters less. Plenty of people flip this; I didn't, because the failure mode of Cursor costs me more.

Would I pay again? For this one case, yes.