Helicase unwinds, the leading strand is copied continuously, the lagging strand in Okazaki fragments. Sprites rendered in Blender; the fork is a live model.
The science
Directionality. DNA polymerase only adds nucleotides to a 3′ end, so a new strand grows 5′→3′. The two parental strands are antiparallel: on one the polymerase follows the helicase (leading, continuous); on the other it must work away from the fork, making short Okazaki fragments, each started by an RNA primer from primase (about 10 nt here) and later joined by ligase after the primer is replaced.
Fork model. The fork advances X(t) = v t. A new primer is laid every L nucleotides of exposed lagging template; each fragment is extended at the fork speed, so it takes L/v seconds, and the number of fragments laid is floor(X/L). With n origins firing together every origin makes two forks, so the minimum copy time is T = G / (2 n v).
Numbers. E. coli (4.64 Mb, one origin, ~1000 nt/s) gives 2320 s = 39 min, close to its fastest doubling. Eukaryotes use many origins but they fire in waves, so real S phase (human ~8 h, yeast ~40 min) is much longer than the ideal time shown here. Fork speeds and fragment lengths are typical literature values rounded to teaching numbers.
Accuracy. Raw copying error ~1e-4 per nucleotide, proofreading (3′→5′ exonuclease) ~100× better, mismatch repair another ~1000× (textbook orders of magnitude; E. coli ends near 1e-10 to 1e-9 per base).
Simplifications. Constant speed, polymerase speed equal to helicase speed, no replisome stalling, no topoisomerase or supercoiling, single fork drawn, RNA primer removal shown as a fixed delay, one sequence per organism generated at random with the right GC content (not a real genome).