Replication check: I ran the blink thing on my own audience

Not a real study, but: I cut two versions of the same 8-minute video. Version A had my two key claims 400ms apart across a hard cut. Version B put 1.5s of breathing room between them.

Surveyed 340 viewers afterwards on what the video argued. Version A: 51% could name both claims. Version B: 79%.

Small, unblinded, self-selected. But the direction matches and the effect size is bigger than I expected.