I study within-day behavioral responses to time-varying congestion pricing using New York City's Congestion Relief Zone (CRZ), implemented on January 5, 2025. Exploiting the CRZ's geographic boundary together with sharp, pre-announced toll schedule thresholds, I estimate causal effects using a difference-in-discontinuities (DiD-RD) design applied to high-frequency mobility data on hourly visits to points of interest (POIs). Across the four primary time boundaries implied by the toll schedule, I find evidence of temporal substitution concentrated at a single threshold: at the weekday 5 a.m. transition into peak pricing, visits during the discounted overnight window (midnight-4 a.m.) increase by approximately 5.9 percent relative to post-5 a.m. visits in the CRZ, compared with a control city and the pre-policy period. In contrast, estimates at the weekday evening boundary and both weekend boundaries are small and precisely near zero. Placebo tests using earlier implementation dates yield no detectable effects, and shifting the cutoff by one hour weakens the weekday-morning estimate, suggesting that re-timing is concentrated at this boundary.